diff --git a/.flake8 b/.flake8 new file mode 100644 index 00000000..1170053a --- /dev/null +++ b/.flake8 @@ -0,0 +1,8 @@ +[flake8] +exclude = + .git, + __pycache__, + polytracker/src/compiler-rt, + examples, + the_klondike, + third_party diff --git a/.github/workflows/dockerimage.yml b/.github/workflows/dockerimage.yml index 31b72060..8ed2a2ad 100644 --- a/.github/workflows/dockerimage.yml +++ b/.github/workflows/dockerimage.yml @@ -11,3 +11,5 @@ jobs: submodules: recursive - name: Build the base image run: docker build . --file Dockerfile --tag trailofbits/polytracker --no-cache + - name: Run the PolyTracker tests + run: docker run --rm --workdir /polytracker trailofbits/polytracker pytest diff --git a/.github/workflows/python.yml b/.github/workflows/python.yml index 2dd78c8d..65e33ff3 100644 --- a/.github/workflows/python.yml +++ b/.github/workflows/python.yml @@ -39,5 +39,3 @@ jobs: - name: MyPy run: | mypy --python-version ${{ matrix.python-version }} --ignore-missing-imports polytracker tests build_in_docker eval --exclude compiler-rt - - name: PolyTracker tests - run: docker run --rm trailofbits/polytracker pytest \ No newline at end of file diff --git a/examples/Dockerfile-acropalypse.demo b/examples/Dockerfile-acropalypse.demo new file mode 100644 index 00000000..3c55f6e9 --- /dev/null +++ b/examples/Dockerfile-acropalypse.demo @@ -0,0 +1,28 @@ + +FROM trailofbits/polytracker +LABEL org.opencontainers.image.authors="henrik.brodin@trailofbits.com" + +WORKDIR /polytracker/acropalypse + + +RUN curl -L https://downloads.sourceforge.net/libpng/libpng-1.6.39.tar.xz --output libpng.tar.xz +RUN tar xvf libpng.tar.xz +WORKDIR /polytracker/acropalypse/libpng-1.6.39 + +RUN curl -L https://zlib.net/zlib-1.2.13.tar.xz --output zlib.tar.xz +RUN tar xf zlib.tar.xz + +WORKDIR /polytracker/acropalypse/libpng-1.6.39/zlib-1.2.13/ +RUN polytracker build ./configure +RUN polytracker build make +RUN polytracker extract-bc -o ../libz.bc libz.a + +WORKDIR /polytracker/acropalypse/libpng-1.6.39 + +RUN CPPFLAGS="-I$(pwd)/zlib-1.2.13/include" LDFLAGS="-L$(pwd)/zlib-1.2.13/lib" polytracker build ./configure --disable-shared +RUN CPPFLAGS="-I$(pwd)/zlib-1.2.13/include" LDFLAGS="-L$(pwd)/zlib-1.2.13/lib" polytracker build make pngtest +RUN polytracker extract-bc -o pngtest.bc pngtest + +RUN llvm-link -o pngtest-linked.bc pngtest.bc libz.bc +RUN polytracker instrument-bc --taint --ftrace pngtest-linked.bc -o instrumented.bc +RUN polytracker lower-bc instrumented.bc -t pngtest -o pngtest.instrumented diff --git a/examples/analysis/README.md b/examples/analysis/README.md index 4a50f89e..03308126 100644 --- a/examples/analysis/README.md +++ b/examples/analysis/README.md @@ -1,7 +1,10 @@ # Analysis Scripts -## last updated Nov 22, kelly.kaoudis@trailofbits.com +## last updated May 2023, kelly.kaoudis@trailofbits.com -Small scripts and other assorted tooling which might be copied into containers or run in the native working environment to automate learning about how Polytracker works. +### analysis/nitf/ +Some early experiments with NITF parser exploration that eventually resulted in the paper referenced below. + +### analysis/ubet/ +Tooling, examples, and other artifacts of the LangSec '23 paper [Automatically Detecting Variability Bugs Through Hybrid Control and Data Flow Analysis](https://langsec.org/spw23/papers.html#variability) that are not yet integrated into main PolyTracker. This tooling is generally referenced from Dockerfiles in the parent directory. -The point of these is reproduceability and knowledge sharing. \ No newline at end of file diff --git a/examples/analysis/ubet/Dockerfile b/examples/analysis/ubet/Dockerfile new file mode 100644 index 00000000..ac4b199a --- /dev/null +++ b/examples/analysis/ubet/Dockerfile @@ -0,0 +1,10 @@ +FROM ubuntu:jammy +ENV DEBIAN_FRONTEND=noninteractive + +# We just need the PolyTracker python analysis code to run eval_nitro.py, and not the PolyTracker llvm environment. +RUN apt-get update && \ + apt-get upgrade -y && \ + apt-get install -y clang g++ python3 python3-pip && \ + pip3 install polytracker + +WORKDIR /polytracker/the_klondike/nitro/build/ubet \ No newline at end of file diff --git a/examples/analysis/ubet/Dockerfile.nitro b/examples/analysis/ubet/Dockerfile.nitro new file mode 100644 index 00000000..3549b341 --- /dev/null +++ b/examples/analysis/ubet/Dockerfile.nitro @@ -0,0 +1,65 @@ +FROM trailofbits/polytracker:latest +ENV DEBIAN_FRONTEND=noninteractive +LABEL org.opencontainers.image.authors="kelly.kaoudis@trailofbits.com, henrik.brodin@trailofbits.com" +WORKDIR /polytracker/the_klondike + +# kaoudis, May 2023: Nitro (or more likely, coda-oss) has done something weird +# and Nitro no longer can find PYTHON_HOME. Using ENABLE_PYTHON=OFF for now +# so that this Dockerfile at least builds. Enable the Nitro SWIG bindings +# at your own risk! + +RUN apt-get update && \ + apt-get install -y libcurl4-openssl-dev libssl-dev python3 + +RUN pip install cxxfilt + +RUN git clone https://github.com/mdaus/nitro.git +WORKDIR /polytracker/the_klondike/nitro +RUN git checkout b39ccc4c07e84e6c05cecb9ae24143373a3ed8e2 +WORKDIR /polytracker/the_klondike/nitro/build/release + +# Build Nitro: Release +RUN polytracker build cmake ../.. \ + -DCMAKE_C_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable" \ + -DCMAKE_CXX_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable" \ + -DCMAKE_BUILD_TYPE=Release \ + -DCODA_BUILD_TESTS=OFF \ + -DENABLE_PYTHON=OFF + +RUN polytracker build cmake --build . -j$((`nproc`+1)) --target show_nitf++ --config Release + +RUN cp modules/c++/nitf/show_nitf++ nitro_Release + +RUN polytracker instrument-targets \ + --taint \ + --ftrace \ + --cflog \ + show_nitf++ + +RUN mv show_nitf++.instrumented nitro_trackRelease + +# Build Nitro: Debug +WORKDIR /polytracker/the_klondike/nitro/build/debug +RUN polytracker build cmake ../.. \ + -DCMAKE_C_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable" \ + -DCMAKE_CXX_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable" \ + -DCMAKE_BUILD_TYPE=Debug \ +-DCODA_BUILD_TESTS=OFF \ +-DENABLE_PYTHON=OFF + +RUN polytracker build cmake --build . -j$((`nproc`+1)) --clean-first --target show_nitf++ --config Debug +RUN cp modules/c++/nitf/show_nitf++ nitro_Debug +RUN polytracker instrument-targets \ + --taint \ + --ftrace \ + --cflog \ + show_nitf++ + +RUN mv show_nitf++.instrumented nitro_trackDebug + +# If this Dockerfile is run with run.sh, this will link to the external +# location where the evaluation scripts live, and you'll be dropped into a +# shell so you can work in a configured environment. +# Note to the unwary: compiler-rt sanitizers and Polytracker are NOT COMPATIBLE. +# If you need compiler-rt, please use Dockerfile.nitro.sanitizers. +WORKDIR /polytracker/the_klondike/nitro/build/ubet \ No newline at end of file diff --git a/examples/analysis/ubet/Dockerfile.nitro.sanitizers b/examples/analysis/ubet/Dockerfile.nitro.sanitizers new file mode 100644 index 00000000..babdfe30 --- /dev/null +++ b/examples/analysis/ubet/Dockerfile.nitro.sanitizers @@ -0,0 +1,34 @@ +FROM ubuntu:focal +ENV DEBIAN_FRONTEND=noninteractive +LABEL org.opencontainers.image.authors="kelly.kaoudis@trailofbits.com, henrik.brodin@trailofbits.com" +WORKDIR /nitro + +RUN apt-get update && \ + apt-get install -y libcurl4-openssl-dev libssl-dev git cmake clang-12 build-essential python python-numpy + +RUN git clone https://github.com/mdaus/nitro.git +WORKDIR /polytracker/the_klondike/nitro +RUN git checkout b39ccc4c07e84e6c05cecb9ae24143373a3ed8e2 +WORKDIR /nitro/nitro/build/release + +RUN cmake ../.. \ + -DCMAKE_C_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable -fsanitize=address,undefined" \ + -DCMAKE_CXX_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable -fsanitize=address,undefined" \ + -DCMAKE_LINK_FLAGS="-fsanitize=address,undefined" \ + -DCMAKE_BUILD_TYPE=Release -DCODA_BUILD_TESTS=OFF -DENABLE_PYTHON=OFF + +RUN cmake --build . -j$((`nproc`+1)) --target show_nitf++ --config Release + +RUN cp modules/c++/nitf/show_nitf++ nitro_Release + +WORKDIR /nitro/nitro/build/debug + +RUN cmake ../.. \ + -DCMAKE_C_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable -fsanitize=address,undefined" \ + -DCMAKE_CXX_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable -fsanitize=address,undefined" \ + -DCMAKE_LINK_FLAGS="-fsanitize=address,undefined" \ + -DCMAKE_BUILD_TYPE=Debug -DCODA_BUILD_TESTS=OFF -DENABLE_PYTHON=OFF + +RUN cmake --build . -j$((`nproc`+1)) --target show_nitf++ --config Debug + +RUN cp modules/c++/nitf/show_nitf++ nitro_Debug \ No newline at end of file diff --git a/examples/analysis/ubet/Dockerfile.polytracker b/examples/analysis/ubet/Dockerfile.polytracker new file mode 100644 index 00000000..37c6ca93 --- /dev/null +++ b/examples/analysis/ubet/Dockerfile.polytracker @@ -0,0 +1,5 @@ +FROM trailofbits/polytracker:latest + +RUN apt-get update && apt-get upgrade -y +RUN apt-get install -y g++ +WORKDIR /workdir diff --git a/examples/analysis/ubet/README.md b/examples/analysis/ubet/README.md new file mode 100644 index 00000000..e464b6c0 --- /dev/null +++ b/examples/analysis/ubet/README.md @@ -0,0 +1,43 @@ +# UBet + +In general this directory contains the analysis scripts, configuration, and *most* other things necessary to reproduce our results from the LangSec '23 paper [Automatically Detecting Variability Bugs Through Hybrid Control and Data Flow Analysis](https://langsec.org/spw23/papers.html#variability). + +## Reproducing our results +| :wrench: Getting Started | +| ------------------------ | +If you have not yet done so, clone [Nitro](https://github.com/mdaus/nitro) and ensure it builds with PolyTracker in Docker for you, as demonstrated in the base NITF Dockerfile `polytracker/examples/Dockerfile-nitro-nitf.demo`. You will notice our build process is somewhat different compared to [how the Nitro maintainers recommend building the software](https://github.com/mdaus/nitro#building-nitro) since that requires GCC, or MSVC. + +### :whale: Dockerfiles here +- `Dockerfile` creates a clean, reproducible testing environment we used to build the toy motivation examples and to build some earlier experiments that didn't make it into the LangSec version of our paper +- `Dockerfile.nitro` builds an instrumented version of Nitro with the needed dependencies available to reproduce the experiments described in the paper +- `Dockerfile.polytracker` creates a clean, reproducible PolyTracker based testing environment. The compiler-rt sanitizers aren't available here, since PolyTracker a) requires the WLLVM/gclang compiler front-end (it *does* work with Clang, but is really intended to work with gclang) and b) alters the ABI list and other critical items in a way that is not compatible with base dfsan and the rest of LLVM compiler-rt anymore. You will get weird errors if you try to run compiler-rt sanitizers in a PolyTracker based environment. +- `Dockerfile.nitro.sanitizers` builds Nitro with UBSan and ASan and attempts to use them to show some of the issues inherent in Nitro. We build Nitro with these compiler-rt sanitizers in a way as close to the way we build Nitro for PolyTracker as possible. + +### NITF +The examples we reference in the paper primarily relate to the [NITF](https://jitc.fhu.disa.mil/projects/nitf/testdata.aspx) (National Imagery Transmission Format) reference parser Nitro, though in our motivation section we also use some specifically targeted toy examples, available under `polytracker/examples/analysis/ubet/examples/motivation` and named by listing. + +NITF is a binary image file format. Each NITF packages one or more visual data representations (video, fingerprints, CAT scan, JPEG, etc.) with extra metadata and other conditionally included information e.g., captions, information for rendering visual redactions, or geo-reference data. Nitro parses multiple mutually incompatible versions of the NITF specification. To simulate the effects of encountering a particular bad input we would like reproduce the effects of in a testing, local, or staging environment we applied Nitro instrumented with UBet to a corpus of 148 valid and known-invalid NITF files. + +#### :blue_book: NITF standard +There are three publicly available versions of MIL-STD-2500 (A, B, and C) that collectively describe [NITF](https://www.wikidata.org/wiki/Q26218335) as Nitro understands it. As *we* understand it, MIL-STD-2500a and MIL-STD-2500b together describe NITF 2.0 (note most NITF 2.0 files will map better to the fields described in MIL-STD-2500a, but some NITF 2.0 files will map better to the fields described in MIL-STD-2500b!). MIL-STD-2500c is closest to NITF 2.1. [MIL-STD-1300a](https://web.archive.org/web/20130217094453/http://www.gwg.nga.mil/ntb/baseline/docs/1300a/1300a.pdf) may also be relevant to understanding the format. `NSIF` is another closely related format that is good to understand to figure out the overlaps between the A, B, and C NITF standards. + +#### Reproducing our results, or making results like them +From the current working directory (`examples/analysis/ubet`): + +``` +docker build -t trailofbits/polytracker-nitro -f +docker run -ti --rm -v $(pwd):workdir trailofbits/polytracker-nitro +cd /workdir +find nitfdir/ -type f | python3 eval_nitro.py --locate +mkdir output +python3 eval_nitro.py --cflog --compare output/U_2001E.NTF/ +``` + +There is also a script `run.sh` in the cwd that you can use to just drop into an appropriately configured environment using one of the above Dockerfiles for any experiments you'd like to run. + +| :exclamation: Note for the unwary | +| --------------------------------- | +Nitro replaces an old semi-custom build system known as [WAF](https://github.com/mdaus/nitro#building-with-waf) with a new build layer on top of CMake, [coda-oss](https://github.com/mdaus/coda-oss) that bakes in a bespoke stdlib implementation. We've had to [macro some of this out](https://github.com/trailofbits/polytracker/blob/master/examples/Dockerfile-nitro-nitf.demo#L16), since it relies on implementation-specific behaviour of GCC and is not entirely compatible with Clang. We are aware of other implementation-specific and undefined behaviour related issues within the coda-oss code that we are in the process of gathering more data on using this analysis and instrumentation code, in order to report to the Nitro maintainers, beyond the bugs discussed in the LangSec paper. + +### Dead code in Nitro +Nitro repository also contains some [possibly-dead](https://github.com/mdaus/nitro#platforms) code that we did not evaluate or interact with - namely the Matlab and Java and related bindings located there. We focused on building and instrumenting its C++ implementation initially. This also applies to Nitro's Python, since Nitro uses SWIG to generate Python bindings. diff --git a/examples/analysis/ubet/build_nitro.sh b/examples/analysis/ubet/build_nitro.sh new file mode 100755 index 00000000..d9a13a2a --- /dev/null +++ b/examples/analysis/ubet/build_nitro.sh @@ -0,0 +1,47 @@ +#!/usr/bin/bash + +# NASSERT/NDEBUG builds "O3" +mkdir release +cd release || exit +polytracker build cmake ../.. \ + -DCMAKE_C_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable" \ + -DCMAKE_CXX_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable" \ + -DCMAKE_BUILD_TYPE=Debug -DNASSERT=1 -DNDEBUG=1 -DCODA_BUILD_TESTS=OFF + +polytracker build cmake --build . -j$(($(nproc) + 1)) --target show_nitf++ --config Debug +polytracker extract-bc -o baseO3.bc modules/c++/nitf/show_nitf++ +opt -load "${COMPILER_DIR}/pass/libPolytrackerPass.so" -load-pass-plugin "${COMPILER_DIR}/pass/libPolytrackerPass.so" -passes=pt-tcf -o "after_preoptO3.bc" "baseO3.bc" +echo "Optmize bitcode" +polytracker opt-bc --output O3.bc after_preoptO3.bc +echo "Instrument optimized bitcode" +polytracker instrument-bc --ftrace --taint --output instrumentedO3.bc O3.bc +echo "Lower optimized bitcode" +polytracker lower-bc -t show_nitf++ -o nitro_trackRelease instrumentedO3.bc + +cd .. || exit + +# O0 build +mkdir debug +cd debug || exit +polytracker build cmake ../.. \ + -DCMAKE_C_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable" \ + -DCMAKE_CXX_FLAGS="-w -D_POSIX_C_SOURCE=200809L -DCODA_OSS_NO_is_trivially_copyable" \ + -DCMAKE_BUILD_TYPE=Debug -DCODA_BUILD_TESTS=OFF + +polytracker build cmake --build . -j$(($(nproc) + 1)) --target show_nitf++ --config Debug +polytracker extract-bc -o baseO0.bc modules/c++/nitf/show_nitf++ + +opt -load "${COMPILER_DIR}/pass/libPolytrackerPass.so" -load-pass-plugin "${COMPILER_DIR}/pass/libPolytrackerPass.so" -passes=pt-tcf -o "after_preoptO0.bc" "baseO0.bc" + +cp after_preoptO0.bc O0.bc + +echo "Instrument non-optimized bitcode" +polytracker instrument-bc --ftrace --taint --output instrumentedO0.bc O0.bc + +echo "Lower non-optimized bitcode" +polytracker lower-bc -t show_nitf++ -o nitro_trackDebug instrumentedO0.bc + +cd .. || exit + +cp release/nitro_trackRelease . +cp debug/nitro_trackDebug . diff --git a/examples/analysis/ubet/compress_tdag.py b/examples/analysis/ubet/compress_tdag.py new file mode 100644 index 00000000..df4bef89 --- /dev/null +++ b/examples/analysis/ubet/compress_tdag.py @@ -0,0 +1,85 @@ +from typing import List +from polytracker import taint_dag +from argparse import ArgumentParser +from pathlib import Path +import os + +# Note: This is being integrated into PolyTracker directly as `polytracker compress ` + + +def copy_section( + fin, fout, section_in: taint_dag.TDSectionMeta, section_out: taint_dag.TDSectionMeta +): + assert section_in.size == section_out.size + assert section_in.tag == section_out.tag + assert section_in.align == section_out.align + os.copy_file_range( + fin.fileno(), + fout.fileno(), + section_in.size, + section_in.offset, + section_out.offset, + ) + + +def compact_section( + starting_offset: int, section_in: taint_dag.TDSectionMeta +) -> taint_dag.TDSectionMeta: + section_out = taint_dag.TDSectionMeta() + section_out.offset = section_in.align * round(starting_offset / section_in.align) + section_out.align = section_in.align + section_out.size = section_in.size + section_out.tag = section_in.tag + return section_out + + +def main(): + parser = ArgumentParser( + prog="compress_tdag", description="Compress a sparse tdag file" + ) + parser.add_argument( + "-i", + "--input", + help="Sparse input (source) tdag file", + type=Path, + required=True, + ) + parser.add_argument( + "-o", + "--output", + help="Dense output (destination) tdag file", + type=Path, + required=True, + ) + + args = parser.parse_args() + + with open(args.input, "rb") as fin, open(args.output, "wb") as fout: + fmeta_in = taint_dag.TDFileMeta() + sections_in = [] + sections_out = [] + + fin.readinto(fmeta_in) + for n in range(fmeta_in.section_count): + section = taint_dag.TDSectionMeta() + fin.readinto(section) + sections_in.append(section) + header_len = fin.tell() + print(fmeta_in, sections_in, header_len) + + starting_offset = fin.tell() + fout.write(fmeta_in) + for section in sections_in: + section_out = compact_section(starting_offset, section) + sections_out.append(section_out) + fout.write(section_out) + starting_offset = section_out.offset + section_out.size + + print("COPY!") + for section_in, section_out in zip(sections_in, sections_out): + print(section_in, section_out) + copy_section(fin, fout, section_in, section_out) + + +if __name__ == "__main__": + main() diff --git a/examples/analysis/ubet/eval.py b/examples/analysis/ubet/eval.py new file mode 100644 index 00000000..36344f34 --- /dev/null +++ b/examples/analysis/ubet/eval.py @@ -0,0 +1,143 @@ +# /usr/bin/python +import os +import random +import sys +import subprocess +from typing import List, Tuple +from pathlib import Path + +from polytracker import PolyTrackerTrace + + +src_arg = Path(sys.argv[1]) +no_build = "nobuild" == sys.argv[2] if len(sys.argv) > 2 else False +src_dir = src_arg.parent +src_name = src_arg.name +print(f"DIR {src_dir} name {src_name}") + + +def build_command_line(src, dst, compiler, optlevel): + return [compiler, "-o", dst, "-std=c++17", f"-O{optlevel}"] + [src] + + +def flag_and_compiler_to_filename(src, flag, compiler): + return f"{src}-{compiler}-O{flag}" + + +def run_binary_record_output(binary, input): + return subprocess.check_output([f"./{binary}"], input=input, cwd=src_dir).decode( + "utf-8" + ) + + +def polytracker_build(cmdline): + command = ["/usr/bin/env", "polytracker", "build"] + cmdline + if not no_build: + subprocess.call(command, cwd=src_dir) + + +def polytracker_instrument(bin): + command = ["/usr/bin/env", "polytracker", "instrument-targets", "--taint", bin] + target_name = f"{bin}.instrumented" + if not no_build: + subprocess.call(command, cwd=src_dir) + os.rename(f"{src_dir}/ub.instrumented", f"{src_dir}/{target_name}") + return target_name + + +def build_binaries(compiler, optlevel, src) -> Tuple[str, str]: + fn = flag_and_compiler_to_filename(src, optlevel, compiler) + polytracker_build(build_command_line(src, fn, compiler, optlevel)) + instrumented_fn = polytracker_instrument(fn) + return (fn, instrumented_fn) + + +def instrumented_run_on_input(bin, input, db): + env = os.environ.copy() + env["POLYTRACKER_STDOUT_SINK"] = "1" + env["POLYTRACKER_STDIN_SOURCE"] = "1" + env["POLYDB"] = db + + command = [f"./{bin}"] + return subprocess.check_output(command, env=env, input=input, cwd=src_dir).decode( + "utf-8" + ) + + +def binary_to_db(bin): + return f"{bin}.db" + + +def compare_src(db_files): + print([str(x) for x in db_files]) + tdags = [PolyTrackerTrace.load(src_dir / fn) for fn in db_files] + tdfiles = [trace.tdfile for trace in tdags] + + # Use file zero as reference + for input_label in tdfiles[0].input_labels(): + nodes = [tdf.decode_node(input_label) for tdf in tdfiles] + print(f"label: {input_label} {nodes}") + + for i, tup in enumerate(map(lambda *x: tuple(x), *(tdf.sinks for tdf in tdfiles))): + print( + f"{i}: {tup} {[' <- DIFFERENCE', ''][all(e.label == tup[0].label for e in tup)]}" + ) + + for f in tdfiles: + for n in f.nodes: + print(f"{n}") + + # TODO(hbrodin): Random idea: could we use something like gspan to compare graphs? + + +def main(): + optlevel = [0, 3] + compilers = ["clang++"] + # compilers = ["clang++", "g++"] + + binaries = [] + instrumented_binaries = [] + compiler_opt = [] + + # Phase 1 build all versions of a binary + src = src_name + for o in optlevel: + for compiler in compilers: + fn, instrumented_fn = build_binaries(compiler, o, src) + binaries.append(fn) + instrumented_binaries.append(instrumented_fn) + + # Phase 2 give each binary same input and store output + results: List[Tuple[bytes, List[str]]] = [] + iter_count = 10 + + for i in range(0, iter_count): + input = random.randbytes(2) + output = [run_binary_record_output(binary, input) for binary in binaries] + + results.append((input, output)) + + for result in results: + print("=============================") + print(f"input: {result[0]}") + + if all(e == result[1][0] for e in result[1]): + print("OK!") + else: + for i, out in enumerate(result[1]): + print(f"---- {binaries[i]} -----") + print(f"{out}") + if instrumented_binaries[i] != "": + print(f"Run instrumented {instrumented_binaries[i]}") + output = instrumented_run_on_input( + instrumented_binaries[i], + result[0], + binary_to_db(instrumented_binaries[i]), + ) + if output != out: + print(f"Results differ {out} vs {output}") + compare_src([binary_to_db(x) for x in instrumented_binaries]) + + +if __name__ == "__main__": + main() diff --git a/examples/analysis/ubet/eval_nitro.py b/examples/analysis/ubet/eval_nitro.py new file mode 100644 index 00000000..c188aabb --- /dev/null +++ b/examples/analysis/ubet/eval_nitro.py @@ -0,0 +1,464 @@ +import argparse +from collections import defaultdict +import subprocess +import os +import sys +from typing import Optional, Set, Iterator, Tuple, Dict +from polytracker import PolyTrackerTrace, taint_dag +from polytracker.taint_dag import TDFile, TDNode, TDSourceNode, TDUnionNode, TDRangeNode +from polytracker.mapping import InputOutputMapping +from pathlib import Path + +# To Silence TQDM! +from tqdm import tqdm +from functools import partialmethod + +import cxxfilt + +tqdm.__init__ = partialmethod(tqdm.__init__, disable=True) + + +OUTPUT_COLUMN_WIDTH = 40 + + +def build_dir(is_debug): + if os.environ.get("UBET_BUILD_DIR", "") != "": + return Path(os.environ["UBET_BUILD_DIR"]) / ["release", "debug"][is_debug] + else: + return ( + Path("/polytracker/the_klondike/nitro/build") + / ["release", "debug"][is_debug] + ) + + +def instrumented_bin_path(is_debug): + return build_dir(is_debug) / ["nitro_trackRelease", "nitro_trackDebug"][is_debug] + + +def bin_path(is_debug): + return build_dir(is_debug) / ["nitro_Release", "nitro_Debug"][is_debug] + + +def function_id_path(is_debug): + return build_dir(is_debug) / "functionid.json" + + +def db_name(is_debug): + return ["Release", "Debug"][is_debug] + ".tdag" + + +LabelType = int +OffsetType = int +FileOffsetType = Tuple[Path, OffsetType] +CavityType = Tuple[OffsetType, OffsetType] + + +class OutputInputMapping: + def __init__(self, f: TDFile): + self.tdfile: TDFile = f + + def dfs_walk( + self, label: LabelType, seen: Optional[Set[LabelType]] = None + ) -> Iterator[Tuple[LabelType, TDNode]]: + if seen is None: + seen = set() + + stack = [label] + while stack: + lbl = stack.pop() + + if lbl in seen: + continue + + seen.add(lbl) + + n = self.tdfile.decode_node(lbl) + + yield (lbl, n) + + if isinstance(n, TDSourceNode): + continue + + elif isinstance(n, TDUnionNode): + stack.append(n.left) + stack.append(n.right) + + elif isinstance(n, TDRangeNode): + stack.extend(range(n.first, n.last + 1)) + + def mapping(self) -> Dict[FileOffsetType, Set[FileOffsetType]]: + result: Dict[FileOffsetType, Set[FileOffsetType]] = defaultdict(set) + for s in list(self.tdfile.sinks): + for _, n in self.dfs_walk(s.label): + if isinstance(n, TDSourceNode): + sp = self.tdfile.fd_headers[s.fdidx][0] + np = self.tdfile.fd_headers[n.idx][0] + result[(sp, s.offset)].add((np, n.offset)) + + return result + + +def eq(n1, n2): + if type(n1) is not type(n2): + return False + + if n1.affects_control_flow != n2.affects_control_flow: + return False + + if isinstance(n1, taint_dag.TDSourceNode): + return n1.idx == n2.idx and n1.offset == n2.offset + elif isinstance(n1, taint_dag.TDUnionTaint): + return n1.left == n2.left and n1.right == n2.right + elif isinstance(n1, taint_dag.TDRangeNode): + return n1.first == n2.first and n1.last == n2.last + + assert isinstance(n1, taint_dag.TDUntaintedNode) + return True + + +def input_offsets(tdf): + ret = {} + for input_label in tdf.input_labels(): + node = tdf.decode_node(input_label) + offset = node.offset + if offset in ret: + ret[offset].append(node) + else: + ret[offset] = [node] + + # Squash multiple labels at same offset if they are equal + for k, v in ret.items(): + if all(eq(vals, v[0]) for vals in v): + ret[k] = v[:1] + return ret + + +def enum_diff(dbg_tdfile, rel_tdfile): + offset_dbg = input_offsets(dbg_tdfile) + offset_rel = input_offsets(rel_tdfile) + + i = 0 + maxl = max(offset_dbg, key=int) + maxr = max(offset_rel, key=int) + maxtot = max(maxl, maxr) + while i <= maxtot: + if i in offset_dbg and i not in offset_rel: + print(f"Only DBG: {offset_dbg[i]}") + elif i in offset_rel and i not in offset_dbg: + print(f"Only REL: {offset_rel[i]}") + elif i in offset_dbg and i in offset_rel: + if len(offset_dbg[i]) == 1 and len(offset_rel[i]) == 1: + if not eq(offset_dbg[i][0], offset_rel[i][0]): + print(f"DBG {offset_dbg[i][0]} - REL {offset_rel[i][0]}") + elif offset_dbg[i] != offset_rel[i]: + print(f"ED (count): DBG {offset_dbg[i]} - REL {offset_rel[i]}") + + i += 1 + + +def compare_inputs_used(dbg_tdfile, rel_tdfile): + dbg_mapping = OutputInputMapping(dbg_tdfile).mapping() + rel_mapping = OutputInputMapping(rel_tdfile).mapping() + inputs_dbg = set(x[1] for x in dbg_mapping) + inputs_rel = set(x[1] for x in rel_mapping) + print(f"Input diffs: {sorted(inputs_rel-inputs_dbg)}") + + +def compare_run_trace(tdfdbg, tdfrel): + # TODO(hbrodin): Just outputing runtrace for release atm. + for e in tdfrel.events: + fn = cxxfilt.demangle(tdfdbg.fn_headers[e.fnidx][0]) + print(f"{e}: {fn}") + + +def input_set(first_label: int, tdag) -> Set[int]: + q = [first_label] + ret = set() + seen = set() + while q: + label = q.pop() + if label in seen: + continue + seen.add(label) + + n = tdag.decode_node(label) + if isinstance(n, taint_dag.TDSourceNode): + ret.add(n) + elif isinstance(n, taint_dag.TDUnionNode): + q.append(n.left) + q.append(n.right) + else: + for lbl in range(n.first, n.last + 1): + q.append(lbl) + + return ret + + +def input_offsets(first_label: int, tdag) -> Set[int]: + return sorted(map(lambda n: n.offset, input_set(first_label, tdag))) + + +def run_nitro(is_debug, filename): + args = [bin_path(is_debug), filename] + return subprocess.run(args, stdout=subprocess.PIPE, stderr=subprocess.PIPE) + + +def run_instrumented(is_debug: bool, inputfile: Path, targetdir: Path): + args = [instrumented_bin_path(is_debug), inputfile] + db = db_name(is_debug) + + e = { + "POLYDB": str(db), + "POLYTRACKER_STDOUT_SINK": "1", + "POLYTRACKER_LOG_CONTROL_FLOW": "1", + } + ret = subprocess.run(args, env=e, stdout=subprocess.PIPE, stderr=subprocess.PIPE) + os.rename(db, targetdir / db) + + +def locate_candindates(): + print("Locating candidates") + for filename in sys.stdin: + fn = Path(filename.rstrip()).absolute() + if not fn.exists(): + print(f"Skipping non-existing {fn}.") + continue + + print(f"Processing: {fn}") + + dbg = run_nitro(True, fn) + rel = run_nitro(False, fn) + if dbg.stdout != rel.stdout or dbg.stderr != rel.stderr: + targetdir = Path("./output") / fn.name + targetdir = targetdir.absolute() + if not targetdir.exists(): + targetdir.mkdir(0o755) + log = targetdir / "log.txt" + + with open(log, "w") as f: + f.write(f"FILE: {fn}\n") + f.write(f"DBG-stdout(utf-8): {dbg.stdout.decode('utf-8')}\n") + f.write(f"DBG-stderr(utf-8): {dbg.stderr.decode('utf-8')}\n") + f.write(f"REL-stdout(utf-8): {rel.stdout.decode('utf-8')}\n") + f.write(f"REL-stderr(utf-8): {rel.stderr.decode('utf-8')}\n") + + with open(targetdir / "stdout-dbg-raw", "wb") as f: + f.write(dbg.stdout) + with open(targetdir / "stdout-rel-raw", "wb") as f: + f.write(rel.stdout) + with open(targetdir / "stderr-dbg-raw", "wb") as f: + f.write(dbg.stderr) + with open(targetdir / "stderr-rel-raw", "wb") as f: + f.write(rel.stderr) + + run_instrumented(True, fn, targetdir) + run_instrumented(False, fn, targetdir) + + +def print_cols(dbg, release, additional=""): + print( + (dbg.ljust(OUTPUT_COLUMN_WIDTH)) + + release.ljust(OUTPUT_COLUMN_WIDTH) + + additional + ) + + +def compare_cflog(dbg_tdfile, rel_tdfile): + import json + + def get_cflog_entires(tdfile, is_debug): + with open(function_id_path(is_debug)) as f: + function_id = json.load(f) + cflog = tdfile._get_section(taint_dag.TDControlFlowLogSection) + cflog.function_id_mapping(list(map(cxxfilt.demangle, function_id))) + return list( + map( + lambda e: (input_offsets(e.label, tdfile), e.callstack), + filter( + lambda e: isinstance(e, taint_dag.TDTaintedControlFlowEvent), cflog + ), + ) + ) + + dbg = get_cflog_entires(dbg_tdfile, True) + rel = get_cflog_entires(rel_tdfile, False) + + print("COMPARE CONTROL FLOW LOGS") + n = max(len(dbg), len(rel)) + + len_dbg = len(dbg) + len_rel = len(rel) + + dbgidx = 0 + relidx = 0 + while dbgidx < len_dbg or relidx < len_rel: + dbg_entry = dbg[dbgidx] if dbgidx < len_dbg else None + rel_entry = rel[relidx] if relidx < len_rel else None + + if dbg_entry is None: + print_cols("", str(rel_entry), "") + relidx += 1 + continue + + if rel_entry is None: + print_cols(str(dbg_entry), "", "") + dbgidx += 1 + continue + + dbg_callstack = str(dbg_entry[1]) + rel_callstack = str(rel_entry[1]) + + if dbg_entry[0] == rel_entry[0]: + print_cols( + str(dbg_entry[0]), + str(rel_entry[0]), + f" !!! DBG: {dbg_callstack} != REL: {rel_callstack}" + if dbg_callstack != rel_callstack + else "", + ) + dbgidx += 1 + relidx += 1 + else: + # check if we should be stepping debug or release + # depending on shortest path + debug_steps = 0 + release_steps = 0 + + while dbgidx + debug_steps < len_dbg: + if dbg[dbgidx + debug_steps][0] == rel_entry[0]: + break + debug_steps += 1 + + while relidx + release_steps < len_rel: + if rel[relidx + release_steps][0] == dbg_entry[0]: + break + release_steps += 1 + + if debug_steps < release_steps: + print_cols(str(dbg_entry[0]), "", dbg_callstack) + dbgidx += 1 + else: + print_cols("", str(rel_entry[0]), rel_callstack) + relidx += 1 + + return + + +def compare_input_output(dbg_tdfile, rel_tdfile): + dbg_mapping = InputOutputMapping(dbg_tdfile).mapping() + rel_mapping = InputOutputMapping(rel_tdfile).mapping() + print("=============== INPUT -> OUTPUT =================") + keydiff = set(dbg_mapping) - set(rel_mapping) + print(f"KeyDiff {keydiff}") + + for k_dbg, v_dbg in dbg_mapping.items(): + v_rel = rel_mapping[k_dbg] + if v_dbg != v_rel: + print(f"{k_dbg}: DBG {v_dbg} REL {sorted(v_rel)}") + + +def compare_output_input(dbg_tdfile, rel_tdfile): + dbg_mapping = OutputInputMapping(dbg_tdfile).mapping() + rel_mapping = OutputInputMapping(rel_tdfile).mapping() + print("=============== OUTPUT -> INPUT =================") + keydiff = set(dbg_mapping) - set(rel_mapping) + print(f"KeyDiff {keydiff}") + + for k_dbg, v_dbg in dbg_mapping.items(): + v_rel = rel_mapping[k_dbg] + if v_dbg != v_rel: + print(f"{k_dbg}: DBG {v_dbg} REL {sorted(v_rel)}") + + +def do_comparison(path: Path, args): + dbg_tdag = path / db_name(True) + rel_tdag = path / db_name(False) + print(f"Compare {dbg_tdag} and {rel_tdag}") + + dbg_trace = PolyTrackerTrace.load(dbg_tdag) + rel_trace = PolyTrackerTrace.load(rel_tdag) + dbg_tdfile = dbg_trace.tdfile + rel_tdfile = rel_trace.tdfile + + if args.cflog: + compare_cflog(dbg_tdfile, rel_tdfile) + + if args.inout: + compare_input_output(dbg_tdfile, rel_tdfile) + + if args.outin: + compare_output_input(dbg_tdfile, rel_tdfile) + + if args.runtrace: + compare_run_trace(dbg_tdfile, rel_tdfile) + + if args.inputsused: + compare_inputs_used(dbg_tdfile, rel_tdfile) + + if args.enumdiff: + enum_diff(dbg_tdfile, rel_tdfile) + + +def main(): + parser = argparse.ArgumentParser( + prog="eval_nitro", + description="Evaluate unwanted/unexpected/undefined/implementation defined behaviours in Nitro", + ) + parser.add_argument( + "-l", + "--locate", + action="store_true", + help="Filenames read from stdin are run in Nitro and any discrepancies between debug/release builds are store in the output directory. Can be executed as 'find dir -type f | python3 eval_nitro.py -l'", + ) + parser.add_argument( + "-c", + "--compare", + type=Path, + help="Compare the Debug/Release tdags in the directory.", + ) + parser.add_argument( + "--cflog", + action="store_true", + help="Compare Control Flow Logs (requires --compare)", + ) + parser.add_argument( + "--inout", + action="store_true", + help="Compare Input-Output mapping (requires --compare)", + ) + parser.add_argument( + "--outin", + action="store_true", + help="Compare Output-Input mapping (requires --compare)", + ) + parser.add_argument( + "--runtrace", action="store_true", help="Compare runtrace (requires --compare)" + ) + parser.add_argument( + "--inputsused", + action="store_true", + help="Compare inputs used (requires --compare)", + ) + parser.add_argument( + "--enumdiff", + action="store_true", + help="Enumerate differences (kind of) (requires --compare)", + ) + + args = parser.parse_args() + + if args.locate and args.compare: + print("Error: Can both locate and compare") + parser.print_help() + return + elif args.locate: + locate_candindates() + elif args.compare: + do_comparison(args.compare, args) + else: + print("Error: Specify either -locate or -compare") + parser.print_help() + + +if __name__ == "__main__": + main() diff --git a/examples/analysis/ubet/motivation_examples/listing_1.cpp b/examples/analysis/ubet/motivation_examples/listing_1.cpp new file mode 100644 index 00000000..38a36591 --- /dev/null +++ b/examples/analysis/ubet/motivation_examples/listing_1.cpp @@ -0,0 +1,31 @@ +/* Consider the following (buggy) toy parser, which accepts at least one + * command-line argument. + * + * With optimizations disabled, Clang / LLVM 15.0.0 run on the code in Listing + * 1 will produce a binary that operates consistently dependent on input. Due + * to undefined behavior in the first branch of the conditional, a program + * version compiled with optimizations (-03) will only return 0; + * Clang elides the undefined behavior during optimization. + * + * Listing 1: According to the C++20 specification, a bitwise left shift + * operation (as on line 9 here) results in undefined behavior if the right + * operand is negative. + * + * Listing 2: Assembly of the program in Listing 1 when compiled with Clang’s + * -O3 optimization level. All of the control flow has been silently elided, + * and the program will always return 0. + * main: + * xor %eax, %eax + * retq + * The clear difference in the simple examples of program control flow in + * Listings 1 and 2 reflect potential effects of optimization passes run + * on code invoking undefined behavior. + */ + +int main(int argc, char* argv[]) { + if (argc > 1) { + return (int)*argv[argc - 1] << -2; + } else { + return 0; + } +} \ No newline at end of file diff --git a/examples/analysis/ubet/motivation_examples/listing_3.cpp b/examples/analysis/ubet/motivation_examples/listing_3.cpp new file mode 100644 index 00000000..d23299f3 --- /dev/null +++ b/examples/analysis/ubet/motivation_examples/listing_3.cpp @@ -0,0 +1,52 @@ +#if defined(PRODUCTION) +#define NDEBUG +#endif + +#include +#include + +/* + * Consider now Listing 3, for which we assume there are distinct debug + * and production build configurations. + * + * Listing 3: A bitwise left shift operation in the following toy program + * results in undefined behavior if shift is greater than the data type’s + * max bitwise capacity. Undefined behavior on line 22 here occurs dependent + * on user input and build configuration. + * + * Suppose, as above, a particular contributor writes control flow relying on + * an assertion to check the user-provided value of shift is within size bounds + * of the container int type, but another contributor later adds the NDEBUG + * macro to prevent assertion usage (as on Listing 3 lines 1–3) when the code + * is compiled with -DPRODUCTION. + * + * Now suppose a third programmer without knowledge of the source code observes + * their deployment of the production build allows a shift value of 63 + * (causing integer overflow), though all tests pass. Our third (debugging) + * programmer runs the debug binary version to reproduce the issue locally, + * where the assert() fails and integer overflow does not occur. If the binary + * is then instrumented at compile time with a common sanitizer such as UBSan + * [4] and the program receives 63 as its argument, UBSan will warn that a + * shift exponent of 63 is too large for the 32-bit int type, but will neither + * show that the NDEBUG macro redefines the assert() implementation to a + * no-op, nor show that an assertion guards a risky computation accepting + * unsanitized user input, where a conditional should be instead. While a + * static analyser could potentially provide some of this information, + * particularly if the codebase under analysis were more complex, it would be + * buried in a large “maybe” state space of potentially dangerous flows to + * analyse [3], and would not take into account the context that the input + * value 63 is problematic. If an ordinary programmer without significant + * knowledge of the codebase beforehand debugging a similar issue aims to + * quickly fix the real cause in a more complex codebase, neither of these + * common methods applies cleanly. + */ + +int main(int argc, char* argv[]) { + if (argc > 1) { + int shift = std::atoi(argv[1]); + assert(shift > 0 && shift < 32); + return 0xff << shift; + } else { + return 0; + } +} diff --git a/examples/analysis/ubet/motivation_examples/listing_4.cpp b/examples/analysis/ubet/motivation_examples/listing_4.cpp new file mode 100644 index 00000000..accbec18 --- /dev/null +++ b/examples/analysis/ubet/motivation_examples/listing_4.cpp @@ -0,0 +1,33 @@ +#include + +/* + * Compiled with Clang with optimizations enabled, when run, Listing 4 + * immediately exits after printing “Hello world!” [25]. + * + * Listing 4: A C++ program that one would expect to either enter an infinite + * busy loop, or immediately exit with code zero. A bug in the latest version + * of Clang/LLVM (15.0.0) causes this program to erroneously print “Hello + * world!” when compiled with optimizations enabled. + * + * When optimizing out the infinite loop (an operation the C++ standard + * allows), Clang fails to add an implicit return at the end of main(). + * Execution thus falls through to the code directly after main(): + * unreachable(). A binary built without optimizations does run the + * infinite while loop as expected; Listing 4’s execution and output only + * change when optimizations are enabled. This begs the question of whether + * a commodity sanitizer such as UBSan could poten- tially expose such an + * UB-adjacent issue. Yet when built with Clang (with and without + * optimizations) and UBSan, via the -fsanitize=undefined option + * (which includes the - fsanitize=return check intended to alert + * when the end of a value-returning function is reached without returning + * a value) the missing return is not caught. Such bugs and their full effects + * on program control and data flow are difficult to diagnose, particularly in + * complex programs. This paper proposes a technique to automatically trace + * back to the source lines most closely related to the origins of such bugs. + */ + +int main() { + while (true); +} + +void unreachable() { std::cout << "Hello world!" << std::endl; } diff --git a/examples/analysis/ubet/run.sh b/examples/analysis/ubet/run.sh new file mode 100755 index 00000000..2e153848 --- /dev/null +++ b/examples/analysis/ubet/run.sh @@ -0,0 +1,33 @@ +#!/usr/bin/env bash + +# This script and the associated Dockerfiles are known to work on Debian and Ubuntu and have not been tested in other environments. + +PASSTHROUGH_DOCKER_ARGS="" +DOCKERFILE="Dockerfile.nitro" +NO_CACHE="--no-cache" +IMAGE_NAME="ub-container" + +while getopts bp: arg; do + case "${arg}" in + b) + echo "(Re)building ${DOCKERFILE} container and saving as ${IMAGE_NAME} before running..." + docker build "${NO_CACHE}" -t "${IMAGE_NAME}" -f "${DOCKERFILE}" . + ;; + p) + PASSTHROUGH_DOCKER_ARGS=${OPTARG} + ;; + *) + echo "== run.sh: run UB Docker container-as-executable ==" + echo "Optional: -b (builds the ${IMAGE_NAME} image first from ${DOCKERFILE})" + echo "Optional: -p \"passthrough Docker executable arguments, in quotes\", c.f.:" + python3 eval_nitro.py --help + exit 1 + ;; + esac +done + +INTERNAL="/polytracker/the_klondike/nitro/build/ubet" +echo "run.sh: using $(pwd) as the volume attached to ${IMAGE_NAME} container internal location ${INTERNAL}" + +#shellcheck disable=SC2086 +docker run -it --rm --volume "$(pwd)":"${INTERNAL}" ub-container /usr/bin/bash ${PASSTHROUGH_DOCKER_ARGS} diff --git a/examples/analysis/ubet/tags/README.md b/examples/analysis/ubet/tags/README.md new file mode 100644 index 00000000..3e20a8ee --- /dev/null +++ b/examples/analysis/ubet/tags/README.md @@ -0,0 +1,15 @@ +# how to use a .tags file + +Our [PolyFile](https://github.com/trailofbits/polyfile) support for NITF is at time of writing limited to NITF 2.1, so in looking for quick ways to examine and tag the fields in NITF 2.0 files, I came across the VSCode extension "Hex Editor With Tags" (VSCode extension ID: notblank00.hexeditor), a mod of the Microsoft Hex Editor VSCode extension that adds colorizing. It's frankly a lot more manual than and nowhere near as nice as Polyfile, but it worked for some quick examples. + +## i_3034c.ntf + +With the VSCode extension installed, pull the [test NITF directory from the Galois Format Analysis Workbench](https://github.com/GaloisInc/FAW/tree/master/test_files/nitf) as referenced and used in our earlier NITF explorations - see polytracker/examples/analysis/nitf/base.sh. Place `i_3034c.ntf.tags` in the same directory as [i_3034c.ntf](https://github.com/GaloisInc/FAW/blob/master/test_files/nitf/i_3034c.ntf) and load the .ntf file into VSCode. VScode extension should automatically read the tags file and colorize the fields in the .ntf file for your visual exploration. + +## Initial Results: Table 1 + +Unfortunately, the NITF corpus we used in the paper to evaluate Nitro is not public and we do not own the rights to these files. Table 1 shows the results of running a PolyTracker and Ubet instrumeneted Nitro binary on one of these files, which is of NITF 2.0 format. + +`table_1_annotated_diff_with_fields.txt` is the results of combining UBet output with manual NITF field annotation using the Hex Editor with Tags extension. + +As discussed in the paper, we will replace this manual input annotation and exploration step with PolyFile, but have not yet gotten the time to do so. \ No newline at end of file diff --git a/examples/analysis/ubet/tags/i_3034c.ntf.tags b/examples/analysis/ubet/tags/i_3034c.ntf.tags new file mode 100644 index 00000000..f7f85c3b --- /dev/null +++ b/examples/analysis/ubet/tags/i_3034c.ntf.tags @@ -0,0 +1 @@ +[{"from":0,"to":3,"color":"red","caption":"File Profile Name (FHDR) "},{"from":4,"to":8,"color":"orange","caption":"File Version (FVER)"},{"from":9,"to":10,"color":"yellow","caption":"Complexity Level (CLEVEL) "},{"from":11,"to":14,"color":"lime","caption":"Standard Type (STYPE) "},{"from":15,"to":24,"color":"green","caption":"Originating Station ID (OSTAID) "},{"from":25,"to":38,"color":"aqua","caption":"File Date and Time (FDT) "},{"from":119,"to":119,"color":"purple","caption":"File Security Classification (FSCLAS) "},{"from":39,"to":118,"color":"blue","caption":"File Title (FTITLE) "},{"from":120,"to":121,"color":"pink","caption":"File Classification Security System (FSCLSY) "},{"from":122,"to":132,"color":"red","caption":"File Codewords (FSCODE) "},{"from":133,"to":134,"color":"orange","caption":"File Control and Handling (FSCTLH)"},{"from":135,"to":153,"color":"yellow","caption":"File Releasing Instructions (FSREL) "},{"from":154,"to":155,"color":"lime","caption":"File Declassification Type (FSDCTP) "},{"from":156,"to":163,"color":"green","caption":"File Declassification Date (FSDCDT) "},{"from":164,"to":167,"color":"aqua","caption":"File Declassification Exemption (FSDCXM) "},{"from":168,"to":168,"color":"blue","caption":"File Downgrade (FSDG) "},{"from":169,"to":176,"color":"purple","caption":"File Downgrade Date (FSDGDT) "},{"from":177,"to":219,"color":"pink","caption":"File Classification Text (FSCLTX) "},{"from":220,"to":220,"color":"red","caption":"File Classification Authority Type (FSCATP) "},{"from":221,"to":260,"color":"orange","caption":"File Classification Authority (FSCAUT) "},{"from":261,"to":261,"color":"yellow","caption":"File Classification Reason (FSCRSN) "},{"from":262,"to":269,"color":"lime","caption":"File Security Source Date (FSSRDT) "},{"from":270,"to":285,"color":"aqua","caption":"File Security Control Number (FSCTLN) "},{"from":286,"to":290,"color":"green","caption":"File Copy Number (FSCOP) "},{"from":291,"to":295,"color":"aqua","caption":"File Number of Copies (FSCPYS) "},{"from":291,"to":295,"color":"blue","caption":"File Number of Copies (FSCPYS) "},{"from":296,"to":296,"color":"pink","caption":"Encryption (ENCRYP) "},{"from":297,"to":299,"color":"purple","caption":"File Background Color (FBKGC) "},{"from":300,"to":323,"color":"red","caption":"Originator's Name (ONAME) "},{"from":324,"to":341,"color":"orange","caption":"Originator's Phone Number (OPHONE)"},{"from":342,"to":353,"color":"yellow","caption":"File Length (FL)"},{"from":354,"to":359,"color":"lime","caption":"NITF File Header Length (HL) "},{"from":360,"to":362,"color":"green","caption":"Number of Image Segments (NUMI) "},{"from":363,"to":368,"color":"aqua","caption":"Length of 1st Image Subheader (LISH001) "},{"from":369,"to":378,"color":"blue","caption":"Length of 1st Image Segment (LI001) "},{"from":379,"to":403,"color":"pink","caption":"??? remainder of nitf file header"},{"from":404,"to":405,"color":"red","caption":"NITF image subheader: File Part Type (IM)"},{"from":406,"to":415,"color":"orange","caption":"Image Identifier 1 (IID1) "},{"from":416,"to":429,"color":"yellow","caption":"Image Date and Time (IDATIM) "},{"from":430,"to":446,"color":"lime","caption":"Target Identifier (TGTID) "},{"from":447,"to":526,"color":"green","caption":"Image Identifier 2 (IID2) "},{"from":527,"to":527,"color":"aqua","caption":"Image Security Classification (ISCLAS) "},{"from":528,"to":529,"color":"blue","caption":"Image Classification Security System (ISCLSY) "},{"from":530,"to":540,"color":"purple","caption":"Image Codewords (ISCODE) "},{"from":541,"to":542,"color":"pink","caption":"Image Control and Handling (ISCTLH) "},{"from":543,"to":562,"color":"red","caption":"Image Releasing Instructions (ISREL) "},{"from":563,"to":564,"color":"orange","caption":"Image Declassification Type (ISDCTP) "},{"from":565,"to":572,"color":"yellow","caption":"Image Declassification Date (ISDCDT) "},{"from":573,"to":576,"color":"lime","caption":"Image Declassification Exemption (ISDCXM) "},{"from":577,"to":577,"color":"green","caption":"Image Downgrade (ISDG) "},{"from":578,"to":585,"color":"aqua","caption":"Image Downgrade Date (ISDGDT) "},{"from":586,"to":628,"color":"blue","caption":"Image Classification Text (ISCLTX) "},{"from":629,"to":629,"color":"purple","caption":"Image Classification Authority Type (ISCATP) "},{"from":630,"to":669,"color":"red","caption":"Image Classification Authority (ISCAUT) "},{"from":670,"to":670,"color":"pink","caption":"Image Classification Reason (ISCRSN) "},{"from":671,"to":678,"color":"orange","caption":"Image Security Source Date (ISSRDT) "},{"from":679,"to":693,"color":"yellow","caption":"Image Security Control Number (ISCTLN) "},{"from":694,"to":694,"color":"green","caption":"Encryption (ENCRYP) "},{"from":695,"to":736,"color":"aqua","caption":"Image Source (ISORCE) "},{"from":737,"to":744,"color":"blue","caption":"Number of Significant Rows in image (NROWS) "},{"from":745,"to":752,"color":"purple","caption":"Number of Significant Columns in image (NCOLS) "},{"from":753,"to":755,"color":"pink","caption":"Pixel Value Type (PVTYPE) "},{"from":764,"to":771,"color":"yellow","caption":"Image Category (ICAT) "},{"from":756,"to":763,"color":"orange","caption":"Image Representation (IREP) "},{"from":772,"to":773,"color":"red","caption":"Actual Bits-Per-Pixel Per Band (ABPP)"},{"from":774,"to":774,"color":"lime","caption":"Pixel Justification (PJUST) "},{"from":775,"to":775,"color":"pink","caption":"Image Coordinate Representation (ICORDS) "},{"from":776,"to":776,"color":"aqua","caption":"Number of Image Comments (NICOM) "},{"from":777,"to":778,"color":"red","caption":"Image Compression (IC) "},{"from":779,"to":779,"color":"lime","caption":"Number of Bands (NBANDS) "},{"from":780,"to":781,"color":"blue","caption":"1st Band Representation (IREPBAND1) "},{"from":782,"to":787,"color":"purple","caption":"1st Band Subcategory (ISUBCAT1) "},{"from":788,"to":788,"color":"green","caption":"1st Band Image Filter Condition (IFC1) "},{"from":789,"to":791,"color":"yellow","caption":"1st Band Standard Image Filter Code (IMFLT1) "},{"from":805,"to":805,"color":"red","caption":"Image Mode (IMODE) "},{"from":804,"to":804,"color":"orange","caption":"Image Sync Code (ISYNC) "},{"from":806,"to":809,"color":"lime","caption":"Number of Blocks per Row (NBPR) "},{"from":810,"to":813,"color":"purple","caption":"Number of Blocks per Column (NBPC) "},{"from":803,"to":803,"color":"pink","caption":"Number of LUTs for the 1st Image Band (NLUTS1)"},{"from":792,"to":802,"color":"red","caption":"???"},{"from":840,"to":843,"color":"purple","caption":"Image Magnification? (IMAG)"},{"from":854,"to":933,"color":"red","caption":"image data? "}] \ No newline at end of file diff --git a/examples/analysis/ubet/tags/table_1_annotated_diff_with_fields.txt b/examples/analysis/ubet/tags/table_1_annotated_diff_with_fields.txt new file mode 100644 index 00000000..76ff2bc0 --- /dev/null +++ b/examples/analysis/ubet/tags/table_1_annotated_diff_with_fields.txt @@ -0,0 +1,173 @@ +DIFF +--- Debug.tdag + ++++ Release.tdag + +@@ -1,55 +1,158 @@ + + [360, 361, 362] 3 byte field: Number of Images (NUMI) replaced with Number of Image Segments in 2500c + [360, 361, 362] 3 byte field: Number of Images (NUMI) + [360, 361, 362] 3 byte field: Number of Images (NUMI) + [379, 380, 381] 3 byte field: Number of Symbols (NUMS) + [382, 383, 384] 3 byte field: Number of Labels (NUML) + [385, 386, 387] 3 byte field: Number of Text Files (NUMT) + [388, 389, 390] 3 byte field: Number of Data Extension Segments (NUMDES) + [391, 392, 393] 3 byte field: Number of Reserved Extension Segements (NUMRES) + [394, 395, 396, 397, 398] + [399, 400, 401, 402, 403] + [354, 355, 356, 357, 358, 359] NITF File Header Length (HL) + [360, 361, 362] 3 byte field: Number of Images (NUMI) + [775] Image Coordinate System (ICORDS) +-[775] Image Coordinate System (ICORDS) +-[775] Image Coordinate System (ICORDS) + [776] Number of Image Comments (NICOM) +-[783] Number of Bands (NBANDS) 3 meaning RGB +-[783] + [783] + [783] + [783] + [783] + [783] + [796] 1st (Y) Band Number of LUTS (NLUTS1) + [783] Number of Bands (NBANDS) 3 meaning RGB + [809] nth (2nd; Cr) Band Number of LUTS (NLUTSnn) + [783] Number of Bands (NBANDS) 3 meaning RGB + [822] nth (3rd; Cb) Band Number of LUTS (NLUTSnn) + [783] Number of Bands (NBANDS) 3 meaning RGB + [863, 864, 865, 866, 867] User Defined Image Data Length (UDIDL) + [868, 869, 870, 871, 872] Extended Subheader Data Length (IXSHDL) + [363, 364, 365, 366, 367, 368] Length of 1st Image Subheader (LISH001) ++[369, 370, 371, 372, 373, 374, 375, 376, 377, 378] Length of 1st Image (LI001) + [360, 361, 362] 3 byte field: Number of Images (NUMI) + [379, 380, 381] 3 byte field: Number of Symbols (NUMS) + [382, 383, 384] 3 byte field: Number of Labels (NUML) + [385, 386, 387] 3 byte field: Number of Text Files (NUMT) + [388, 389, 390] 3 byte field: Number of Data Extension Segments (NUMDES) + [391, 392, 393] 3 byte field: Number of Reserved Extension Segements (NUMRES) + [360, 361, 362] 3 byte field: Number of Images (NUMI) + [360, 361, 362] 3 byte field: Number of Images (NUMI) + [360, 361, 362] 3 byte field: Number of Images (NUMI) + [379, 380, 381] 3 byte field: Number of Symbols (NUMS) + [382, 383, 384] 3 byte field: Number of Labels (NUML) + [385, 386, 387] 3 byte field: Number of Text Files (NUMT) + [388, 389, 390] 3 byte field: Number of Data Extension Segments (NUMDES) + [391, 392, 393] 3 byte field: Number of Reserved Extension Segements (NUMRES) + [360, 361, 362] 3 byte field: Number of Images (NUMI) + [360, 361, 362] 3 byte field: Number of Images (NUMI) + [360, 361, 362] 3 byte field: Number of Images (NUMI) ++[778] 2nd byte of IC field. Image Compression (IC) C3 is JPEG ++[778] 2nd byte of IC field. Image Compression (IC) C3 is JPEG ++[777] 1st byte of IC field. Image Compression (IC) C3 is JPEG ++[777] 1st byte of IC field. Image Compression (IC) C3 is JPEG ++[782] last byte of Compression Rate Code (COMRAT) ??? ++[782] last byte of Compression Rate Code (COMRAT) ??? ++[779] 1st byte of Compression Rate Code (COMRAT) ??? ++[779] 1st byte of Compression Rate Code (COMRAT) ??? ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[379, 380, 381] 3 byte field: Number of Symbols (NUMS) ++[379, 380, 381] 3 byte field: Number of Symbols (NUMS) ++[379, 380, 381] 3 byte field: Number of Symbols (NUMS) ++[382, 383, 384] 3 byte field: Number of Labels (NUML) ++[382, 383, 384] 3 byte field: Number of Labels (NUML) ++[382, 383, 384] 3 byte field: Number of Labels (NUML) ++[385, 386, 387] 3 byte field: Number of Text Files (NUMT) ++[385, 386, 387] 3 byte field: Number of Text Files (NUMT) ++[385, 386, 387] 3 byte field: Number of Text Files (NUMT) ++[388, 389, 390] 3 byte field: Number of Data Extension Segments (NUMDES) ++[388, 389, 390] 3 byte field: Number of Data Extension Segments (NUMDES) ++[388, 389, 390] 3 byte field: Number of Data Extension Segments (NUMDES) ++[391, 392, 393] 3 byte field: Number of Reserved Extension Segements (NUMRES) ++[391, 392, 393] 3 byte field: Number of Reserved Extension Segements (NUMRES) ++[391, 392, 393] 3 byte field: Number of Reserved Extension Segements (NUMRES) + [360, 361, 362] 3 byte field: Number of Images (NUMI) + [360, 361, 362] 3 byte field: Number of Images (NUMI) +-[360, 361, 362] 3 byte field: Number of Images (NUMI) +-[360, 361, 362] 3 byte field: Number of Images (NUMI) +-[360, 361, 362] 3 byte field: Number of Images (NUMI) +-[360, 361, 362] 3 byte field: Number of Images (NUMI) ++[379, 380, 381] 3 byte field: Number of Symbols (NUMS) ++[382, 383, 384] 3 byte field: Number of Labels (NUML) ++[385, 386, 387] 3 byte field: Number of Text Files (NUMT) ++[388, 389, 390] 3 byte field: Number of Data Extension Segments (NUMDES) ++[391, 392, 393] 3 byte field: Number of Reserved Extension Segements (NUMRES) ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB ++[783] Number of Bands (NBANDS) 3 meaning RGB \ No newline at end of file diff --git a/examples/http/picohttpparser/.dockerignore b/examples/http/picohttpparser/.dockerignore new file mode 100644 index 00000000..ab43995c --- /dev/null +++ b/examples/http/picohttpparser/.dockerignore @@ -0,0 +1,3 @@ +picohttpparser +example_picohttpparser +example_picohttpparser.o \ No newline at end of file diff --git a/examples/http/picohttpparser/Dockerfile b/examples/http/picohttpparser/Dockerfile new file mode 100644 index 00000000..89cb0fe0 --- /dev/null +++ b/examples/http/picohttpparser/Dockerfile @@ -0,0 +1,18 @@ +# Create a separate image with the latest source +FROM trailofbits/polytracker:latest +LABEL org.opencontainers.image.authors="lisa.overall@trailofbits.com" + +RUN rm -rf /polytracker/examples/http/picohttpparser && mkdir -p /polytracker/examples/http/picohttpparser + +WORKDIR /polytracker/examples/http/picohttpparser +RUN git clone https://github.com/h2o/picohttpparser.git +COPY Makefile example_picohttpparser.c /polytracker/examples/http/picohttpparser/ + +# Build and instrument +RUN polytracker build make -j$((`nproc`+1)) +RUN polytracker instrument-targets --taint --ftrace example_picohttpparser +RUN mv example_picohttpparser.instrumented example_picohttpparser_track + +# Note, the /workdir and /testcase directories are intended to be mounted at runtime +VOLUME ["/workdir", "/testcase"] +WORKDIR /workdir diff --git a/examples/http/picohttpparser/Makefile b/examples/http/picohttpparser/Makefile new file mode 100644 index 00000000..f1e58fd1 --- /dev/null +++ b/examples/http/picohttpparser/Makefile @@ -0,0 +1,18 @@ +.phony: all +all: example_picohttpparser + +picohttpparser: + git clone https://github.com/h2o/picohttpparser.git + +picohttpparser/picohttpparser.a picohttpparser/picohttpparser.h: picohttpparser + cd picohttpparser && gcc -c picohttpparser.c && ar -rc picohttpparser.a picohttpparser.o + +example_picohttpparser.o : example_picohttpparser.c picohttpparser/picohttpparser.h + $(CC) -Ipicohttpparser -O0 -c $< -o $@ + +example_picohttpparser : example_picohttpparser.o picohttpparser/picohttpparser.a + $(CC) -O0 $^ -o $@ + +.phony: clean +clean: + rm -rf picohttpparser example_picohttpparser example_picohttpparser.o \ No newline at end of file diff --git a/examples/http/picohttpparser/example_picohttpparser.c b/examples/http/picohttpparser/example_picohttpparser.c new file mode 100644 index 00000000..56125528 --- /dev/null +++ b/examples/http/picohttpparser/example_picohttpparser.c @@ -0,0 +1,90 @@ +#include +#include +#include +#include +#include "picohttpparser.h" + +#define ERROR_IO -1 +#define ERROR_PARSE -2 +#define ERROR_REQUEST_TOO_LONG -3 + +// tiny hack to make a union (range) for a memory range and ensure it is considered a taint sink +void make_union(char *p, size_t len) { + int sum = 0; + for (size_t i=0;i 0) + break; /* successfully parsed the request */ + else if (pret == -1) + return ERROR_PARSE; + /* request is incomplete, continue the loop */ + assert(pret == -2); + if (buflen == sizeof(buf)) + return ERROR_REQUEST_TOO_LONG; + } + + printf("request is %d bytes long\n", pret); + make_union(method, method_len); + printf("method is %.*s\n", (int)method_len, method); + make_union(path, path_len); + printf("path is %.*s\n", (int)path_len, path); + make_union((char*)&minor_version, sizeof(minor_version)); + printf("HTTP version is 1.%d\n", minor_version); + printf("headers:\n"); + for (i = 0; i != num_headers; ++i) { + make_union(headers[i].name, headers[i].name_len); + make_union(headers[i].value, headers[i].value_len); + printf("%.*s: %.*s\n", (int)headers[i].name_len, headers[i].name, + (int)headers[i].value_len, headers[i].value); + } + + fclose(infile); + return 0; +} + +int main(int argc, char *argv[]) { + int i; + if(argc < 2) { + return 1; + } + for(i=1; i/dev/null)" == "" ]]; then + docker build -t trailofbits/polytracker -f ../../Dockerfile ../../ +fi +if [[ "$(docker images -q trailofbits/polytracker-demo-http-picohttpparser 2>/dev/null)" == "" ]]; then + docker build -t trailofbits/polytracker-demo-http-picohttpparser . +fi + +HOST_PATH=$(realpath "$1") +BASENAME=$(basename "$HOST_PATH") +HOST_DIR=$(dirname "$HOST_PATH") + +# mount the file if it's not already in /workdir +SCRIPT_DIR=$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" &>/dev/null && pwd) +if [[ "$HOST_DIR" == "$SCRIPT_DIR" ]]; then + docker run --read-only -ti --rm -e POLYPATH="$1" -e POLYDB="$1.tdag" -e POLYTRACKER_STDOUT_SINK=1\ + --mount type=bind,source="$(pwd)",target=/workdir trailofbits/polytracker-demo-http-picohttpparser:latest \ + /polytracker/examples/http/picohttpparser/example_picohttpparser_track "$1" +else + CONTAINER_PATH=/testcase/"$BASENAME" + docker run --read-only -ti --rm -e POLYPATH="$CONTAINER_PATH" -e POLYDB=/workdir/"$BASENAME".tdag -e POLYTRACKER_STDOUT_SINK=1 \ + --mount type=bind,source="$(pwd)",target=/workdir \ + --mount type=bind,source="$HOST_PATH",target="$CONTAINER_PATH" \ + trailofbits/polytracker-demo-http-picohttpparser:latest \ + /polytracker/examples/http/picohttpparser/example_picohttpparser_track "$CONTAINER_PATH" +fi diff --git a/mypy.ini b/mypy.ini index 976ba029..248a5b65 100644 --- a/mypy.ini +++ b/mypy.ini @@ -1,2 +1,3 @@ [mypy] ignore_missing_imports = True +exclude = polytracker/src/compiler-rt diff --git a/polytracker/build.py b/polytracker/build.py index d350671c..eeded59c 100644 --- a/polytracker/build.py +++ b/polytracker/build.py @@ -147,6 +147,26 @@ def _optimize_bitcode(input_bitcode: Path, output_bitcode: Path) -> None: subprocess.check_call(cmd) +def _preopt_instrument_bitcode(input_bitcode: Path, output_bitcode: Path) -> None: + POLY_PASS_PATH: Path = _ensure_path_exists( + _compiler_dir_path() / "pass" / "libPolytrackerPass.so" + ) + + cmd = [ + "opt", + "-load", + str(POLY_PASS_PATH), + "-load-pass-plugin", + str(POLY_PASS_PATH), + "-passes=pt-tcf", + str(input_bitcode), + "-o", + str(output_bitcode), + ] + # execute `cmd` + subprocess.check_call(cmd) + + def _instrument_bitcode( input_bitcode: Path, output_bitcode: Path, @@ -398,16 +418,27 @@ def __init_arguments__(self, parser: argparse.ArgumentParser): help="specify additional ignore lists to polytracker", ) + parser.add_argument( + "--cflog", + action="store_true", + help="instrument with control affecting dataflow logging", + ) + def run(self, args: argparse.Namespace): for target in args.targets: blight_cmds = _read_blight_journal(args.journal_path) target_cmd, target_path = _find_target(target, blight_cmds) bc_path = target_path.with_suffix(".bc") + opt_bc = bc_path.with_suffix(".opt.bc") _extract_bitcode(target_path, bc_path) - _optimize_bitcode(bc_path, bc_path) + if args.cflog: + # Control affecting data flow logging happens before optimization + _preopt_instrument_bitcode(bc_path, bc_path) + + _optimize_bitcode(bc_path, opt_bc) inst_bc_path = Path(f"{bc_path.stem}.instrumented.bc") _instrument_bitcode( - bc_path, + opt_bc, inst_bc_path, args.ignore_lists, args.taint, diff --git a/polytracker/custom_abi/dfsan_abilist.txt b/polytracker/custom_abi/dfsan_abilist.txt index 62176f43..ff1afb51 100644 --- a/polytracker/custom_abi/dfsan_abilist.txt +++ b/polytracker/custom_abi/dfsan_abilist.txt @@ -32,6 +32,10 @@ fun:getrandom=discard ########################################## # Polytracker functions ######################################### +fun:__polytracker_leave_function=uninstrumented +fun:__polytracker_leave_function=discard +fun:__polytracker_enter_function=uninstrumented +fun:__polytracker_enter_function=discard fun:__polytracker_log_func_entry=uninstrumented fun:__polytracker_log_func_entry=discard fun:__polytracker_log_func_exit=uninstrumented @@ -45,6 +49,8 @@ fun:__polytracker_log_taint_op=uninstrumented fun:__polytracker_log_taint_op=custom fun:__polytracker_log_conditional_branch=uninstrumented fun:__polytracker_log_conditional_branch=custom +fun:__polytracker_log_tainted_control_flow=uninstrumented +fun:__polytracker_log_tainted_control_flow=custom # -- end fun:__polytracker_dump=uninstrumented fun:__polytracker_dump=discard diff --git a/polytracker/include/polytracker/passes/tainted_control_flow.h b/polytracker/include/polytracker/passes/tainted_control_flow.h new file mode 100644 index 00000000..b9d22f6a --- /dev/null +++ b/polytracker/include/polytracker/passes/tainted_control_flow.h @@ -0,0 +1,66 @@ +/* + * Copyright (c) 2023-present, Trail of Bits, Inc. + * All rights reserved. + * + * This source code is licensed in accordance with the terms specified in + * the LICENSE file found in the root directory of this source tree. + */ + +#pragma once + +#include +#include +#include + +namespace polytracker { +namespace detail { +struct FunctionMappingJSONWriter; +} + +class TaintedControlFlowPass + : public llvm::PassInfoMixin, + public llvm::InstVisitor { + // + llvm::IntegerType *label_ty{nullptr}; + // Taint tracking startup + llvm::FunctionCallee taint_start_fn; + // Log taint label affecting control flow + llvm::FunctionCallee cond_br_log_fn; + // Log enter/leave functions + llvm::FunctionCallee fn_enter_log_fn; + llvm::FunctionCallee fn_leave_log_fn; + + // Helpers + void insertCondBrLogCall(llvm::Instruction &inst, llvm::Value *val); + void insertTaintStartupCall(llvm::Module &mod); + void declareLoggingFunctions(llvm::Module &mod); + + llvm::ConstantInt *get_function_id_const(llvm::Function &f); + llvm::ConstantInt *get_function_id_const(llvm::Instruction &i); + +public: + using function_id = uint32_t; + + TaintedControlFlowPass(); + TaintedControlFlowPass(TaintedControlFlowPass &&); + ~TaintedControlFlowPass(); + + llvm::PreservedAnalyses run(llvm::Module &mod, + llvm::ModuleAnalysisManager &mam); + void visitGetElementPtrInst(llvm::GetElementPtrInst &gep); + void visitBranchInst(llvm::BranchInst &bi); + void visitSwitchInst(llvm::SwitchInst &si); + void visitSelectInst(llvm::SelectInst &si); + + void instrumentFunctionEnter(llvm::Function &func); + void visitReturnInst(llvm::ReturnInst &ri); + + function_id function_mapping(llvm::Function &func); + + std::unordered_map function_ids_; + function_id function_counter_{0}; + + std::unique_ptr function_mapping_writer_; +}; + +} // namespace polytracker \ No newline at end of file diff --git a/polytracker/include/taintdag/control_flow_log.h b/polytracker/include/taintdag/control_flow_log.h new file mode 100644 index 00000000..1bcff380 --- /dev/null +++ b/polytracker/include/taintdag/control_flow_log.h @@ -0,0 +1,88 @@ + +/* + * Copyright (c) 2022-present, Trail of Bits, Inc. + * All rights reserved. + * + * This source code is licensed in accordance with the terms specified in + * the LICENSE file found in the root directory of this source tree. + */ + +#pragma once + +#include "taintdag/outputfile.h" +#include "taintdag/section.h" +#include "taintdag/taint.h" +#include "taintdag/util.h" + +namespace taintdag { + +namespace detail { +// A uint32_t varint encoded by setting highest bit for all but the final byte. +// Requires up to 5 bytes of storage as each output byte uses 7 input bits. +// Total maximum need is floor(32/7) = 5. Returns number of bytes required. +size_t varint_encode(uint32_t val, uint8_t *buffer) { + auto orig_buffer = buffer; + while (val >= 0x80) { + *buffer++ = 0x80 | (val & 0x7f); + val >>= 7; + } + *buffer++ = val & 0x7f; + return buffer - orig_buffer; +} +// TODO (hbrodin): Should probably used std::span +} // namespace detail + +struct ControlFlowLog : public SectionBase { + enum EventType { + EnterFunction = 0, + LeaveFunction = 1, + TaintedControlFlow = 2, + }; + + static constexpr uint8_t tag{8}; + static constexpr size_t align_of{1}; + static constexpr size_t allocation_size{1024 * 1024 * 1024}; + + template + ControlFlowLog(SectionArg of) : SectionBase(of.range) {} + + void function_event(EventType evt, uint32_t function_id) { + uint8_t buffer[6]; + buffer[0] = static_cast(evt); + auto used = detail::varint_encode(function_id, &buffer[1]); + auto total = used + 1; + + if (auto wctx = write(total)) { + std::copy(&buffer[0], &buffer[total], wctx->mem.begin()); + } else { + error_exit("Failed to write ", total, + " bytes of output to the ControlFlowLog Section."); + } + } + void enter_function(uint32_t function_id) { + function_event(EnterFunction, function_id); + } + + void leave_function(uint32_t function_id) { + function_event(LeaveFunction, function_id); + } + + void tainted_control_flow(label_t label, uint32_t function_id) { + // 1 byte event, <= 5 bytes function id, <= 5 bytes label + uint8_t buffer[11]; + buffer[0] = static_cast(TaintedControlFlow); + auto used = detail::varint_encode(function_id, &buffer[1]); + auto total = used + 1; + used = detail::varint_encode(label, &buffer[total]); + total += used; + + if (auto wctx = write(total)) { + std::copy(&buffer[0], &buffer[total], wctx->mem.begin()); + } else { + error_exit("Failed to write ", total, + " bytes of output to the ControlFlowLog Section."); + } + } +}; + +} // namespace taintdag diff --git a/polytracker/include/taintdag/labels.h b/polytracker/include/taintdag/labels.h index 7c8915ac..66fcb5b4 100644 --- a/polytracker/include/taintdag/labels.h +++ b/polytracker/include/taintdag/labels.h @@ -19,7 +19,8 @@ namespace taintdag { struct Labels : public FixedSizeAlloc { static constexpr uint8_t tag{2}; - static constexpr size_t allocation_size{max_label + 1}; + static constexpr size_t allocation_size{sizeof(storage_t) * + (static_cast(max_label) + 1)}; // How many labels to scan backwards to detect if the same Taint is about to // be produced. diff --git a/polytracker/include/taintdag/polytracker.h b/polytracker/include/taintdag/polytracker.h index 08211fef..751f6d17 100644 --- a/polytracker/include/taintdag/polytracker.h +++ b/polytracker/include/taintdag/polytracker.h @@ -12,6 +12,7 @@ #include #include "taintdag/bitmap_section.h" +#include "taintdag/control_flow_log.h" #include "taintdag/fnmapping.h" #include "taintdag/fntrace.h" #include "taintdag/labels.h" @@ -54,6 +55,21 @@ class PolyTracker { // Update the label, it affects control flow void affects_control_flow(label_t taint_label); + // Instrumentation callback for when control flow is influenced by a + // a tainted value + void log_tainted_control_flow(label_t taint_label, uint32_t function_id); + + // Instrumentation callback for when execution enters a function + // NOTE: There is a overlap in functionality between this and `function_entry` + // they will co-exist for now as they operate slightly different. The + // underlying reason is that this was developed separately to support the + // Tainted Control Flow logging mechanism. + void enter_function(uint32_t function_id); + + // Instrumentation callback for when execution leaves a function + // NOTE: Se `enter_function` comment about overlap. + void leave_function(uint32_t function_id); + // Log tainted data flowed into the sink void taint_sink(int fd, util::Offset offset, void const *mem, size_t length); // Same as before, but use same label for all data @@ -79,7 +95,7 @@ class PolyTracker { // sections and in which order they appear. using ConcreteOutputFile = OutputFile; + SourceLabelIndexSection, Functions, Events, ControlFlowLog>; ConcreteOutputFile output_file_; // Tracking source offsets for streams (where offsets can be determined by diff --git a/polytracker/mapping.py b/polytracker/mapping.py index ffab2287..845015ba 100644 --- a/polytracker/mapping.py +++ b/polytracker/mapping.py @@ -61,7 +61,7 @@ def mapping(self) -> Dict[FileOffsetType, Set[FileOffsetType]]: return result - def marker_to_ranges(self, m: bytearray) -> List[CavityType]: + def marker_to_ranges(self, m: bytes) -> List[CavityType]: ranges = [] start = None for i, v in enumerate(m): @@ -87,27 +87,31 @@ def file_cavities(self) -> Dict[Path, List[CavityType]]: # iterating over sinks as any taint node that affects control flow will # already have all of its source taints affecting control flow, and thus # be in the marker array already. - for source_label in self.tdfile.input_labels(): - source_node = self.tdfile.decode_node(source_label) - assert isinstance(source_node, TDSourceNode) - source_index = source_node.idx - source_offset = source_node.offset - - if source_index not in markers: - # Attempt to get the size of the file, to prevent reallocation of the markers array. - # Use whatever size is greater (size hint will be zero for failures) to allocate the - # array. - fdheader = self.tdfile.fd_headers[source_index][1] - size = source_offset + 1 if fdheader.invalid_size() else fdheader.size - markers[source_index] = bytearray(size) - - marker = markers[source_index] - if source_offset >= len(marker): - marker = marker.ljust(source_offset + 1, b"\0") - markers[source_index] = marker - - if source_node.affects_control_flow: - marker[source_offset] = 1 + with tqdm(desc="indexing taint sources", unit="labels", leave=False) as t: + for source_label in self.tdfile.input_labels(): + t.update(1) + source_node = self.tdfile.decode_node(source_label) + assert isinstance(source_node, TDSourceNode) + source_index = source_node.idx + source_offset = source_node.offset + + if source_index not in markers: + # Attempt to get the size of the file, to prevent reallocation of the markers array. + # Use whatever size is greater (size hint will be zero for failures) to allocate the + # array. + fdheader = self.tdfile.fd_headers[source_index][1] + size = ( + source_offset + 1 if fdheader.invalid_size() else fdheader.size + ) + markers[source_index] = bytearray(size) + + marker = markers[source_index] + if source_offset >= len(marker): + marker = marker.ljust(source_offset + 1, b"\0") + markers[source_index] = marker + + if source_node.affects_control_flow: + marker[source_offset] = 1 # Now, iterate all taint labels written to outputs (sinks). Walk them backwards to reach # source nodes and mark any source offset contributing to outputs. If a node affects @@ -133,11 +137,18 @@ def file_cavities(self) -> Dict[Path, List[CavityType]]: elif isinstance(n, TDRangeNode): seen.update(range(n.first, n.last + 1)) + # Flatten all files by name in case files are opened multiple times + merged: Dict[Path, bytes] = {} + + for k, v in markers.items(): + fname = self.tdfile.fd_headers[k][0] + if fname in merged: + merged[fname] = bytes(a | b for (a, b) in zip(merged[fname], v)) + else: + merged[fname] = bytes(v) + # Convert the source index to the source path and marker bit arrays to ranges - return { - self.tdfile.fd_headers[k][0]: self.marker_to_ranges(v) - for (k, v) in markers.items() - } + return {k: self.marker_to_ranges(v) for (k, v) in merged.items()} class MapInputsToOutputs(Command): diff --git a/polytracker/src/compiler-rt/lib/dfsan/scripts/build-libc-list.py b/polytracker/src/compiler-rt/lib/dfsan/scripts/build-libc-list.py index 40805c01..0eb59a49 100755 --- a/polytracker/src/compiler-rt/lib/dfsan/scripts/build-libc-list.py +++ b/polytracker/src/compiler-rt/lib/dfsan/scripts/build-libc-list.py @@ -1,11 +1,11 @@ #!/usr/bin/env python -#===- lib/dfsan/scripts/build-libc-list.py ---------------------------------===# +# ===- lib/dfsan/scripts/build-libc-list.py ---------------------------------===# # # Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. # See https://llvm.org/LICENSE.txt for license information. # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception # -#===------------------------------------------------------------------------===# +# ===------------------------------------------------------------------------===# # The purpose of this script is to identify every function symbol in a set of # libraries (in this case, libc and libgcc) so that they can be marked as # uninstrumented, thus allowing the instrumentation pass to treat calls to those @@ -16,80 +16,108 @@ import sys from optparse import OptionParser + def defined_function_list(object): - functions = [] - readelf_proc = subprocess.Popen(['readelf', '-s', '-W', object], - stdout=subprocess.PIPE) - readelf = readelf_proc.communicate()[0].split('\n') - if readelf_proc.returncode != 0: - raise subprocess.CalledProcessError(readelf_proc.returncode, 'readelf') - for line in readelf: - if (line[31:35] == 'FUNC' or line[31:36] == 'IFUNC') and \ - line[39:44] != 'LOCAL' and \ - line[55:58] != 'UND': - function_name = line[59:].split('@')[0] - functions.append(function_name) - return functions + functions = [] + readelf_proc = subprocess.Popen( + ["readelf", "-s", "-W", object], stdout=subprocess.PIPE + ) + readelf = readelf_proc.communicate()[0].split("\n") + if readelf_proc.returncode != 0: + raise subprocess.CalledProcessError(readelf_proc.returncode, "readelf") + for line in readelf: + if ( + (line[31:35] == "FUNC" or line[31:36] == "IFUNC") + and line[39:44] != "LOCAL" + and line[55:58] != "UND" + ): + function_name = line[59:].split("@")[0] + functions.append(function_name) + return functions + p = OptionParser() -p.add_option('--libc-dso-path', metavar='PATH', - help='path to libc DSO directory', - default='/lib/x86_64-linux-gnu') -p.add_option('--libc-archive-path', metavar='PATH', - help='path to libc archive directory', - default='/usr/lib/x86_64-linux-gnu') +p.add_option( + "--libc-dso-path", + metavar="PATH", + help="path to libc DSO directory", + default="/lib/x86_64-linux-gnu", +) +p.add_option( + "--libc-archive-path", + metavar="PATH", + help="path to libc archive directory", + default="/usr/lib/x86_64-linux-gnu", +) -p.add_option('--libgcc-dso-path', metavar='PATH', - help='path to libgcc DSO directory', - default='/lib/x86_64-linux-gnu') -p.add_option('--libgcc-archive-path', metavar='PATH', - help='path to libgcc archive directory', - default='/usr/lib/gcc/x86_64-linux-gnu/4.6') +p.add_option( + "--libgcc-dso-path", + metavar="PATH", + help="path to libgcc DSO directory", + default="/lib/x86_64-linux-gnu", +) +p.add_option( + "--libgcc-archive-path", + metavar="PATH", + help="path to libgcc archive directory", + default="/usr/lib/gcc/x86_64-linux-gnu/4.6", +) -p.add_option('--with-libstdcxx', action='store_true', - dest='with_libstdcxx', - help='include libstdc++ in the list (inadvisable)') -p.add_option('--libstdcxx-dso-path', metavar='PATH', - help='path to libstdc++ DSO directory', - default='/usr/lib/x86_64-linux-gnu') +p.add_option( + "--with-libstdcxx", + action="store_true", + dest="with_libstdcxx", + help="include libstdc++ in the list (inadvisable)", +) +p.add_option( + "--libstdcxx-dso-path", + metavar="PATH", + help="path to libstdc++ DSO directory", + default="/usr/lib/x86_64-linux-gnu", +) (options, args) = p.parse_args() -libs = [os.path.join(options.libc_dso_path, name) for name in - ['ld-linux-x86-64.so.2', - 'libanl.so.1', - 'libBrokenLocale.so.1', - 'libcidn.so.1', - 'libcrypt.so.1', - 'libc.so.6', - 'libdl.so.2', - 'libm.so.6', - 'libnsl.so.1', - 'libpthread.so.0', - 'libresolv.so.2', - 'librt.so.1', - 'libthread_db.so.1', - 'libutil.so.1']] -libs += [os.path.join(options.libc_archive_path, name) for name in - ['libc_nonshared.a', - 'libpthread_nonshared.a']] +libs = [ + os.path.join(options.libc_dso_path, name) + for name in [ + "ld-linux-x86-64.so.2", + "libanl.so.1", + "libBrokenLocale.so.1", + "libcidn.so.1", + "libcrypt.so.1", + "libc.so.6", + "libdl.so.2", + "libm.so.6", + "libnsl.so.1", + "libpthread.so.0", + "libresolv.so.2", + "librt.so.1", + "libthread_db.so.1", + "libutil.so.1", + ] +] +libs += [ + os.path.join(options.libc_archive_path, name) + for name in ["libc_nonshared.a", "libpthread_nonshared.a"] +] -libs.append(os.path.join(options.libgcc_dso_path, 'libgcc_s.so.1')) -libs.append(os.path.join(options.libgcc_archive_path, 'libgcc.a')) +libs.append(os.path.join(options.libgcc_dso_path, "libgcc_s.so.1")) +libs.append(os.path.join(options.libgcc_archive_path, "libgcc.a")) if options.with_libstdcxx: - libs.append(os.path.join(options.libstdcxx_dso_path, 'libstdc++.so.6')) + libs.append(os.path.join(options.libstdcxx_dso_path, "libstdc++.so.6")) functions = [] for l in libs: - if os.path.exists(l): - functions += defined_function_list(l) - else: - print >> sys.stderr, 'warning: library %s not found' % l + if os.path.exists(l): + functions += defined_function_list(l) + else: + print >> sys.stderr, "warning: library %s not found" % l functions = list(set(functions)) functions.sort() for f in functions: - print 'fun:%s=uninstrumented' % f + print("fun:{f}=uninstrumented") diff --git a/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/cpplint.py b/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/cpplint.py index 1262e5b1..965ee6c4 100755 --- a/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/cpplint.py +++ b/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/cpplint.py @@ -54,9 +54,9 @@ import sysconfig try: - xrange # Python 2 + xrange # Python 2 except NameError: - xrange = range # Python 3 + xrange = range # Python 3 _USAGE = """ @@ -202,250 +202,253 @@ # If you add a new error message with a new category, add it to the list # here! cpplint_unittest.py should tell you if you forget to do this. _ERROR_CATEGORIES = [ - 'build/class', - 'build/c++11', - 'build/c++14', - 'build/c++tr1', - 'build/deprecated', - 'build/endif_comment', - 'build/explicit_make_pair', - 'build/forward_decl', - 'build/header_guard', - 'build/include', - 'build/include_alpha', - 'build/include_order', - 'build/include_what_you_use', - 'build/namespaces', - 'build/printf_format', - 'build/storage_class', - 'legal/copyright', - 'readability/alt_tokens', - 'readability/braces', - 'readability/casting', - 'readability/check', - 'readability/constructors', - 'readability/fn_size', - 'readability/inheritance', - 'readability/multiline_comment', - 'readability/multiline_string', - 'readability/namespace', - 'readability/nolint', - 'readability/nul', - 'readability/strings', - 'readability/todo', - 'readability/utf8', - 'runtime/arrays', - 'runtime/casting', - 'runtime/explicit', - 'runtime/int', - 'runtime/init', - 'runtime/invalid_increment', - 'runtime/member_string_references', - 'runtime/memset', - 'runtime/indentation_namespace', - 'runtime/operator', - 'runtime/printf', - 'runtime/printf_format', - 'runtime/references', - 'runtime/string', - 'runtime/threadsafe_fn', - 'runtime/vlog', - 'whitespace/blank_line', - 'whitespace/braces', - 'whitespace/comma', - 'whitespace/comments', - 'whitespace/empty_conditional_body', - 'whitespace/empty_if_body', - 'whitespace/empty_loop_body', - 'whitespace/end_of_line', - 'whitespace/ending_newline', - 'whitespace/forcolon', - 'whitespace/indent', - 'whitespace/line_length', - 'whitespace/newline', - 'whitespace/operators', - 'whitespace/parens', - 'whitespace/semicolon', - 'whitespace/tab', - 'whitespace/todo', - ] + "build/class", + "build/c++11", + "build/c++14", + "build/c++tr1", + "build/deprecated", + "build/endif_comment", + "build/explicit_make_pair", + "build/forward_decl", + "build/header_guard", + "build/include", + "build/include_alpha", + "build/include_order", + "build/include_what_you_use", + "build/namespaces", + "build/printf_format", + "build/storage_class", + "legal/copyright", + "readability/alt_tokens", + "readability/braces", + "readability/casting", + "readability/check", + "readability/constructors", + "readability/fn_size", + "readability/inheritance", + "readability/multiline_comment", + "readability/multiline_string", + "readability/namespace", + "readability/nolint", + "readability/nul", + "readability/strings", + "readability/todo", + "readability/utf8", + "runtime/arrays", + "runtime/casting", + "runtime/explicit", + "runtime/int", + "runtime/init", + "runtime/invalid_increment", + "runtime/member_string_references", + "runtime/memset", + "runtime/indentation_namespace", + "runtime/operator", + "runtime/printf", + "runtime/printf_format", + "runtime/references", + "runtime/string", + "runtime/threadsafe_fn", + "runtime/vlog", + "whitespace/blank_line", + "whitespace/braces", + "whitespace/comma", + "whitespace/comments", + "whitespace/empty_conditional_body", + "whitespace/empty_if_body", + "whitespace/empty_loop_body", + "whitespace/end_of_line", + "whitespace/ending_newline", + "whitespace/forcolon", + "whitespace/indent", + "whitespace/line_length", + "whitespace/newline", + "whitespace/operators", + "whitespace/parens", + "whitespace/semicolon", + "whitespace/tab", + "whitespace/todo", +] # These error categories are no longer enforced by cpplint, but for backwards- # compatibility they may still appear in NOLINT comments. _LEGACY_ERROR_CATEGORIES = [ - 'readability/streams', - 'readability/function', - ] + "readability/streams", + "readability/function", +] # The default state of the category filter. This is overridden by the --filter= # flag. By default all errors are on, so only add here categories that should be # off by default (i.e., categories that must be enabled by the --filter= flags). # All entries here should start with a '-' or '+', as in the --filter= flag. -_DEFAULT_FILTERS = ['-build/include_alpha'] +_DEFAULT_FILTERS = ["-build/include_alpha"] # The default list of categories suppressed for C (not C++) files. _DEFAULT_C_SUPPRESSED_CATEGORIES = [ - 'readability/casting', - ] + "readability/casting", +] # The default list of categories suppressed for Linux Kernel files. _DEFAULT_KERNEL_SUPPRESSED_CATEGORIES = [ - 'whitespace/tab', - ] + "whitespace/tab", +] # We used to check for high-bit characters, but after much discussion we # decided those were OK, as long as they were in UTF-8 and didn't represent # hard-coded international strings, which belong in a separate i18n file. # C++ headers -_CPP_HEADERS = frozenset([ - # Legacy - 'algobase.h', - 'algo.h', - 'alloc.h', - 'builtinbuf.h', - 'bvector.h', - 'complex.h', - 'defalloc.h', - 'deque.h', - 'editbuf.h', - 'fstream.h', - 'function.h', - 'hash_map', - 'hash_map.h', - 'hash_set', - 'hash_set.h', - 'hashtable.h', - 'heap.h', - 'indstream.h', - 'iomanip.h', - 'iostream.h', - 'istream.h', - 'iterator.h', - 'list.h', - 'map.h', - 'multimap.h', - 'multiset.h', - 'ostream.h', - 'pair.h', - 'parsestream.h', - 'pfstream.h', - 'procbuf.h', - 'pthread_alloc', - 'pthread_alloc.h', - 'rope', - 'rope.h', - 'ropeimpl.h', - 'set.h', - 'slist', - 'slist.h', - 'stack.h', - 'stdiostream.h', - 'stl_alloc.h', - 'stl_relops.h', - 'streambuf.h', - 'stream.h', - 'strfile.h', - 'strstream.h', - 'tempbuf.h', - 'tree.h', - 'type_traits.h', - 'vector.h', - # 17.6.1.2 C++ library headers - 'algorithm', - 'array', - 'atomic', - 'bitset', - 'chrono', - 'codecvt', - 'complex', - 'condition_variable', - 'deque', - 'exception', - 'forward_list', - 'fstream', - 'functional', - 'future', - 'initializer_list', - 'iomanip', - 'ios', - 'iosfwd', - 'iostream', - 'istream', - 'iterator', - 'limits', - 'list', - 'locale', - 'map', - 'memory', - 'mutex', - 'new', - 'numeric', - 'ostream', - 'queue', - 'random', - 'ratio', - 'regex', - 'scoped_allocator', - 'set', - 'sstream', - 'stack', - 'stdexcept', - 'streambuf', - 'string', - 'strstream', - 'system_error', - 'thread', - 'tuple', - 'typeindex', - 'typeinfo', - 'type_traits', - 'unordered_map', - 'unordered_set', - 'utility', - 'valarray', - 'vector', - # 17.6.1.2 C++ headers for C library facilities - 'cassert', - 'ccomplex', - 'cctype', - 'cerrno', - 'cfenv', - 'cfloat', - 'cinttypes', - 'ciso646', - 'climits', - 'clocale', - 'cmath', - 'csetjmp', - 'csignal', - 'cstdalign', - 'cstdarg', - 'cstdbool', - 'cstddef', - 'cstdint', - 'cstdio', - 'cstdlib', - 'cstring', - 'ctgmath', - 'ctime', - 'cuchar', - 'cwchar', - 'cwctype', - ]) +_CPP_HEADERS = frozenset( + [ + # Legacy + "algobase.h", + "algo.h", + "alloc.h", + "builtinbuf.h", + "bvector.h", + "complex.h", + "defalloc.h", + "deque.h", + "editbuf.h", + "fstream.h", + "function.h", + "hash_map", + "hash_map.h", + "hash_set", + "hash_set.h", + "hashtable.h", + "heap.h", + "indstream.h", + "iomanip.h", + "iostream.h", + "istream.h", + "iterator.h", + "list.h", + "map.h", + "multimap.h", + "multiset.h", + "ostream.h", + "pair.h", + "parsestream.h", + "pfstream.h", + "procbuf.h", + "pthread_alloc", + "pthread_alloc.h", + "rope", + "rope.h", + "ropeimpl.h", + "set.h", + "slist", + "slist.h", + "stack.h", + "stdiostream.h", + "stl_alloc.h", + "stl_relops.h", + "streambuf.h", + "stream.h", + "strfile.h", + "strstream.h", + "tempbuf.h", + "tree.h", + "type_traits.h", + "vector.h", + # 17.6.1.2 C++ library headers + "algorithm", + "array", + "atomic", + "bitset", + "chrono", + "codecvt", + "complex", + "condition_variable", + "deque", + "exception", + "forward_list", + "fstream", + "functional", + "future", + "initializer_list", + "iomanip", + "ios", + "iosfwd", + "iostream", + "istream", + "iterator", + "limits", + "list", + "locale", + "map", + "memory", + "mutex", + "new", + "numeric", + "ostream", + "queue", + "random", + "ratio", + "regex", + "scoped_allocator", + "set", + "sstream", + "stack", + "stdexcept", + "streambuf", + "string", + "strstream", + "system_error", + "thread", + "tuple", + "typeindex", + "typeinfo", + "type_traits", + "unordered_map", + "unordered_set", + "utility", + "valarray", + "vector", + # 17.6.1.2 C++ headers for C library facilities + "cassert", + "ccomplex", + "cctype", + "cerrno", + "cfenv", + "cfloat", + "cinttypes", + "ciso646", + "climits", + "clocale", + "cmath", + "csetjmp", + "csignal", + "cstdalign", + "cstdarg", + "cstdbool", + "cstddef", + "cstdint", + "cstdio", + "cstdlib", + "cstring", + "ctgmath", + "ctime", + "cuchar", + "cwchar", + "cwctype", + ] +) # Type names _TYPES = re.compile( - r'^(?:' + r"^(?:" # [dcl.type.simple] - r'(char(16_t|32_t)?)|wchar_t|' - r'bool|short|int|long|signed|unsigned|float|double|' + r"(char(16_t|32_t)?)|wchar_t|" + r"bool|short|int|long|signed|unsigned|float|double|" # [support.types] - r'(ptrdiff_t|size_t|max_align_t|nullptr_t)|' + r"(ptrdiff_t|size_t|max_align_t|nullptr_t)|" # [cstdint.syn] - r'(u?int(_fast|_least)?(8|16|32|64)_t)|' - r'(u?int(max|ptr)_t)|' - r')$') + r"(u?int(_fast|_least)?(8|16|32|64)_t)|" + r"(u?int(max|ptr)_t)|" + r")$" +) # These headers are excluded from [build/include] and [build/include_order] @@ -454,38 +457,52 @@ # uppercase character, such as Python.h or nsStringAPI.h, for example). # - Lua headers. _THIRD_PARTY_HEADERS_PATTERN = re.compile( - r'^(?:[^/]*[A-Z][^/]*\.h|lua\.h|lauxlib\.h|lualib\.h)$') + r"^(?:[^/]*[A-Z][^/]*\.h|lua\.h|lauxlib\.h|lualib\.h)$" +) # Pattern for matching FileInfo.BaseName() against test file name -_TEST_FILE_SUFFIX = r'(_test|_unittest|_regtest)$' +_TEST_FILE_SUFFIX = r"(_test|_unittest|_regtest)$" # Pattern that matches only complete whitespace, possibly across multiple lines. -_EMPTY_CONDITIONAL_BODY_PATTERN = re.compile(r'^\s*$', re.DOTALL) +_EMPTY_CONDITIONAL_BODY_PATTERN = re.compile(r"^\s*$", re.DOTALL) # Assertion macros. These are defined in base/logging.h and # testing/base/public/gunit.h. _CHECK_MACROS = [ - 'DCHECK', 'CHECK', - 'EXPECT_TRUE', 'ASSERT_TRUE', - 'EXPECT_FALSE', 'ASSERT_FALSE', - ] + "DCHECK", + "CHECK", + "EXPECT_TRUE", + "ASSERT_TRUE", + "EXPECT_FALSE", + "ASSERT_FALSE", +] # Replacement macros for CHECK/DCHECK/EXPECT_TRUE/EXPECT_FALSE _CHECK_REPLACEMENT = dict([(m, {}) for m in _CHECK_MACROS]) -for op, replacement in [('==', 'EQ'), ('!=', 'NE'), - ('>=', 'GE'), ('>', 'GT'), - ('<=', 'LE'), ('<', 'LT')]: - _CHECK_REPLACEMENT['DCHECK'][op] = 'DCHECK_%s' % replacement - _CHECK_REPLACEMENT['CHECK'][op] = 'CHECK_%s' % replacement - _CHECK_REPLACEMENT['EXPECT_TRUE'][op] = 'EXPECT_%s' % replacement - _CHECK_REPLACEMENT['ASSERT_TRUE'][op] = 'ASSERT_%s' % replacement - -for op, inv_replacement in [('==', 'NE'), ('!=', 'EQ'), - ('>=', 'LT'), ('>', 'LE'), - ('<=', 'GT'), ('<', 'GE')]: - _CHECK_REPLACEMENT['EXPECT_FALSE'][op] = 'EXPECT_%s' % inv_replacement - _CHECK_REPLACEMENT['ASSERT_FALSE'][op] = 'ASSERT_%s' % inv_replacement +for op, replacement in [ + ("==", "EQ"), + ("!=", "NE"), + (">=", "GE"), + (">", "GT"), + ("<=", "LE"), + ("<", "LT"), +]: + _CHECK_REPLACEMENT["DCHECK"][op] = "DCHECK_%s" % replacement + _CHECK_REPLACEMENT["CHECK"][op] = "CHECK_%s" % replacement + _CHECK_REPLACEMENT["EXPECT_TRUE"][op] = "EXPECT_%s" % replacement + _CHECK_REPLACEMENT["ASSERT_TRUE"][op] = "ASSERT_%s" % replacement + +for op, inv_replacement in [ + ("==", "NE"), + ("!=", "EQ"), + (">=", "LT"), + (">", "LE"), + ("<=", "GT"), + ("<", "GE"), +]: + _CHECK_REPLACEMENT["EXPECT_FALSE"][op] = "EXPECT_%s" % inv_replacement + _CHECK_REPLACEMENT["ASSERT_FALSE"][op] = "ASSERT_%s" % inv_replacement # Alternative tokens and their replacements. For full list, see section 2.5 # Alternative tokens [lex.digraph] in the C++ standard. @@ -493,18 +510,18 @@ # Digraphs (such as '%:') are not included here since it's a mess to # match those on a word boundary. _ALT_TOKEN_REPLACEMENT = { - 'and': '&&', - 'bitor': '|', - 'or': '||', - 'xor': '^', - 'compl': '~', - 'bitand': '&', - 'and_eq': '&=', - 'or_eq': '|=', - 'xor_eq': '^=', - 'not': '!', - 'not_eq': '!=' - } + "and": "&&", + "bitor": "|", + "or": "||", + "xor": "^", + "compl": "~", + "bitand": "&", + "and_eq": "&=", + "or_eq": "|=", + "xor_eq": "^=", + "not": "!", + "not_eq": "!=", +} # Compile regular expression that matches all the above keywords. The "[ =()]" # bit is meant to avoid matching these keywords outside of boolean expressions. @@ -512,7 +529,8 @@ # False positives include C-style multi-line comments and multi-line strings # but those have always been troublesome for cpplint. _ALT_TOKEN_REPLACEMENT_PATTERN = re.compile( - r'[ =()](' + ('|'.join(_ALT_TOKEN_REPLACEMENT.keys())) + r')(?=[ (]|$)') + r"[ =()](" + ("|".join(_ALT_TOKEN_REPLACEMENT.keys())) + r")(?=[ (]|$)" +) # These constants define types of headers for use with @@ -524,22 +542,23 @@ _OTHER_HEADER = 5 # These constants define the current inline assembly state -_NO_ASM = 0 # Outside of inline assembly block -_INSIDE_ASM = 1 # Inside inline assembly block -_END_ASM = 2 # Last line of inline assembly block -_BLOCK_ASM = 3 # The whole block is an inline assembly block +_NO_ASM = 0 # Outside of inline assembly block +_INSIDE_ASM = 1 # Inside inline assembly block +_END_ASM = 2 # Last line of inline assembly block +_BLOCK_ASM = 3 # The whole block is an inline assembly block # Match start of assembly blocks -_MATCH_ASM = re.compile(r'^\s*(?:asm|_asm|__asm|__asm__)' - r'(?:\s+(volatile|__volatile__))?' - r'\s*[{(]') +_MATCH_ASM = re.compile( + r"^\s*(?:asm|_asm|__asm|__asm__)" r"(?:\s+(volatile|__volatile__))?" r"\s*[{(]" +) # Match strings that indicate we're working on a C (not C++) file. -_SEARCH_C_FILE = re.compile(r'\b(?:LINT_C_FILE|' - r'vim?:\s*.*(\s*|:)filetype=c(\s*|:|$))') +_SEARCH_C_FILE = re.compile( + r"\b(?:LINT_C_FILE|" r"vim?:\s*.*(\s*|:)filetype=c(\s*|:|$))" +) # Match string that indicates we're working on a Linux Kernel file. -_SEARCH_KERNEL_FILE = re.compile(r'\b(?:LINT_KERNEL_FILE)') +_SEARCH_KERNEL_FILE = re.compile(r"\b(?:LINT_KERNEL_FILE)") _regexp_compile_cache = {} @@ -558,701 +577,737 @@ # The allowed extensions for file names # This is set by --extensions flag. -_valid_extensions = set(['cc', 'h', 'cpp', 'cu', 'cuh']) +_valid_extensions = set(["cc", "h", "cpp", "cu", "cuh"]) # Treat all headers starting with 'h' equally: .h, .hpp, .hxx etc. # This is set by --headers flag. -_hpp_headers = set(['h']) +_hpp_headers = set(["h"]) # {str, bool}: a map from error categories to booleans which indicate if the # category should be suppressed for every line. _global_error_suppressions = {} + def ProcessHppHeadersOption(val): - global _hpp_headers - try: - _hpp_headers = set(val.split(',')) - # Automatically append to extensions list so it does not have to be set 2 times - _valid_extensions.update(_hpp_headers) - except ValueError: - PrintUsage('Header extensions must be comma separated list.') + global _hpp_headers + try: + _hpp_headers = set(val.split(",")) + # Automatically append to extensions list so it does not have to be set 2 times + _valid_extensions.update(_hpp_headers) + except ValueError: + PrintUsage("Header extensions must be comma separated list.") + def IsHeaderExtension(file_extension): - return file_extension in _hpp_headers + return file_extension in _hpp_headers + def ParseNolintSuppressions(filename, raw_line, linenum, error): - """Updates the global list of line error-suppressions. - - Parses any NOLINT comments on the current line, updating the global - error_suppressions store. Reports an error if the NOLINT comment - was malformed. - - Args: - filename: str, the name of the input file. - raw_line: str, the line of input text, with comments. - linenum: int, the number of the current line. - error: function, an error handler. - """ - matched = Search(r'\bNOLINT(NEXTLINE)?\b(\([^)]+\))?', raw_line) - if matched: - if matched.group(1): - suppressed_line = linenum + 1 - else: - suppressed_line = linenum - category = matched.group(2) - if category in (None, '(*)'): # => "suppress all" - _error_suppressions.setdefault(None, set()).add(suppressed_line) - else: - if category.startswith('(') and category.endswith(')'): - category = category[1:-1] - if category in _ERROR_CATEGORIES: - _error_suppressions.setdefault(category, set()).add(suppressed_line) - elif category not in _LEGACY_ERROR_CATEGORIES: - error(filename, linenum, 'readability/nolint', 5, - 'Unknown NOLINT error category: %s' % category) + """Updates the global list of line error-suppressions. + + Parses any NOLINT comments on the current line, updating the global + error_suppressions store. Reports an error if the NOLINT comment + was malformed. + + Args: + filename: str, the name of the input file. + raw_line: str, the line of input text, with comments. + linenum: int, the number of the current line. + error: function, an error handler. + """ + matched = Search(r"\bNOLINT(NEXTLINE)?\b(\([^)]+\))?", raw_line) + if matched: + if matched.group(1): + suppressed_line = linenum + 1 + else: + suppressed_line = linenum + category = matched.group(2) + if category in (None, "(*)"): # => "suppress all" + _error_suppressions.setdefault(None, set()).add(suppressed_line) + else: + if category.startswith("(") and category.endswith(")"): + category = category[1:-1] + if category in _ERROR_CATEGORIES: + _error_suppressions.setdefault(category, set()).add(suppressed_line) + elif category not in _LEGACY_ERROR_CATEGORIES: + error( + filename, + linenum, + "readability/nolint", + 5, + "Unknown NOLINT error category: %s" % category, + ) def ProcessGlobalSuppresions(lines): - """Updates the list of global error suppressions. + """Updates the list of global error suppressions. - Parses any lint directives in the file that have global effect. + Parses any lint directives in the file that have global effect. - Args: - lines: An array of strings, each representing a line of the file, with the - last element being empty if the file is terminated with a newline. - """ - for line in lines: - if _SEARCH_C_FILE.search(line): - for category in _DEFAULT_C_SUPPRESSED_CATEGORIES: - _global_error_suppressions[category] = True - if _SEARCH_KERNEL_FILE.search(line): - for category in _DEFAULT_KERNEL_SUPPRESSED_CATEGORIES: - _global_error_suppressions[category] = True + Args: + lines: An array of strings, each representing a line of the file, with the + last element being empty if the file is terminated with a newline. + """ + for line in lines: + if _SEARCH_C_FILE.search(line): + for category in _DEFAULT_C_SUPPRESSED_CATEGORIES: + _global_error_suppressions[category] = True + if _SEARCH_KERNEL_FILE.search(line): + for category in _DEFAULT_KERNEL_SUPPRESSED_CATEGORIES: + _global_error_suppressions[category] = True def ResetNolintSuppressions(): - """Resets the set of NOLINT suppressions to empty.""" - _error_suppressions.clear() - _global_error_suppressions.clear() + """Resets the set of NOLINT suppressions to empty.""" + _error_suppressions.clear() + _global_error_suppressions.clear() def IsErrorSuppressedByNolint(category, linenum): - """Returns true if the specified error category is suppressed on this line. + """Returns true if the specified error category is suppressed on this line. - Consults the global error_suppressions map populated by - ParseNolintSuppressions/ProcessGlobalSuppresions/ResetNolintSuppressions. + Consults the global error_suppressions map populated by + ParseNolintSuppressions/ProcessGlobalSuppresions/ResetNolintSuppressions. - Args: - category: str, the category of the error. - linenum: int, the current line number. - Returns: - bool, True iff the error should be suppressed due to a NOLINT comment or - global suppression. - """ - return (_global_error_suppressions.get(category, False) or - linenum in _error_suppressions.get(category, set()) or - linenum in _error_suppressions.get(None, set())) + Args: + category: str, the category of the error. + linenum: int, the current line number. + Returns: + bool, True iff the error should be suppressed due to a NOLINT comment or + global suppression. + """ + return ( + _global_error_suppressions.get(category, False) + or linenum in _error_suppressions.get(category, set()) + or linenum in _error_suppressions.get(None, set()) + ) def Match(pattern, s): - """Matches the string with the pattern, caching the compiled regexp.""" - # The regexp compilation caching is inlined in both Match and Search for - # performance reasons; factoring it out into a separate function turns out - # to be noticeably expensive. - if pattern not in _regexp_compile_cache: - _regexp_compile_cache[pattern] = sre_compile.compile(pattern) - return _regexp_compile_cache[pattern].match(s) + """Matches the string with the pattern, caching the compiled regexp.""" + # The regexp compilation caching is inlined in both Match and Search for + # performance reasons; factoring it out into a separate function turns out + # to be noticeably expensive. + if pattern not in _regexp_compile_cache: + _regexp_compile_cache[pattern] = sre_compile.compile(pattern) + return _regexp_compile_cache[pattern].match(s) def ReplaceAll(pattern, rep, s): - """Replaces instances of pattern in a string with a replacement. + """Replaces instances of pattern in a string with a replacement. - The compiled regex is kept in a cache shared by Match and Search. + The compiled regex is kept in a cache shared by Match and Search. - Args: - pattern: regex pattern - rep: replacement text - s: search string + Args: + pattern: regex pattern + rep: replacement text + s: search string - Returns: - string with replacements made (or original string if no replacements) - """ - if pattern not in _regexp_compile_cache: - _regexp_compile_cache[pattern] = sre_compile.compile(pattern) - return _regexp_compile_cache[pattern].sub(rep, s) + Returns: + string with replacements made (or original string if no replacements) + """ + if pattern not in _regexp_compile_cache: + _regexp_compile_cache[pattern] = sre_compile.compile(pattern) + return _regexp_compile_cache[pattern].sub(rep, s) def Search(pattern, s): - """Searches the string for the pattern, caching the compiled regexp.""" - if pattern not in _regexp_compile_cache: - _regexp_compile_cache[pattern] = sre_compile.compile(pattern) - return _regexp_compile_cache[pattern].search(s) + """Searches the string for the pattern, caching the compiled regexp.""" + if pattern not in _regexp_compile_cache: + _regexp_compile_cache[pattern] = sre_compile.compile(pattern) + return _regexp_compile_cache[pattern].search(s) def _IsSourceExtension(s): - """File extension (excluding dot) matches a source file extension.""" - return s in ('c', 'cc', 'cpp', 'cxx') + """File extension (excluding dot) matches a source file extension.""" + return s in ("c", "cc", "cpp", "cxx") class _IncludeState(object): - """Tracks line numbers for includes, and the order in which includes appear. - - include_list contains list of lists of (header, line number) pairs. - It's a lists of lists rather than just one flat list to make it - easier to update across preprocessor boundaries. - - Call CheckNextIncludeOrder() once for each header in the file, passing - in the type constants defined above. Calls in an illegal order will - raise an _IncludeError with an appropriate error message. - - """ - # self._section will move monotonically through this set. If it ever - # needs to move backwards, CheckNextIncludeOrder will raise an error. - _INITIAL_SECTION = 0 - _MY_H_SECTION = 1 - _C_SECTION = 2 - _CPP_SECTION = 3 - _OTHER_H_SECTION = 4 - - _TYPE_NAMES = { - _C_SYS_HEADER: 'C system header', - _CPP_SYS_HEADER: 'C++ system header', - _LIKELY_MY_HEADER: 'header this file implements', - _POSSIBLE_MY_HEADER: 'header this file may implement', - _OTHER_HEADER: 'other header', - } - _SECTION_NAMES = { - _INITIAL_SECTION: "... nothing. (This can't be an error.)", - _MY_H_SECTION: 'a header this file implements', - _C_SECTION: 'C system header', - _CPP_SECTION: 'C++ system header', - _OTHER_H_SECTION: 'other header', - } - - def __init__(self): - self.include_list = [[]] - self.ResetSection('') - - def FindHeader(self, header): - """Check if a header has already been included. - - Args: - header: header to check. - Returns: - Line number of previous occurrence, or -1 if the header has not - been seen before. - """ - for section_list in self.include_list: - for f in section_list: - if f[0] == header: - return f[1] - return -1 - - def ResetSection(self, directive): - """Reset section checking for preprocessor directive. - - Args: - directive: preprocessor directive (e.g. "if", "else"). - """ - # The name of the current section. - self._section = self._INITIAL_SECTION - # The path of last found header. - self._last_header = '' - - # Update list of includes. Note that we never pop from the - # include list. - if directive in ('if', 'ifdef', 'ifndef'): - self.include_list.append([]) - elif directive in ('else', 'elif'): - self.include_list[-1] = [] - - def SetLastHeader(self, header_path): - self._last_header = header_path + """Tracks line numbers for includes, and the order in which includes appear. - def CanonicalizeAlphabeticalOrder(self, header_path): - """Returns a path canonicalized for alphabetical comparison. + include_list contains list of lists of (header, line number) pairs. + It's a lists of lists rather than just one flat list to make it + easier to update across preprocessor boundaries. - - replaces "-" with "_" so they both cmp the same. - - removes '-inl' since we don't require them to be after the main header. - - lowercase everything, just in case. - - Args: - header_path: Path to be canonicalized. - - Returns: - Canonicalized path. - """ - return header_path.replace('-inl.h', '.h').replace('-', '_').lower() - - def IsInAlphabeticalOrder(self, clean_lines, linenum, header_path): - """Check if a header is in alphabetical order with the previous header. - - Args: - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - header_path: Canonicalized header to be checked. + Call CheckNextIncludeOrder() once for each header in the file, passing + in the type constants defined above. Calls in an illegal order will + raise an _IncludeError with an appropriate error message. - Returns: - Returns true if the header is in alphabetical order. """ - # If previous section is different from current section, _last_header will - # be reset to empty string, so it's always less than current header. - # - # If previous line was a blank line, assume that the headers are - # intentionally sorted the way they are. - if (self._last_header > header_path and - Match(r'^\s*#\s*include\b', clean_lines.elided[linenum - 1])): - return False - return True - - def CheckNextIncludeOrder(self, header_type): - """Returns a non-empty error message if the next header is out of order. - This function also updates the internal state to be ready to check - the next include. - - Args: - header_type: One of the _XXX_HEADER constants defined above. + # self._section will move monotonically through this set. If it ever + # needs to move backwards, CheckNextIncludeOrder will raise an error. + _INITIAL_SECTION = 0 + _MY_H_SECTION = 1 + _C_SECTION = 2 + _CPP_SECTION = 3 + _OTHER_H_SECTION = 4 + + _TYPE_NAMES = { + _C_SYS_HEADER: "C system header", + _CPP_SYS_HEADER: "C++ system header", + _LIKELY_MY_HEADER: "header this file implements", + _POSSIBLE_MY_HEADER: "header this file may implement", + _OTHER_HEADER: "other header", + } + _SECTION_NAMES = { + _INITIAL_SECTION: "... nothing. (This can't be an error.)", + _MY_H_SECTION: "a header this file implements", + _C_SECTION: "C system header", + _CPP_SECTION: "C++ system header", + _OTHER_H_SECTION: "other header", + } - Returns: - The empty string if the header is in the right order, or an - error message describing what's wrong. + def __init__(self): + self.include_list = [[]] + self.ResetSection("") + + def FindHeader(self, header): + """Check if a header has already been included. + + Args: + header: header to check. + Returns: + Line number of previous occurrence, or -1 if the header has not + been seen before. + """ + for section_list in self.include_list: + for f in section_list: + if f[0] == header: + return f[1] + return -1 + + def ResetSection(self, directive): + """Reset section checking for preprocessor directive. + + Args: + directive: preprocessor directive (e.g. "if", "else"). + """ + # The name of the current section. + self._section = self._INITIAL_SECTION + # The path of last found header. + self._last_header = "" + + # Update list of includes. Note that we never pop from the + # include list. + if directive in ("if", "ifdef", "ifndef"): + self.include_list.append([]) + elif directive in ("else", "elif"): + self.include_list[-1] = [] + + def SetLastHeader(self, header_path): + self._last_header = header_path + + def CanonicalizeAlphabeticalOrder(self, header_path): + """Returns a path canonicalized for alphabetical comparison. + + - replaces "-" with "_" so they both cmp the same. + - removes '-inl' since we don't require them to be after the main header. + - lowercase everything, just in case. + + Args: + header_path: Path to be canonicalized. + + Returns: + Canonicalized path. + """ + return header_path.replace("-inl.h", ".h").replace("-", "_").lower() + + def IsInAlphabeticalOrder(self, clean_lines, linenum, header_path): + """Check if a header is in alphabetical order with the previous header. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + header_path: Canonicalized header to be checked. + + Returns: + Returns true if the header is in alphabetical order. + """ + # If previous section is different from current section, _last_header will + # be reset to empty string, so it's always less than current header. + # + # If previous line was a blank line, assume that the headers are + # intentionally sorted the way they are. + if self._last_header > header_path and Match( + r"^\s*#\s*include\b", clean_lines.elided[linenum - 1] + ): + return False + return True - """ - error_message = ('Found %s after %s' % - (self._TYPE_NAMES[header_type], - self._SECTION_NAMES[self._section])) - - last_section = self._section - - if header_type == _C_SYS_HEADER: - if self._section <= self._C_SECTION: - self._section = self._C_SECTION - else: - self._last_header = '' - return error_message - elif header_type == _CPP_SYS_HEADER: - if self._section <= self._CPP_SECTION: - self._section = self._CPP_SECTION - else: - self._last_header = '' - return error_message - elif header_type == _LIKELY_MY_HEADER: - if self._section <= self._MY_H_SECTION: - self._section = self._MY_H_SECTION - else: - self._section = self._OTHER_H_SECTION - elif header_type == _POSSIBLE_MY_HEADER: - if self._section <= self._MY_H_SECTION: - self._section = self._MY_H_SECTION - else: - # This will always be the fallback because we're not sure - # enough that the header is associated with this file. - self._section = self._OTHER_H_SECTION - else: - assert header_type == _OTHER_HEADER - self._section = self._OTHER_H_SECTION + def CheckNextIncludeOrder(self, header_type): + """Returns a non-empty error message if the next header is out of order. + + This function also updates the internal state to be ready to check + the next include. + + Args: + header_type: One of the _XXX_HEADER constants defined above. + + Returns: + The empty string if the header is in the right order, or an + error message describing what's wrong. + + """ + error_message = "Found %s after %s" % ( + self._TYPE_NAMES[header_type], + self._SECTION_NAMES[self._section], + ) + + last_section = self._section + + if header_type == _C_SYS_HEADER: + if self._section <= self._C_SECTION: + self._section = self._C_SECTION + else: + self._last_header = "" + return error_message + elif header_type == _CPP_SYS_HEADER: + if self._section <= self._CPP_SECTION: + self._section = self._CPP_SECTION + else: + self._last_header = "" + return error_message + elif header_type == _LIKELY_MY_HEADER: + if self._section <= self._MY_H_SECTION: + self._section = self._MY_H_SECTION + else: + self._section = self._OTHER_H_SECTION + elif header_type == _POSSIBLE_MY_HEADER: + if self._section <= self._MY_H_SECTION: + self._section = self._MY_H_SECTION + else: + # This will always be the fallback because we're not sure + # enough that the header is associated with this file. + self._section = self._OTHER_H_SECTION + else: + assert header_type == _OTHER_HEADER + self._section = self._OTHER_H_SECTION - if last_section != self._section: - self._last_header = '' + if last_section != self._section: + self._last_header = "" - return '' + return "" class _CppLintState(object): - """Maintains module-wide state..""" - - def __init__(self): - self.verbose_level = 1 # global setting. - self.error_count = 0 # global count of reported errors - # filters to apply when emitting error messages - self.filters = _DEFAULT_FILTERS[:] - # backup of filter list. Used to restore the state after each file. - self._filters_backup = self.filters[:] - self.counting = 'total' # In what way are we counting errors? - self.errors_by_category = {} # string to int dict storing error counts - self.quiet = False # Suppress non-error messagess? - - # output format: - # "emacs" - format that emacs can parse (default) - # "vs7" - format that Microsoft Visual Studio 7 can parse - self.output_format = 'emacs' - - def SetOutputFormat(self, output_format): - """Sets the output format for errors.""" - self.output_format = output_format - - def SetQuiet(self, quiet): - """Sets the module's quiet settings, and returns the previous setting.""" - last_quiet = self.quiet - self.quiet = quiet - return last_quiet - - def SetVerboseLevel(self, level): - """Sets the module's verbosity, and returns the previous setting.""" - last_verbose_level = self.verbose_level - self.verbose_level = level - return last_verbose_level - - def SetCountingStyle(self, counting_style): - """Sets the module's counting options.""" - self.counting = counting_style - - def SetFilters(self, filters): - """Sets the error-message filters. - - These filters are applied when deciding whether to emit a given - error message. - - Args: - filters: A string of comma-separated filters (eg "+whitespace/indent"). - Each filter should start with + or -; else we die. + """Maintains module-wide state..""" + + def __init__(self): + self.verbose_level = 1 # global setting. + self.error_count = 0 # global count of reported errors + # filters to apply when emitting error messages + self.filters = _DEFAULT_FILTERS[:] + # backup of filter list. Used to restore the state after each file. + self._filters_backup = self.filters[:] + self.counting = "total" # In what way are we counting errors? + self.errors_by_category = {} # string to int dict storing error counts + self.quiet = False # Suppress non-error messagess? + + # output format: + # "emacs" - format that emacs can parse (default) + # "vs7" - format that Microsoft Visual Studio 7 can parse + self.output_format = "emacs" + + def SetOutputFormat(self, output_format): + """Sets the output format for errors.""" + self.output_format = output_format + + def SetQuiet(self, quiet): + """Sets the module's quiet settings, and returns the previous setting.""" + last_quiet = self.quiet + self.quiet = quiet + return last_quiet + + def SetVerboseLevel(self, level): + """Sets the module's verbosity, and returns the previous setting.""" + last_verbose_level = self.verbose_level + self.verbose_level = level + return last_verbose_level + + def SetCountingStyle(self, counting_style): + """Sets the module's counting options.""" + self.counting = counting_style + + def SetFilters(self, filters): + """Sets the error-message filters. + + These filters are applied when deciding whether to emit a given + error message. + + Args: + filters: A string of comma-separated filters (eg "+whitespace/indent"). + Each filter should start with + or -; else we die. + + Raises: + ValueError: The comma-separated filters did not all start with '+' or '-'. + E.g. "-,+whitespace,-whitespace/indent,whitespace/badfilter" + """ + # Default filters always have less priority than the flag ones. + self.filters = _DEFAULT_FILTERS[:] + self.AddFilters(filters) + + def AddFilters(self, filters): + """Adds more filters to the existing list of error-message filters.""" + for filt in filters.split(","): + clean_filt = filt.strip() + if clean_filt: + self.filters.append(clean_filt) + for filt in self.filters: + if not (filt.startswith("+") or filt.startswith("-")): + raise ValueError( + "Every filter in --filters must start with + or -" + " (%s does not)" % filt + ) + + def BackupFilters(self): + """Saves the current filter list to backup storage.""" + self._filters_backup = self.filters[:] + + def RestoreFilters(self): + """Restores filters previously backed up.""" + self.filters = self._filters_backup[:] + + def ResetErrorCounts(self): + """Sets the module's error statistic back to zero.""" + self.error_count = 0 + self.errors_by_category = {} + + def IncrementErrorCount(self, category): + """Bumps the module's error statistic.""" + self.error_count += 1 + if self.counting in ("toplevel", "detailed"): + if self.counting != "detailed": + category = category.split("/")[0] + if category not in self.errors_by_category: + self.errors_by_category[category] = 0 + self.errors_by_category[category] += 1 + + def PrintErrorCounts(self): + """Print a summary of errors by category, and the total.""" + for category, count in self.errors_by_category.iteritems(): + sys.stderr.write("Category '%s' errors found: %d\n" % (category, count)) + sys.stdout.write("Total errors found: %d\n" % self.error_count) - Raises: - ValueError: The comma-separated filters did not all start with '+' or '-'. - E.g. "-,+whitespace,-whitespace/indent,whitespace/badfilter" - """ - # Default filters always have less priority than the flag ones. - self.filters = _DEFAULT_FILTERS[:] - self.AddFilters(filters) - - def AddFilters(self, filters): - """ Adds more filters to the existing list of error-message filters. """ - for filt in filters.split(','): - clean_filt = filt.strip() - if clean_filt: - self.filters.append(clean_filt) - for filt in self.filters: - if not (filt.startswith('+') or filt.startswith('-')): - raise ValueError('Every filter in --filters must start with + or -' - ' (%s does not)' % filt) - - def BackupFilters(self): - """ Saves the current filter list to backup storage.""" - self._filters_backup = self.filters[:] - - def RestoreFilters(self): - """ Restores filters previously backed up.""" - self.filters = self._filters_backup[:] - - def ResetErrorCounts(self): - """Sets the module's error statistic back to zero.""" - self.error_count = 0 - self.errors_by_category = {} - - def IncrementErrorCount(self, category): - """Bumps the module's error statistic.""" - self.error_count += 1 - if self.counting in ('toplevel', 'detailed'): - if self.counting != 'detailed': - category = category.split('/')[0] - if category not in self.errors_by_category: - self.errors_by_category[category] = 0 - self.errors_by_category[category] += 1 - - def PrintErrorCounts(self): - """Print a summary of errors by category, and the total.""" - for category, count in self.errors_by_category.iteritems(): - sys.stderr.write('Category \'%s\' errors found: %d\n' % - (category, count)) - sys.stdout.write('Total errors found: %d\n' % self.error_count) _cpplint_state = _CppLintState() def _OutputFormat(): - """Gets the module's output format.""" - return _cpplint_state.output_format + """Gets the module's output format.""" + return _cpplint_state.output_format def _SetOutputFormat(output_format): - """Sets the module's output format.""" - _cpplint_state.SetOutputFormat(output_format) + """Sets the module's output format.""" + _cpplint_state.SetOutputFormat(output_format) + def _Quiet(): - """Return's the module's quiet setting.""" - return _cpplint_state.quiet + """Return's the module's quiet setting.""" + return _cpplint_state.quiet + def _SetQuiet(quiet): - """Set the module's quiet status, and return previous setting.""" - return _cpplint_state.SetQuiet(quiet) + """Set the module's quiet status, and return previous setting.""" + return _cpplint_state.SetQuiet(quiet) def _VerboseLevel(): - """Returns the module's verbosity setting.""" - return _cpplint_state.verbose_level + """Returns the module's verbosity setting.""" + return _cpplint_state.verbose_level def _SetVerboseLevel(level): - """Sets the module's verbosity, and returns the previous setting.""" - return _cpplint_state.SetVerboseLevel(level) + """Sets the module's verbosity, and returns the previous setting.""" + return _cpplint_state.SetVerboseLevel(level) def _SetCountingStyle(level): - """Sets the module's counting options.""" - _cpplint_state.SetCountingStyle(level) + """Sets the module's counting options.""" + _cpplint_state.SetCountingStyle(level) def _Filters(): - """Returns the module's list of output filters, as a list.""" - return _cpplint_state.filters + """Returns the module's list of output filters, as a list.""" + return _cpplint_state.filters def _SetFilters(filters): - """Sets the module's error-message filters. - - These filters are applied when deciding whether to emit a given - error message. + """Sets the module's error-message filters. - Args: - filters: A string of comma-separated filters (eg "whitespace/indent"). - Each filter should start with + or -; else we die. - """ - _cpplint_state.SetFilters(filters) - -def _AddFilters(filters): - """Adds more filter overrides. - - Unlike _SetFilters, this function does not reset the current list of filters - available. - - Args: - filters: A string of comma-separated filters (eg "whitespace/indent"). - Each filter should start with + or -; else we die. - """ - _cpplint_state.AddFilters(filters) - -def _BackupFilters(): - """ Saves the current filter list to backup storage.""" - _cpplint_state.BackupFilters() - -def _RestoreFilters(): - """ Restores filters previously backed up.""" - _cpplint_state.RestoreFilters() - -class _FunctionState(object): - """Tracks current function name and the number of lines in its body.""" - - _NORMAL_TRIGGER = 250 # for --v=0, 500 for --v=1, etc. - _TEST_TRIGGER = 400 # about 50% more than _NORMAL_TRIGGER. - - def __init__(self): - self.in_a_function = False - self.lines_in_function = 0 - self.current_function = '' - - def Begin(self, function_name): - """Start analyzing function body. + These filters are applied when deciding whether to emit a given + error message. Args: - function_name: The name of the function being tracked. + filters: A string of comma-separated filters (eg "whitespace/indent"). + Each filter should start with + or -; else we die. """ - self.in_a_function = True - self.lines_in_function = 0 - self.current_function = function_name + _cpplint_state.SetFilters(filters) - def Count(self): - """Count line in current function body.""" - if self.in_a_function: - self.lines_in_function += 1 - def Check(self, error, filename, linenum): - """Report if too many lines in function body. +def _AddFilters(filters): + """Adds more filter overrides. + + Unlike _SetFilters, this function does not reset the current list of filters + available. Args: - error: The function to call with any errors found. - filename: The name of the current file. - linenum: The number of the line to check. + filters: A string of comma-separated filters (eg "whitespace/indent"). + Each filter should start with + or -; else we die. """ - if not self.in_a_function: - return - - if Match(r'T(EST|est)', self.current_function): - base_trigger = self._TEST_TRIGGER - else: - base_trigger = self._NORMAL_TRIGGER - trigger = base_trigger * 2**_VerboseLevel() + _cpplint_state.AddFilters(filters) - if self.lines_in_function > trigger: - error_level = int(math.log(self.lines_in_function / base_trigger, 2)) - # 50 => 0, 100 => 1, 200 => 2, 400 => 3, 800 => 4, 1600 => 5, ... - if error_level > 5: - error_level = 5 - error(filename, linenum, 'readability/fn_size', error_level, - 'Small and focused functions are preferred:' - ' %s has %d non-comment lines' - ' (error triggered by exceeding %d lines).' % ( - self.current_function, self.lines_in_function, trigger)) - def End(self): - """Stop analyzing function body.""" - self.in_a_function = False +def _BackupFilters(): + """Saves the current filter list to backup storage.""" + _cpplint_state.BackupFilters() -class _IncludeError(Exception): - """Indicates a problem with the include order in a file.""" - pass +def _RestoreFilters(): + """Restores filters previously backed up.""" + _cpplint_state.RestoreFilters() -class FileInfo(object): - """Provides utility functions for filenames. +class _FunctionState(object): + """Tracks current function name and the number of lines in its body.""" + + _NORMAL_TRIGGER = 250 # for --v=0, 500 for --v=1, etc. + _TEST_TRIGGER = 400 # about 50% more than _NORMAL_TRIGGER. + + def __init__(self): + self.in_a_function = False + self.lines_in_function = 0 + self.current_function = "" + + def Begin(self, function_name): + """Start analyzing function body. + + Args: + function_name: The name of the function being tracked. + """ + self.in_a_function = True + self.lines_in_function = 0 + self.current_function = function_name + + def Count(self): + """Count line in current function body.""" + if self.in_a_function: + self.lines_in_function += 1 + + def Check(self, error, filename, linenum): + """Report if too many lines in function body. + + Args: + error: The function to call with any errors found. + filename: The name of the current file. + linenum: The number of the line to check. + """ + if not self.in_a_function: + return + + if Match(r"T(EST|est)", self.current_function): + base_trigger = self._TEST_TRIGGER + else: + base_trigger = self._NORMAL_TRIGGER + trigger = base_trigger * 2 ** _VerboseLevel() + + if self.lines_in_function > trigger: + error_level = int(math.log(self.lines_in_function / base_trigger, 2)) + # 50 => 0, 100 => 1, 200 => 2, 400 => 3, 800 => 4, 1600 => 5, ... + if error_level > 5: + error_level = 5 + error( + filename, + linenum, + "readability/fn_size", + error_level, + "Small and focused functions are preferred:" + " %s has %d non-comment lines" + " (error triggered by exceeding %d lines)." + % (self.current_function, self.lines_in_function, trigger), + ) + + def End(self): + """Stop analyzing function body.""" + self.in_a_function = False - FileInfo provides easy access to the components of a file's path - relative to the project root. - """ - def __init__(self, filename): - self._filename = filename +class _IncludeError(Exception): + """Indicates a problem with the include order in a file.""" - def FullName(self): - """Make Windows paths like Unix.""" - return os.path.abspath(self._filename).replace('\\', '/') + pass - def RepositoryName(self): - """FullName after removing the local path to the repository. - If we have a real absolute path name here we can try to do something smart: - detecting the root of the checkout and truncating /path/to/checkout from - the name so that we get header guards that don't include things like - "C:\Documents and Settings\..." or "/home/username/..." in them and thus - people on different computers who have checked the source out to different - locations won't see bogus errors. - """ - fullname = self.FullName() - - if os.path.exists(fullname): - project_dir = os.path.dirname(fullname) - - if os.path.exists(os.path.join(project_dir, ".svn")): - # If there's a .svn file in the current directory, we recursively look - # up the directory tree for the top of the SVN checkout - root_dir = project_dir - one_up_dir = os.path.dirname(root_dir) - while os.path.exists(os.path.join(one_up_dir, ".svn")): - root_dir = os.path.dirname(root_dir) - one_up_dir = os.path.dirname(one_up_dir) - - prefix = os.path.commonprefix([root_dir, project_dir]) - return fullname[len(prefix) + 1:] - - # Not SVN <= 1.6? Try to find a git, hg, or svn top level directory by - # searching up from the current path. - root_dir = current_dir = os.path.dirname(fullname) - while current_dir != os.path.dirname(current_dir): - if (os.path.exists(os.path.join(current_dir, ".git")) or - os.path.exists(os.path.join(current_dir, ".hg")) or - os.path.exists(os.path.join(current_dir, ".svn"))): - root_dir = current_dir - current_dir = os.path.dirname(current_dir) - - if (os.path.exists(os.path.join(root_dir, ".git")) or - os.path.exists(os.path.join(root_dir, ".hg")) or - os.path.exists(os.path.join(root_dir, ".svn"))): - prefix = os.path.commonprefix([root_dir, project_dir]) - return fullname[len(prefix) + 1:] - - # Don't know what to do; header guard warnings may be wrong... - return fullname - - def Split(self): - """Splits the file into the directory, basename, and extension. - - For 'chrome/browser/browser.cc', Split() would - return ('chrome/browser', 'browser', '.cc') +class FileInfo(object): + """Provides utility functions for filenames. - Returns: - A tuple of (directory, basename, extension). + FileInfo provides easy access to the components of a file's path + relative to the project root. """ - googlename = self.RepositoryName() - project, rest = os.path.split(googlename) - return (project,) + os.path.splitext(rest) - - def BaseName(self): - """File base name - text after the final slash, before the final period.""" - return self.Split()[1] + def __init__(self, filename): + self._filename = filename + + def FullName(self): + """Make Windows paths like Unix.""" + return os.path.abspath(self._filename).replace("\\", "/") + + def RepositoryName(self): + """FullName after removing the local path to the repository. + + If we have a real absolute path name here we can try to do something smart: + detecting the root of the checkout and truncating /path/to/checkout from + the name so that we get header guards that don't include things like + "C:\Documents and Settings\..." or "/home/username/..." in them and thus + people on different computers who have checked the source out to different + locations won't see bogus errors. + """ + fullname = self.FullName() + + if os.path.exists(fullname): + project_dir = os.path.dirname(fullname) + + if os.path.exists(os.path.join(project_dir, ".svn")): + # If there's a .svn file in the current directory, we recursively look + # up the directory tree for the top of the SVN checkout + root_dir = project_dir + one_up_dir = os.path.dirname(root_dir) + while os.path.exists(os.path.join(one_up_dir, ".svn")): + root_dir = os.path.dirname(root_dir) + one_up_dir = os.path.dirname(one_up_dir) + + prefix = os.path.commonprefix([root_dir, project_dir]) + return fullname[len(prefix) + 1 :] + + # Not SVN <= 1.6? Try to find a git, hg, or svn top level directory by + # searching up from the current path. + root_dir = current_dir = os.path.dirname(fullname) + while current_dir != os.path.dirname(current_dir): + if ( + os.path.exists(os.path.join(current_dir, ".git")) + or os.path.exists(os.path.join(current_dir, ".hg")) + or os.path.exists(os.path.join(current_dir, ".svn")) + ): + root_dir = current_dir + current_dir = os.path.dirname(current_dir) + + if ( + os.path.exists(os.path.join(root_dir, ".git")) + or os.path.exists(os.path.join(root_dir, ".hg")) + or os.path.exists(os.path.join(root_dir, ".svn")) + ): + prefix = os.path.commonprefix([root_dir, project_dir]) + return fullname[len(prefix) + 1 :] + + # Don't know what to do; header guard warnings may be wrong... + return fullname + + def Split(self): + """Splits the file into the directory, basename, and extension. + + For 'chrome/browser/browser.cc', Split() would + return ('chrome/browser', 'browser', '.cc') + + Returns: + A tuple of (directory, basename, extension). + """ + + googlename = self.RepositoryName() + project, rest = os.path.split(googlename) + return (project,) + os.path.splitext(rest) + + def BaseName(self): + """File base name - text after the final slash, before the final period.""" + return self.Split()[1] + + def Extension(self): + """File extension - text following the final period.""" + return self.Split()[2] + + def NoExtension(self): + """File has no source file extension.""" + return "/".join(self.Split()[0:2]) + + def IsSource(self): + """File has a source file extension.""" + return _IsSourceExtension(self.Extension()[1:]) - def Extension(self): - """File extension - text following the final period.""" - return self.Split()[2] - def NoExtension(self): - """File has no source file extension.""" - return '/'.join(self.Split()[0:2]) +def _ShouldPrintError(category, confidence, linenum): + """If confidence >= verbose, category passes filter and is not suppressed.""" - def IsSource(self): - """File has a source file extension.""" - return _IsSourceExtension(self.Extension()[1:]) + # There are three ways we might decide not to print an error message: + # a "NOLINT(category)" comment appears in the source, + # the verbosity level isn't high enough, or the filters filter it out. + if IsErrorSuppressedByNolint(category, linenum): + return False + if confidence < _cpplint_state.verbose_level: + return False -def _ShouldPrintError(category, confidence, linenum): - """If confidence >= verbose, category passes filter and is not suppressed.""" + is_filtered = False + for one_filter in _Filters(): + if one_filter.startswith("-"): + if category.startswith(one_filter[1:]): + is_filtered = True + elif one_filter.startswith("+"): + if category.startswith(one_filter[1:]): + is_filtered = False + else: + assert False # should have been checked for in SetFilter. + if is_filtered: + return False - # There are three ways we might decide not to print an error message: - # a "NOLINT(category)" comment appears in the source, - # the verbosity level isn't high enough, or the filters filter it out. - if IsErrorSuppressedByNolint(category, linenum): - return False + return True - if confidence < _cpplint_state.verbose_level: - return False - is_filtered = False - for one_filter in _Filters(): - if one_filter.startswith('-'): - if category.startswith(one_filter[1:]): - is_filtered = True - elif one_filter.startswith('+'): - if category.startswith(one_filter[1:]): - is_filtered = False - else: - assert False # should have been checked for in SetFilter. - if is_filtered: - return False +def Error(filename, linenum, category, confidence, message): + """Logs the fact we've found a lint error. - return True + We log where the error was found, and also our confidence in the error, + that is, how certain we are this is a legitimate style regression, and + not a misidentification or a use that's sometimes justified. + False positives can be suppressed by the use of + "cpplint(category)" comments on the offending line. These are + parsed into _error_suppressions. -def Error(filename, linenum, category, confidence, message): - """Logs the fact we've found a lint error. - - We log where the error was found, and also our confidence in the error, - that is, how certain we are this is a legitimate style regression, and - not a misidentification or a use that's sometimes justified. - - False positives can be suppressed by the use of - "cpplint(category)" comments on the offending line. These are - parsed into _error_suppressions. - - Args: - filename: The name of the file containing the error. - linenum: The number of the line containing the error. - category: A string used to describe the "category" this bug - falls under: "whitespace", say, or "runtime". Categories - may have a hierarchy separated by slashes: "whitespace/indent". - confidence: A number from 1-5 representing a confidence score for - the error, with 5 meaning that we are certain of the problem, - and 1 meaning that it could be a legitimate construct. - message: The error message. - """ - if _ShouldPrintError(category, confidence, linenum): - _cpplint_state.IncrementErrorCount(category) - if _cpplint_state.output_format == 'vs7': - sys.stderr.write('%s(%s): error cpplint: [%s] %s [%d]\n' % ( - filename, linenum, category, message, confidence)) - elif _cpplint_state.output_format == 'eclipse': - sys.stderr.write('%s:%s: warning: %s [%s] [%d]\n' % ( - filename, linenum, message, category, confidence)) - else: - sys.stderr.write('%s:%s: %s [%s] [%d]\n' % ( - filename, linenum, message, category, confidence)) + Args: + filename: The name of the file containing the error. + linenum: The number of the line containing the error. + category: A string used to describe the "category" this bug + falls under: "whitespace", say, or "runtime". Categories + may have a hierarchy separated by slashes: "whitespace/indent". + confidence: A number from 1-5 representing a confidence score for + the error, with 5 meaning that we are certain of the problem, + and 1 meaning that it could be a legitimate construct. + message: The error message. + """ + if _ShouldPrintError(category, confidence, linenum): + _cpplint_state.IncrementErrorCount(category) + if _cpplint_state.output_format == "vs7": + sys.stderr.write( + "%s(%s): error cpplint: [%s] %s [%d]\n" + % (filename, linenum, category, message, confidence) + ) + elif _cpplint_state.output_format == "eclipse": + sys.stderr.write( + "%s:%s: warning: %s [%s] [%d]\n" + % (filename, linenum, message, category, confidence) + ) + else: + sys.stderr.write( + "%s:%s: %s [%s] [%d]\n" + % (filename, linenum, message, category, confidence) + ) # Matches standard C++ escape sequences per 2.13.2.3 of the C++ standard. -_RE_PATTERN_CLEANSE_LINE_ESCAPES = re.compile( - r'\\([abfnrtv?"\\\']|\d+|x[0-9a-fA-F]+)') +_RE_PATTERN_CLEANSE_LINE_ESCAPES = re.compile(r'\\([abfnrtv?"\\\']|\d+|x[0-9a-fA-F]+)') # Match a single C style comment on the same line. -_RE_PATTERN_C_COMMENTS = r'/\*(?:[^*]|\*(?!/))*\*/' +_RE_PATTERN_C_COMMENTS = r"/\*(?:[^*]|\*(?!/))*\*/" # Matches multi-line C style comments. # This RE is a little bit more complicated than one might expect, because we # have to take care of space removals tools so we can handle comments inside @@ -1262,828 +1317,916 @@ def Error(filename, linenum, category, confidence, message): # if this doesn't work we try on left side but only if there's a non-character # on the right. _RE_PATTERN_CLEANSE_LINE_C_COMMENTS = re.compile( - r'(\s*' + _RE_PATTERN_C_COMMENTS + r'\s*$|' + - _RE_PATTERN_C_COMMENTS + r'\s+|' + - r'\s+' + _RE_PATTERN_C_COMMENTS + r'(?=\W)|' + - _RE_PATTERN_C_COMMENTS + r')') + r"(\s*" + + _RE_PATTERN_C_COMMENTS + + r"\s*$|" + + _RE_PATTERN_C_COMMENTS + + r"\s+|" + + r"\s+" + + _RE_PATTERN_C_COMMENTS + + r"(?=\W)|" + + _RE_PATTERN_C_COMMENTS + + r")" +) def IsCppString(line): - """Does line terminate so, that the next symbol is in string constant. + """Does line terminate so, that the next symbol is in string constant. - This function does not consider single-line nor multi-line comments. + This function does not consider single-line nor multi-line comments. - Args: - line: is a partial line of code starting from the 0..n. + Args: + line: is a partial line of code starting from the 0..n. - Returns: - True, if next character appended to 'line' is inside a - string constant. - """ + Returns: + True, if next character appended to 'line' is inside a + string constant. + """ - line = line.replace(r'\\', 'XX') # after this, \\" does not match to \" - return ((line.count('"') - line.count(r'\"') - line.count("'\"'")) & 1) == 1 + line = line.replace(r"\\", "XX") # after this, \\" does not match to \" + return ((line.count('"') - line.count(r"\"") - line.count("'\"'")) & 1) == 1 def CleanseRawStrings(raw_lines): - """Removes C++11 raw strings from lines. - - Before: - static const char kData[] = R"( - multi-line string - )"; - - After: - static const char kData[] = "" - (replaced by blank line) - ""; - - Args: - raw_lines: list of raw lines. - - Returns: - list of lines with C++11 raw strings replaced by empty strings. - """ - - delimiter = None - lines_without_raw_strings = [] - for line in raw_lines: - if delimiter: - # Inside a raw string, look for the end - end = line.find(delimiter) - if end >= 0: - # Found the end of the string, match leading space for this - # line and resume copying the original lines, and also insert - # a "" on the last line. - leading_space = Match(r'^(\s*)\S', line) - line = leading_space.group(1) + '""' + line[end + len(delimiter):] - delimiter = None - else: - # Haven't found the end yet, append a blank line. - line = '""' - - # Look for beginning of a raw string, and replace them with - # empty strings. This is done in a loop to handle multiple raw - # strings on the same line. - while delimiter is None: - # Look for beginning of a raw string. - # See 2.14.15 [lex.string] for syntax. - # - # Once we have matched a raw string, we check the prefix of the - # line to make sure that the line is not part of a single line - # comment. It's done this way because we remove raw strings - # before removing comments as opposed to removing comments - # before removing raw strings. This is because there are some - # cpplint checks that requires the comments to be preserved, but - # we don't want to check comments that are inside raw strings. - matched = Match(r'^(.*?)\b(?:R|u8R|uR|UR|LR)"([^\s\\()]*)\((.*)$', line) - if (matched and - not Match(r'^([^\'"]|\'(\\.|[^\'])*\'|"(\\.|[^"])*")*//', - matched.group(1))): - delimiter = ')' + matched.group(2) + '"' - - end = matched.group(3).find(delimiter) - if end >= 0: - # Raw string ended on same line - line = (matched.group(1) + '""' + - matched.group(3)[end + len(delimiter):]) - delimiter = None - else: - # Start of a multi-line raw string - line = matched.group(1) + '""' - else: - break + """Removes C++11 raw strings from lines. + + Before: + static const char kData[] = R"( + multi-line string + )"; + + After: + static const char kData[] = "" + (replaced by blank line) + ""; + + Args: + raw_lines: list of raw lines. - lines_without_raw_strings.append(line) + Returns: + list of lines with C++11 raw strings replaced by empty strings. + """ - # TODO(unknown): if delimiter is not None here, we might want to - # emit a warning for unterminated string. - return lines_without_raw_strings + delimiter = None + lines_without_raw_strings = [] + for line in raw_lines: + if delimiter: + # Inside a raw string, look for the end + end = line.find(delimiter) + if end >= 0: + # Found the end of the string, match leading space for this + # line and resume copying the original lines, and also insert + # a "" on the last line. + leading_space = Match(r"^(\s*)\S", line) + line = leading_space.group(1) + '""' + line[end + len(delimiter) :] + delimiter = None + else: + # Haven't found the end yet, append a blank line. + line = '""' + + # Look for beginning of a raw string, and replace them with + # empty strings. This is done in a loop to handle multiple raw + # strings on the same line. + while delimiter is None: + # Look for beginning of a raw string. + # See 2.14.15 [lex.string] for syntax. + # + # Once we have matched a raw string, we check the prefix of the + # line to make sure that the line is not part of a single line + # comment. It's done this way because we remove raw strings + # before removing comments as opposed to removing comments + # before removing raw strings. This is because there are some + # cpplint checks that requires the comments to be preserved, but + # we don't want to check comments that are inside raw strings. + matched = Match(r'^(.*?)\b(?:R|u8R|uR|UR|LR)"([^\s\\()]*)\((.*)$', line) + if matched and not Match( + r'^([^\'"]|\'(\\.|[^\'])*\'|"(\\.|[^"])*")*//', matched.group(1) + ): + delimiter = ")" + matched.group(2) + '"' + + end = matched.group(3).find(delimiter) + if end >= 0: + # Raw string ended on same line + line = ( + matched.group(1) + + '""' + + matched.group(3)[end + len(delimiter) :] + ) + delimiter = None + else: + # Start of a multi-line raw string + line = matched.group(1) + '""' + else: + break + + lines_without_raw_strings.append(line) + + # TODO(unknown): if delimiter is not None here, we might want to + # emit a warning for unterminated string. + return lines_without_raw_strings def FindNextMultiLineCommentStart(lines, lineix): - """Find the beginning marker for a multiline comment.""" - while lineix < len(lines): - if lines[lineix].strip().startswith('/*'): - # Only return this marker if the comment goes beyond this line - if lines[lineix].strip().find('*/', 2) < 0: - return lineix - lineix += 1 - return len(lines) + """Find the beginning marker for a multiline comment.""" + while lineix < len(lines): + if lines[lineix].strip().startswith("/*"): + # Only return this marker if the comment goes beyond this line + if lines[lineix].strip().find("*/", 2) < 0: + return lineix + lineix += 1 + return len(lines) def FindNextMultiLineCommentEnd(lines, lineix): - """We are inside a comment, find the end marker.""" - while lineix < len(lines): - if lines[lineix].strip().endswith('*/'): - return lineix - lineix += 1 - return len(lines) + """We are inside a comment, find the end marker.""" + while lineix < len(lines): + if lines[lineix].strip().endswith("*/"): + return lineix + lineix += 1 + return len(lines) def RemoveMultiLineCommentsFromRange(lines, begin, end): - """Clears a range of lines for multi-line comments.""" - # Having // dummy comments makes the lines non-empty, so we will not get - # unnecessary blank line warnings later in the code. - for i in range(begin, end): - lines[i] = '/**/' + """Clears a range of lines for multi-line comments.""" + # Having // dummy comments makes the lines non-empty, so we will not get + # unnecessary blank line warnings later in the code. + for i in range(begin, end): + lines[i] = "/**/" def RemoveMultiLineComments(filename, lines, error): - """Removes multiline (c-style) comments from lines.""" - lineix = 0 - while lineix < len(lines): - lineix_begin = FindNextMultiLineCommentStart(lines, lineix) - if lineix_begin >= len(lines): - return - lineix_end = FindNextMultiLineCommentEnd(lines, lineix_begin) - if lineix_end >= len(lines): - error(filename, lineix_begin + 1, 'readability/multiline_comment', 5, - 'Could not find end of multi-line comment') - return - RemoveMultiLineCommentsFromRange(lines, lineix_begin, lineix_end + 1) - lineix = lineix_end + 1 + """Removes multiline (c-style) comments from lines.""" + lineix = 0 + while lineix < len(lines): + lineix_begin = FindNextMultiLineCommentStart(lines, lineix) + if lineix_begin >= len(lines): + return + lineix_end = FindNextMultiLineCommentEnd(lines, lineix_begin) + if lineix_end >= len(lines): + error( + filename, + lineix_begin + 1, + "readability/multiline_comment", + 5, + "Could not find end of multi-line comment", + ) + return + RemoveMultiLineCommentsFromRange(lines, lineix_begin, lineix_end + 1) + lineix = lineix_end + 1 def CleanseComments(line): - """Removes //-comments and single-line C-style /* */ comments. + """Removes //-comments and single-line C-style /* */ comments. - Args: - line: A line of C++ source. + Args: + line: A line of C++ source. - Returns: - The line with single-line comments removed. - """ - commentpos = line.find('//') - if commentpos != -1 and not IsCppString(line[:commentpos]): - line = line[:commentpos].rstrip() - # get rid of /* ... */ - return _RE_PATTERN_CLEANSE_LINE_C_COMMENTS.sub('', line) + Returns: + The line with single-line comments removed. + """ + commentpos = line.find("//") + if commentpos != -1 and not IsCppString(line[:commentpos]): + line = line[:commentpos].rstrip() + # get rid of /* ... */ + return _RE_PATTERN_CLEANSE_LINE_C_COMMENTS.sub("", line) class CleansedLines(object): - """Holds 4 copies of all lines with different preprocessing applied to them. - - 1) elided member contains lines without strings and comments. - 2) lines member contains lines without comments. - 3) raw_lines member contains all the lines without processing. - 4) lines_without_raw_strings member is same as raw_lines, but with C++11 raw - strings removed. - All these members are of , and of the same length. - """ - - def __init__(self, lines): - self.elided = [] - self.lines = [] - self.raw_lines = lines - self.num_lines = len(lines) - self.lines_without_raw_strings = CleanseRawStrings(lines) - for linenum in range(len(self.lines_without_raw_strings)): - self.lines.append(CleanseComments( - self.lines_without_raw_strings[linenum])) - elided = self._CollapseStrings(self.lines_without_raw_strings[linenum]) - self.elided.append(CleanseComments(elided)) - - def NumLines(self): - """Returns the number of lines represented.""" - return self.num_lines - - @staticmethod - def _CollapseStrings(elided): - """Collapses strings and chars on a line to simple "" or '' blocks. - - We nix strings first so we're not fooled by text like '"http://"' + """Holds 4 copies of all lines with different preprocessing applied to them. + + 1) elided member contains lines without strings and comments. + 2) lines member contains lines without comments. + 3) raw_lines member contains all the lines without processing. + 4) lines_without_raw_strings member is same as raw_lines, but with C++11 raw + strings removed. + All these members are of , and of the same length. + """ + + def __init__(self, lines): + self.elided = [] + self.lines = [] + self.raw_lines = lines + self.num_lines = len(lines) + self.lines_without_raw_strings = CleanseRawStrings(lines) + for linenum in range(len(self.lines_without_raw_strings)): + self.lines.append(CleanseComments(self.lines_without_raw_strings[linenum])) + elided = self._CollapseStrings(self.lines_without_raw_strings[linenum]) + self.elided.append(CleanseComments(elided)) + + def NumLines(self): + """Returns the number of lines represented.""" + return self.num_lines + + @staticmethod + def _CollapseStrings(elided): + """Collapses strings and chars on a line to simple "" or '' blocks. + + We nix strings first so we're not fooled by text like '"http://"' + + Args: + elided: The line being processed. + + Returns: + The line with collapsed strings. + """ + if _RE_PATTERN_INCLUDE.match(elided): + return elided + + # Remove escaped characters first to make quote/single quote collapsing + # basic. Things that look like escaped characters shouldn't occur + # outside of strings and chars. + elided = _RE_PATTERN_CLEANSE_LINE_ESCAPES.sub("", elided) + + # Replace quoted strings and digit separators. Both single quotes + # and double quotes are processed in the same loop, otherwise + # nested quotes wouldn't work. + collapsed = "" + while True: + # Find the first quote character + match = Match(r'^([^\'"]*)([\'"])(.*)$', elided) + if not match: + collapsed += elided + break + head, quote, tail = match.groups() + + if quote == '"': + # Collapse double quoted strings + second_quote = tail.find('"') + if second_quote >= 0: + collapsed += head + '""' + elided = tail[second_quote + 1 :] + else: + # Unmatched double quote, don't bother processing the rest + # of the line since this is probably a multiline string. + collapsed += elided + break + else: + # Found single quote, check nearby text to eliminate digit separators. + # + # There is no special handling for floating point here, because + # the integer/fractional/exponent parts would all be parsed + # correctly as long as there are digits on both sides of the + # separator. So we are fine as long as we don't see something + # like "0.'3" (gcc 4.9.0 will not allow this literal). + if Search(r"\b(?:0[bBxX]?|[1-9])[0-9a-fA-F]*$", head): + match_literal = Match(r"^((?:\'?[0-9a-zA-Z_])*)(.*)$", "'" + tail) + collapsed += head + match_literal.group(1).replace("'", "") + elided = match_literal.group(2) + else: + second_quote = tail.find("'") + if second_quote >= 0: + collapsed += head + "''" + elided = tail[second_quote + 1 :] + else: + # Unmatched single quote + collapsed += elided + break + + return collapsed + + +def FindEndOfExpressionInLine(line, startpos, stack): + """Find the position just after the end of current parenthesized expression. Args: - elided: The line being processed. + line: a CleansedLines line. + startpos: start searching at this position. + stack: nesting stack at startpos. Returns: - The line with collapsed strings. + On finding matching end: (index just after matching end, None) + On finding an unclosed expression: (-1, None) + Otherwise: (-1, new stack at end of this line) """ - if _RE_PATTERN_INCLUDE.match(elided): - return elided - - # Remove escaped characters first to make quote/single quote collapsing - # basic. Things that look like escaped characters shouldn't occur - # outside of strings and chars. - elided = _RE_PATTERN_CLEANSE_LINE_ESCAPES.sub('', elided) - - # Replace quoted strings and digit separators. Both single quotes - # and double quotes are processed in the same loop, otherwise - # nested quotes wouldn't work. - collapsed = '' - while True: - # Find the first quote character - match = Match(r'^([^\'"]*)([\'"])(.*)$', elided) - if not match: - collapsed += elided - break - head, quote, tail = match.groups() - - if quote == '"': - # Collapse double quoted strings - second_quote = tail.find('"') - if second_quote >= 0: - collapsed += head + '""' - elided = tail[second_quote + 1:] - else: - # Unmatched double quote, don't bother processing the rest - # of the line since this is probably a multiline string. - collapsed += elided - break - else: - # Found single quote, check nearby text to eliminate digit separators. - # - # There is no special handling for floating point here, because - # the integer/fractional/exponent parts would all be parsed - # correctly as long as there are digits on both sides of the - # separator. So we are fine as long as we don't see something - # like "0.'3" (gcc 4.9.0 will not allow this literal). - if Search(r'\b(?:0[bBxX]?|[1-9])[0-9a-fA-F]*$', head): - match_literal = Match(r'^((?:\'?[0-9a-zA-Z_])*)(.*)$', "'" + tail) - collapsed += head + match_literal.group(1).replace("'", '') - elided = match_literal.group(2) - else: - second_quote = tail.find('\'') - if second_quote >= 0: - collapsed += head + "''" - elided = tail[second_quote + 1:] - else: - # Unmatched single quote - collapsed += elided - break + for i in xrange(startpos, len(line)): + char = line[i] + if char in "([{": + # Found start of parenthesized expression, push to expression stack + stack.append(char) + elif char == "<": + # Found potential start of template argument list + if i > 0 and line[i - 1] == "<": + # Left shift operator + if stack and stack[-1] == "<": + stack.pop() + if not stack: + return (-1, None) + elif i > 0 and Search(r"\boperator\s*$", line[0:i]): + # operator<, don't add to stack + continue + else: + # Tentative start of template argument list + stack.append("<") + elif char in ")]}": + # Found end of parenthesized expression. + # + # If we are currently expecting a matching '>', the pending '<' + # must have been an operator. Remove them from expression stack. + while stack and stack[-1] == "<": + stack.pop() + if not stack: + return (-1, None) + if ( + (stack[-1] == "(" and char == ")") + or (stack[-1] == "[" and char == "]") + or (stack[-1] == "{" and char == "}") + ): + stack.pop() + if not stack: + return (i + 1, None) + else: + # Mismatched parentheses + return (-1, None) + elif char == ">": + # Found potential end of template argument list. + + # Ignore "->" and operator functions + if i > 0 and ( + line[i - 1] == "-" or Search(r"\boperator\s*$", line[0 : i - 1]) + ): + continue + + # Pop the stack if there is a matching '<'. Otherwise, ignore + # this '>' since it must be an operator. + if stack: + if stack[-1] == "<": + stack.pop() + if not stack: + return (i + 1, None) + elif char == ";": + # Found something that look like end of statements. If we are currently + # expecting a '>', the matching '<' must have been an operator, since + # template argument list should not contain statements. + while stack and stack[-1] == "<": + stack.pop() + if not stack: + return (-1, None) + + # Did not find end of expression or unbalanced parentheses on this line + return (-1, stack) - return collapsed +def CloseExpression(clean_lines, linenum, pos): + """If input points to ( or { or [ or <, finds the position that closes it. -def FindEndOfExpressionInLine(line, startpos, stack): - """Find the position just after the end of current parenthesized expression. - - Args: - line: a CleansedLines line. - startpos: start searching at this position. - stack: nesting stack at startpos. - - Returns: - On finding matching end: (index just after matching end, None) - On finding an unclosed expression: (-1, None) - Otherwise: (-1, new stack at end of this line) - """ - for i in xrange(startpos, len(line)): - char = line[i] - if char in '([{': - # Found start of parenthesized expression, push to expression stack - stack.append(char) - elif char == '<': - # Found potential start of template argument list - if i > 0 and line[i - 1] == '<': - # Left shift operator - if stack and stack[-1] == '<': - stack.pop() - if not stack: - return (-1, None) - elif i > 0 and Search(r'\boperator\s*$', line[0:i]): - # operator<, don't add to stack - continue - else: - # Tentative start of template argument list - stack.append('<') - elif char in ')]}': - # Found end of parenthesized expression. - # - # If we are currently expecting a matching '>', the pending '<' - # must have been an operator. Remove them from expression stack. - while stack and stack[-1] == '<': - stack.pop() - if not stack: - return (-1, None) - if ((stack[-1] == '(' and char == ')') or - (stack[-1] == '[' and char == ']') or - (stack[-1] == '{' and char == '}')): - stack.pop() - if not stack: - return (i + 1, None) - else: - # Mismatched parentheses - return (-1, None) - elif char == '>': - # Found potential end of template argument list. - - # Ignore "->" and operator functions - if (i > 0 and - (line[i - 1] == '-' or Search(r'\boperator\s*$', line[0:i - 1]))): - continue - - # Pop the stack if there is a matching '<'. Otherwise, ignore - # this '>' since it must be an operator. - if stack: - if stack[-1] == '<': - stack.pop() - if not stack: - return (i + 1, None) - elif char == ';': - # Found something that look like end of statements. If we are currently - # expecting a '>', the matching '<' must have been an operator, since - # template argument list should not contain statements. - while stack and stack[-1] == '<': - stack.pop() - if not stack: - return (-1, None) - - # Did not find end of expression or unbalanced parentheses on this line - return (-1, stack) + If lines[linenum][pos] points to a '(' or '{' or '[' or '<', finds the + linenum/pos that correspond to the closing of the expression. + TODO(unknown): cpplint spends a fair bit of time matching parentheses. + Ideally we would want to index all opening and closing parentheses once + and have CloseExpression be just a simple lookup, but due to preprocessor + tricks, this is not so easy. -def CloseExpression(clean_lines, linenum, pos): - """If input points to ( or { or [ or <, finds the position that closes it. - - If lines[linenum][pos] points to a '(' or '{' or '[' or '<', finds the - linenum/pos that correspond to the closing of the expression. - - TODO(unknown): cpplint spends a fair bit of time matching parentheses. - Ideally we would want to index all opening and closing parentheses once - and have CloseExpression be just a simple lookup, but due to preprocessor - tricks, this is not so easy. - - Args: - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - pos: A position on the line. - - Returns: - A tuple (line, linenum, pos) pointer *past* the closing brace, or - (line, len(lines), -1) if we never find a close. Note we ignore - strings and comments when matching; and the line we return is the - 'cleansed' line at linenum. - """ - - line = clean_lines.elided[linenum] - if (line[pos] not in '({[<') or Match(r'<[<=]', line[pos:]): - return (line, clean_lines.NumLines(), -1) + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + pos: A position on the line. - # Check first line - (end_pos, stack) = FindEndOfExpressionInLine(line, pos, []) - if end_pos > -1: - return (line, linenum, end_pos) + Returns: + A tuple (line, linenum, pos) pointer *past* the closing brace, or + (line, len(lines), -1) if we never find a close. Note we ignore + strings and comments when matching; and the line we return is the + 'cleansed' line at linenum. + """ - # Continue scanning forward - while stack and linenum < clean_lines.NumLines() - 1: - linenum += 1 line = clean_lines.elided[linenum] - (end_pos, stack) = FindEndOfExpressionInLine(line, 0, stack) + if (line[pos] not in "({[<") or Match(r"<[<=]", line[pos:]): + return (line, clean_lines.NumLines(), -1) + + # Check first line + (end_pos, stack) = FindEndOfExpressionInLine(line, pos, []) if end_pos > -1: - return (line, linenum, end_pos) + return (line, linenum, end_pos) + + # Continue scanning forward + while stack and linenum < clean_lines.NumLines() - 1: + linenum += 1 + line = clean_lines.elided[linenum] + (end_pos, stack) = FindEndOfExpressionInLine(line, 0, stack) + if end_pos > -1: + return (line, linenum, end_pos) - # Did not find end of expression before end of file, give up - return (line, clean_lines.NumLines(), -1) + # Did not find end of expression before end of file, give up + return (line, clean_lines.NumLines(), -1) def FindStartOfExpressionInLine(line, endpos, stack): - """Find position at the matching start of current expression. - - This is almost the reverse of FindEndOfExpressionInLine, but note - that the input position and returned position differs by 1. - - Args: - line: a CleansedLines line. - endpos: start searching at this position. - stack: nesting stack at endpos. - - Returns: - On finding matching start: (index at matching start, None) - On finding an unclosed expression: (-1, None) - Otherwise: (-1, new stack at beginning of this line) - """ - i = endpos - while i >= 0: - char = line[i] - if char in ')]}': - # Found end of expression, push to expression stack - stack.append(char) - elif char == '>': - # Found potential end of template argument list. - # - # Ignore it if it's a "->" or ">=" or "operator>" - if (i > 0 and - (line[i - 1] == '-' or - Match(r'\s>=\s', line[i - 1:]) or - Search(r'\boperator\s*$', line[0:i]))): - i -= 1 - else: - stack.append('>') - elif char == '<': - # Found potential start of template argument list - if i > 0 and line[i - 1] == '<': - # Left shift operator + """Find position at the matching start of current expression. + + This is almost the reverse of FindEndOfExpressionInLine, but note + that the input position and returned position differs by 1. + + Args: + line: a CleansedLines line. + endpos: start searching at this position. + stack: nesting stack at endpos. + + Returns: + On finding matching start: (index at matching start, None) + On finding an unclosed expression: (-1, None) + Otherwise: (-1, new stack at beginning of this line) + """ + i = endpos + while i >= 0: + char = line[i] + if char in ")]}": + # Found end of expression, push to expression stack + stack.append(char) + elif char == ">": + # Found potential end of template argument list. + # + # Ignore it if it's a "->" or ">=" or "operator>" + if i > 0 and ( + line[i - 1] == "-" + or Match(r"\s>=\s", line[i - 1 :]) + or Search(r"\boperator\s*$", line[0:i]) + ): + i -= 1 + else: + stack.append(">") + elif char == "<": + # Found potential start of template argument list + if i > 0 and line[i - 1] == "<": + # Left shift operator + i -= 1 + else: + # If there is a matching '>', we can pop the expression stack. + # Otherwise, ignore this '<' since it must be an operator. + if stack and stack[-1] == ">": + stack.pop() + if not stack: + return (i, None) + elif char in "([{": + # Found start of expression. + # + # If there are any unmatched '>' on the stack, they must be + # operators. Remove those. + while stack and stack[-1] == ">": + stack.pop() + if not stack: + return (-1, None) + if ( + (char == "(" and stack[-1] == ")") + or (char == "[" and stack[-1] == "]") + or (char == "{" and stack[-1] == "}") + ): + stack.pop() + if not stack: + return (i, None) + else: + # Mismatched parentheses + return (-1, None) + elif char == ";": + # Found something that look like end of statements. If we are currently + # expecting a '<', the matching '>' must have been an operator, since + # template argument list should not contain statements. + while stack and stack[-1] == ">": + stack.pop() + if not stack: + return (-1, None) + i -= 1 - else: - # If there is a matching '>', we can pop the expression stack. - # Otherwise, ignore this '<' since it must be an operator. - if stack and stack[-1] == '>': - stack.pop() - if not stack: - return (i, None) - elif char in '([{': - # Found start of expression. - # - # If there are any unmatched '>' on the stack, they must be - # operators. Remove those. - while stack and stack[-1] == '>': - stack.pop() - if not stack: - return (-1, None) - if ((char == '(' and stack[-1] == ')') or - (char == '[' and stack[-1] == ']') or - (char == '{' and stack[-1] == '}')): - stack.pop() - if not stack: - return (i, None) - else: - # Mismatched parentheses - return (-1, None) - elif char == ';': - # Found something that look like end of statements. If we are currently - # expecting a '<', the matching '>' must have been an operator, since - # template argument list should not contain statements. - while stack and stack[-1] == '>': - stack.pop() - if not stack: - return (-1, None) - - i -= 1 - - return (-1, stack) + + return (-1, stack) def ReverseCloseExpression(clean_lines, linenum, pos): - """If input points to ) or } or ] or >, finds the position that opens it. - - If lines[linenum][pos] points to a ')' or '}' or ']' or '>', finds the - linenum/pos that correspond to the opening of the expression. - - Args: - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - pos: A position on the line. - - Returns: - A tuple (line, linenum, pos) pointer *at* the opening brace, or - (line, 0, -1) if we never find the matching opening brace. Note - we ignore strings and comments when matching; and the line we - return is the 'cleansed' line at linenum. - """ - line = clean_lines.elided[linenum] - if line[pos] not in ')}]>': - return (line, 0, -1) + """If input points to ) or } or ] or >, finds the position that opens it. - # Check last line - (start_pos, stack) = FindStartOfExpressionInLine(line, pos, []) - if start_pos > -1: - return (line, linenum, start_pos) + If lines[linenum][pos] points to a ')' or '}' or ']' or '>', finds the + linenum/pos that correspond to the opening of the expression. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + pos: A position on the line. - # Continue scanning backward - while stack and linenum > 0: - linenum -= 1 + Returns: + A tuple (line, linenum, pos) pointer *at* the opening brace, or + (line, 0, -1) if we never find the matching opening brace. Note + we ignore strings and comments when matching; and the line we + return is the 'cleansed' line at linenum. + """ line = clean_lines.elided[linenum] - (start_pos, stack) = FindStartOfExpressionInLine(line, len(line) - 1, stack) + if line[pos] not in ")}]>": + return (line, 0, -1) + + # Check last line + (start_pos, stack) = FindStartOfExpressionInLine(line, pos, []) if start_pos > -1: - return (line, linenum, start_pos) + return (line, linenum, start_pos) + + # Continue scanning backward + while stack and linenum > 0: + linenum -= 1 + line = clean_lines.elided[linenum] + (start_pos, stack) = FindStartOfExpressionInLine(line, len(line) - 1, stack) + if start_pos > -1: + return (line, linenum, start_pos) - # Did not find start of expression before beginning of file, give up - return (line, 0, -1) + # Did not find start of expression before beginning of file, give up + return (line, 0, -1) def CheckForCopyright(filename, lines, error): - """Logs an error if no Copyright message appears at the top of the file.""" + """Logs an error if no Copyright message appears at the top of the file.""" - # We'll say it should occur by line 10. Don't forget there's a - # dummy line at the front. - for line in xrange(1, min(len(lines), 11)): - if re.search(r'Copyright', lines[line], re.I): break - else: # means no copyright line was found - error(filename, 0, 'legal/copyright', 5, - 'No copyright message found. ' - 'You should have a line: "Copyright [year] "') + # We'll say it should occur by line 10. Don't forget there's a + # dummy line at the front. + for line in xrange(1, min(len(lines), 11)): + if re.search(r"Copyright", lines[line], re.I): + break + else: # means no copyright line was found + error( + filename, + 0, + "legal/copyright", + 5, + "No copyright message found. " + 'You should have a line: "Copyright [year] "', + ) def GetIndentLevel(line): - """Return the number of leading spaces in line. + """Return the number of leading spaces in line. - Args: - line: A string to check. + Args: + line: A string to check. + + Returns: + An integer count of leading spaces, possibly zero. + """ + indent = Match(r"^( *)\S", line) + if indent: + return len(indent.group(1)) + else: + return 0 - Returns: - An integer count of leading spaces, possibly zero. - """ - indent = Match(r'^( *)\S', line) - if indent: - return len(indent.group(1)) - else: - return 0 def PathSplitToList(path): - """Returns the path split into a list by the separator. - - Args: - path: An absolute or relative path (e.g. '/a/b/c/' or '../a') - - Returns: - A list of path components (e.g. ['a', 'b', 'c]). - """ - lst = [] - while True: - (head, tail) = os.path.split(path) - if head == path: # absolute paths end - lst.append(head) - break - if tail == path: # relative paths end - lst.append(tail) - break - - path = head - lst.append(tail) - - lst.reverse() - return lst + """Returns the path split into a list by the separator. -def GetHeaderGuardCPPVariable(filename): - """Returns the CPP variable that should be used as a header guard. + Args: + path: An absolute or relative path (e.g. '/a/b/c/' or '../a') - Args: - filename: The name of a C++ header file. + Returns: + A list of path components (e.g. ['a', 'b', 'c]). + """ + lst = [] + while True: + (head, tail) = os.path.split(path) + if head == path: # absolute paths end + lst.append(head) + break + if tail == path: # relative paths end + lst.append(tail) + break - Returns: - The CPP variable that should be used as a header guard in the - named file. + path = head + lst.append(tail) - """ + lst.reverse() + return lst - # Restores original filename in case that cpplint is invoked from Emacs's - # flymake. - filename = re.sub(r'_flymake\.h$', '.h', filename) - filename = re.sub(r'/\.flymake/([^/]*)$', r'/\1', filename) - # Replace 'c++' with 'cpp'. - filename = filename.replace('C++', 'cpp').replace('c++', 'cpp') - fileinfo = FileInfo(filename) - file_path_from_root = fileinfo.RepositoryName() +def GetHeaderGuardCPPVariable(filename): + """Returns the CPP variable that should be used as a header guard. - def FixupPathFromRoot(): - if _root_debug: - sys.stderr.write("\n_root fixup, _root = '%s', repository name = '%s'\n" - %(_root, fileinfo.RepositoryName())) + Args: + filename: The name of a C++ header file. - # Process the file path with the --root flag if it was set. - if not _root: - if _root_debug: - sys.stderr.write("_root unspecified\n") - return file_path_from_root + Returns: + The CPP variable that should be used as a header guard in the + named file. - def StripListPrefix(lst, prefix): - # f(['x', 'y'], ['w, z']) -> None (not a valid prefix) - if lst[:len(prefix)] != prefix: - return None - # f(['a, 'b', 'c', 'd'], ['a', 'b']) -> ['c', 'd'] - return lst[(len(prefix)):] + """ + + # Restores original filename in case that cpplint is invoked from Emacs's + # flymake. + filename = re.sub(r"_flymake\.h$", ".h", filename) + filename = re.sub(r"/\.flymake/([^/]*)$", r"/\1", filename) + # Replace 'c++' with 'cpp'. + filename = filename.replace("C++", "cpp").replace("c++", "cpp") - # root behavior: - # --root=subdir , lstrips subdir from the header guard - maybe_path = StripListPrefix(PathSplitToList(file_path_from_root), - PathSplitToList(_root)) + fileinfo = FileInfo(filename) + file_path_from_root = fileinfo.RepositoryName() - if _root_debug: - sys.stderr.write(("_root lstrip (maybe_path=%s, file_path_from_root=%s," + - " _root=%s)\n") %(maybe_path, file_path_from_root, _root)) + def FixupPathFromRoot(): + if _root_debug: + sys.stderr.write( + "\n_root fixup, _root = '%s', repository name = '%s'\n" + % (_root, fileinfo.RepositoryName()) + ) + + # Process the file path with the --root flag if it was set. + if not _root: + if _root_debug: + sys.stderr.write("_root unspecified\n") + return file_path_from_root + + def StripListPrefix(lst, prefix): + # f(['x', 'y'], ['w, z']) -> None (not a valid prefix) + if lst[: len(prefix)] != prefix: + return None + # f(['a, 'b', 'c', 'd'], ['a', 'b']) -> ['c', 'd'] + return lst[(len(prefix)) :] + + # root behavior: + # --root=subdir , lstrips subdir from the header guard + maybe_path = StripListPrefix( + PathSplitToList(file_path_from_root), PathSplitToList(_root) + ) + + if _root_debug: + sys.stderr.write( + ( + "_root lstrip (maybe_path=%s, file_path_from_root=%s," + + " _root=%s)\n" + ) + % (maybe_path, file_path_from_root, _root) + ) - if maybe_path: - return os.path.join(*maybe_path) + if maybe_path: + return os.path.join(*maybe_path) - # --root=.. , will prepend the outer directory to the header guard - full_path = fileinfo.FullName() - root_abspath = os.path.abspath(_root) + # --root=.. , will prepend the outer directory to the header guard + full_path = fileinfo.FullName() + root_abspath = os.path.abspath(_root) - maybe_path = StripListPrefix(PathSplitToList(full_path), - PathSplitToList(root_abspath)) + maybe_path = StripListPrefix( + PathSplitToList(full_path), PathSplitToList(root_abspath) + ) - if _root_debug: - sys.stderr.write(("_root prepend (maybe_path=%s, full_path=%s, " + - "root_abspath=%s)\n") %(maybe_path, full_path, root_abspath)) + if _root_debug: + sys.stderr.write( + ("_root prepend (maybe_path=%s, full_path=%s, " + "root_abspath=%s)\n") + % (maybe_path, full_path, root_abspath) + ) - if maybe_path: - return os.path.join(*maybe_path) + if maybe_path: + return os.path.join(*maybe_path) - if _root_debug: - sys.stderr.write("_root ignore, returning %s\n" %(file_path_from_root)) + if _root_debug: + sys.stderr.write("_root ignore, returning %s\n" % (file_path_from_root)) - # --root=FAKE_DIR is ignored - return file_path_from_root + # --root=FAKE_DIR is ignored + return file_path_from_root - file_path_from_root = FixupPathFromRoot() - return re.sub(r'[^a-zA-Z0-9]', '_', file_path_from_root).upper() + '_' + file_path_from_root = FixupPathFromRoot() + return re.sub(r"[^a-zA-Z0-9]", "_", file_path_from_root).upper() + "_" def CheckForHeaderGuard(filename, clean_lines, error): - """Checks that the file contains a header guard. - - Logs an error if no #ifndef header guard is present. For other - headers, checks that the full pathname is used. - - Args: - filename: The name of the C++ header file. - clean_lines: A CleansedLines instance containing the file. - error: The function to call with any errors found. - """ - - # Don't check for header guards if there are error suppression - # comments somewhere in this file. - # - # Because this is silencing a warning for a nonexistent line, we - # only support the very specific NOLINT(build/header_guard) syntax, - # and not the general NOLINT or NOLINT(*) syntax. - raw_lines = clean_lines.lines_without_raw_strings - for i in raw_lines: - if Search(r'//\s*NOLINT\(build/header_guard\)', i): - return - - cppvar = GetHeaderGuardCPPVariable(filename) - - ifndef = '' - ifndef_linenum = 0 - define = '' - endif = '' - endif_linenum = 0 - for linenum, line in enumerate(raw_lines): - linesplit = line.split() - if len(linesplit) >= 2: - # find the first occurrence of #ifndef and #define, save arg - if not ifndef and linesplit[0] == '#ifndef': - # set ifndef to the header guard presented on the #ifndef line. - ifndef = linesplit[1] - ifndef_linenum = linenum - if not define and linesplit[0] == '#define': - define = linesplit[1] - # find the last occurrence of #endif, save entire line - if line.startswith('#endif'): - endif = line - endif_linenum = linenum - - if not ifndef or not define or ifndef != define: - error(filename, 0, 'build/header_guard', 5, - 'No #ifndef header guard found, suggested CPP variable is: %s' % - cppvar) - return - - # The guard should be PATH_FILE_H_, but we also allow PATH_FILE_H__ - # for backward compatibility. - if ifndef != cppvar: - error_level = 0 - if ifndef != cppvar + '_': - error_level = 5 - - ParseNolintSuppressions(filename, raw_lines[ifndef_linenum], ifndef_linenum, - error) - error(filename, ifndef_linenum, 'build/header_guard', error_level, - '#ifndef header guard has wrong style, please use: %s' % cppvar) - - # Check for "//" comments on endif line. - ParseNolintSuppressions(filename, raw_lines[endif_linenum], endif_linenum, - error) - match = Match(r'#endif\s*//\s*' + cppvar + r'(_)?\b', endif) - if match: - if match.group(1) == '_': - # Issue low severity warning for deprecated double trailing underscore - error(filename, endif_linenum, 'build/header_guard', 0, - '#endif line should be "#endif // %s"' % cppvar) - return - - # Didn't find the corresponding "//" comment. If this file does not - # contain any "//" comments at all, it could be that the compiler - # only wants "/**/" comments, look for those instead. - no_single_line_comments = True - for i in xrange(1, len(raw_lines) - 1): - line = raw_lines[i] - if Match(r'^(?:(?:\'(?:\.|[^\'])*\')|(?:"(?:\.|[^"])*")|[^\'"])*//', line): - no_single_line_comments = False - break - - if no_single_line_comments: - match = Match(r'#endif\s*/\*\s*' + cppvar + r'(_)?\s*\*/', endif) + """Checks that the file contains a header guard. + + Logs an error if no #ifndef header guard is present. For other + headers, checks that the full pathname is used. + + Args: + filename: The name of the C++ header file. + clean_lines: A CleansedLines instance containing the file. + error: The function to call with any errors found. + """ + + # Don't check for header guards if there are error suppression + # comments somewhere in this file. + # + # Because this is silencing a warning for a nonexistent line, we + # only support the very specific NOLINT(build/header_guard) syntax, + # and not the general NOLINT or NOLINT(*) syntax. + raw_lines = clean_lines.lines_without_raw_strings + for i in raw_lines: + if Search(r"//\s*NOLINT\(build/header_guard\)", i): + return + + cppvar = GetHeaderGuardCPPVariable(filename) + + ifndef = "" + ifndef_linenum = 0 + define = "" + endif = "" + endif_linenum = 0 + for linenum, line in enumerate(raw_lines): + linesplit = line.split() + if len(linesplit) >= 2: + # find the first occurrence of #ifndef and #define, save arg + if not ifndef and linesplit[0] == "#ifndef": + # set ifndef to the header guard presented on the #ifndef line. + ifndef = linesplit[1] + ifndef_linenum = linenum + if not define and linesplit[0] == "#define": + define = linesplit[1] + # find the last occurrence of #endif, save entire line + if line.startswith("#endif"): + endif = line + endif_linenum = linenum + + if not ifndef or not define or ifndef != define: + error( + filename, + 0, + "build/header_guard", + 5, + "No #ifndef header guard found, suggested CPP variable is: %s" % cppvar, + ) + return + + # The guard should be PATH_FILE_H_, but we also allow PATH_FILE_H__ + # for backward compatibility. + if ifndef != cppvar: + error_level = 0 + if ifndef != cppvar + "_": + error_level = 5 + + ParseNolintSuppressions( + filename, raw_lines[ifndef_linenum], ifndef_linenum, error + ) + error( + filename, + ifndef_linenum, + "build/header_guard", + error_level, + "#ifndef header guard has wrong style, please use: %s" % cppvar, + ) + + # Check for "//" comments on endif line. + ParseNolintSuppressions(filename, raw_lines[endif_linenum], endif_linenum, error) + match = Match(r"#endif\s*//\s*" + cppvar + r"(_)?\b", endif) if match: - if match.group(1) == '_': - # Low severity warning for double trailing underscore - error(filename, endif_linenum, 'build/header_guard', 0, - '#endif line should be "#endif /* %s */"' % cppvar) - return + if match.group(1) == "_": + # Issue low severity warning for deprecated double trailing underscore + error( + filename, + endif_linenum, + "build/header_guard", + 0, + '#endif line should be "#endif // %s"' % cppvar, + ) + return + + # Didn't find the corresponding "//" comment. If this file does not + # contain any "//" comments at all, it could be that the compiler + # only wants "/**/" comments, look for those instead. + no_single_line_comments = True + for i in xrange(1, len(raw_lines) - 1): + line = raw_lines[i] + if Match(r'^(?:(?:\'(?:\.|[^\'])*\')|(?:"(?:\.|[^"])*")|[^\'"])*//', line): + no_single_line_comments = False + break - # Didn't find anything - error(filename, endif_linenum, 'build/header_guard', 5, - '#endif line should be "#endif // %s"' % cppvar) + if no_single_line_comments: + match = Match(r"#endif\s*/\*\s*" + cppvar + r"(_)?\s*\*/", endif) + if match: + if match.group(1) == "_": + # Low severity warning for double trailing underscore + error( + filename, + endif_linenum, + "build/header_guard", + 0, + '#endif line should be "#endif /* %s */"' % cppvar, + ) + return + + # Didn't find anything + error( + filename, + endif_linenum, + "build/header_guard", + 5, + '#endif line should be "#endif // %s"' % cppvar, + ) def CheckHeaderFileIncluded(filename, include_state, error): - """Logs an error if a .cc file does not include its header.""" - - # Do not check test files - fileinfo = FileInfo(filename) - if Search(_TEST_FILE_SUFFIX, fileinfo.BaseName()): - return - - headerfile = filename[0:len(filename) - len(fileinfo.Extension())] + '.h' - if not os.path.exists(headerfile): - return - headername = FileInfo(headerfile).RepositoryName() - first_include = 0 - for section_list in include_state.include_list: - for f in section_list: - if headername in f[0] or f[0] in headername: + """Logs an error if a .cc file does not include its header.""" + + # Do not check test files + fileinfo = FileInfo(filename) + if Search(_TEST_FILE_SUFFIX, fileinfo.BaseName()): return - if not first_include: - first_include = f[1] - error(filename, first_include, 'build/include', 5, - '%s should include its header file %s' % (fileinfo.RepositoryName(), - headername)) + headerfile = filename[0 : len(filename) - len(fileinfo.Extension())] + ".h" + if not os.path.exists(headerfile): + return + headername = FileInfo(headerfile).RepositoryName() + first_include = 0 + for section_list in include_state.include_list: + for f in section_list: + if headername in f[0] or f[0] in headername: + return + if not first_include: + first_include = f[1] + + error( + filename, + first_include, + "build/include", + 5, + "%s should include its header file %s" + % (fileinfo.RepositoryName(), headername), + ) def CheckForBadCharacters(filename, lines, error): - """Logs an error for each line containing bad characters. + """Logs an error for each line containing bad characters. - Two kinds of bad characters: + Two kinds of bad characters: - 1. Unicode replacement characters: These indicate that either the file - contained invalid UTF-8 (likely) or Unicode replacement characters (which - it shouldn't). Note that it's possible for this to throw off line - numbering if the invalid UTF-8 occurred adjacent to a newline. + 1. Unicode replacement characters: These indicate that either the file + contained invalid UTF-8 (likely) or Unicode replacement characters (which + it shouldn't). Note that it's possible for this to throw off line + numbering if the invalid UTF-8 occurred adjacent to a newline. - 2. NUL bytes. These are problematic for some tools. + 2. NUL bytes. These are problematic for some tools. - Args: - filename: The name of the current file. - lines: An array of strings, each representing a line of the file. - error: The function to call with any errors found. - """ - for linenum, line in enumerate(lines): - if u'\ufffd' in line: - error(filename, linenum, 'readability/utf8', 5, - 'Line contains invalid UTF-8 (or Unicode replacement character).') - if '\0' in line: - error(filename, linenum, 'readability/nul', 5, 'Line contains NUL byte.') + Args: + filename: The name of the current file. + lines: An array of strings, each representing a line of the file. + error: The function to call with any errors found. + """ + for linenum, line in enumerate(lines): + if "\ufffd" in line: + error( + filename, + linenum, + "readability/utf8", + 5, + "Line contains invalid UTF-8 (or Unicode replacement character).", + ) + if "\0" in line: + error(filename, linenum, "readability/nul", 5, "Line contains NUL byte.") def CheckForNewlineAtEOF(filename, lines, error): - """Logs an error if there is no newline char at the end of the file. + """Logs an error if there is no newline char at the end of the file. - Args: - filename: The name of the current file. - lines: An array of strings, each representing a line of the file. - error: The function to call with any errors found. - """ + Args: + filename: The name of the current file. + lines: An array of strings, each representing a line of the file. + error: The function to call with any errors found. + """ - # The array lines() was created by adding two newlines to the - # original file (go figure), then splitting on \n. - # To verify that the file ends in \n, we just have to make sure the - # last-but-two element of lines() exists and is empty. - if len(lines) < 3 or lines[-2]: - error(filename, len(lines) - 2, 'whitespace/ending_newline', 5, - 'Could not find a newline character at the end of the file.') + # The array lines() was created by adding two newlines to the + # original file (go figure), then splitting on \n. + # To verify that the file ends in \n, we just have to make sure the + # last-but-two element of lines() exists and is empty. + if len(lines) < 3 or lines[-2]: + error( + filename, + len(lines) - 2, + "whitespace/ending_newline", + 5, + "Could not find a newline character at the end of the file.", + ) def CheckForMultilineCommentsAndStrings(filename, clean_lines, linenum, error): - """Logs an error if we see /* ... */ or "..." that extend past one line. - - /* ... */ comments are legit inside macros, for one line. - Otherwise, we prefer // comments, so it's ok to warn about the - other. Likewise, it's ok for strings to extend across multiple - lines, as long as a line continuation character (backslash) - terminates each line. Although not currently prohibited by the C++ - style guide, it's ugly and unnecessary. We don't do well with either - in this lint program, so we warn about both. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - # Remove all \\ (escaped backslashes) from the line. They are OK, and the - # second (escaped) slash may trigger later \" detection erroneously. - line = line.replace('\\\\', '') - - if line.count('/*') > line.count('*/'): - error(filename, linenum, 'readability/multiline_comment', 5, - 'Complex multi-line /*...*/-style comment found. ' - 'Lint may give bogus warnings. ' - 'Consider replacing these with //-style comments, ' - 'with #if 0...#endif, ' - 'or with more clearly structured multi-line comments.') - - if (line.count('"') - line.count('\\"')) % 2: - error(filename, linenum, 'readability/multiline_string', 5, - 'Multi-line string ("...") found. This lint script doesn\'t ' - 'do well with such strings, and may give bogus warnings. ' - 'Use C++11 raw strings or concatenation instead.') + """Logs an error if we see /* ... */ or "..." that extend past one line. + + /* ... */ comments are legit inside macros, for one line. + Otherwise, we prefer // comments, so it's ok to warn about the + other. Likewise, it's ok for strings to extend across multiple + lines, as long as a line continuation character (backslash) + terminates each line. Although not currently prohibited by the C++ + style guide, it's ugly and unnecessary. We don't do well with either + in this lint program, so we warn about both. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Remove all \\ (escaped backslashes) from the line. They are OK, and the + # second (escaped) slash may trigger later \" detection erroneously. + line = line.replace("\\\\", "") + + if line.count("/*") > line.count("*/"): + error( + filename, + linenum, + "readability/multiline_comment", + 5, + "Complex multi-line /*...*/-style comment found. " + "Lint may give bogus warnings. " + "Consider replacing these with //-style comments, " + "with #if 0...#endif, " + "or with more clearly structured multi-line comments.", + ) + + if (line.count('"') - line.count('\\"')) % 2: + error( + filename, + linenum, + "readability/multiline_string", + 5, + 'Multi-line string ("...") found. This lint script doesn\'t ' + "do well with such strings, and may give bogus warnings. " + "Use C++11 raw strings or concatenation instead.", + ) # (non-threadsafe name, thread-safe alternative, validation pattern) @@ -2098,126 +2241,61 @@ def CheckForMultilineCommentsAndStrings(filename, clean_lines, linenum, error): # in some expression context on the same line by matching on some # operator before the function name. This eliminates constructors and # member function calls. -_UNSAFE_FUNC_PREFIX = r'(?:[-+*/=%^&|(<]\s*|>\s+)' +_UNSAFE_FUNC_PREFIX = r"(?:[-+*/=%^&|(<]\s*|>\s+)" _THREADING_LIST = ( - ('asctime(', 'asctime_r(', _UNSAFE_FUNC_PREFIX + r'asctime\([^)]+\)'), - ('ctime(', 'ctime_r(', _UNSAFE_FUNC_PREFIX + r'ctime\([^)]+\)'), - ('getgrgid(', 'getgrgid_r(', _UNSAFE_FUNC_PREFIX + r'getgrgid\([^)]+\)'), - ('getgrnam(', 'getgrnam_r(', _UNSAFE_FUNC_PREFIX + r'getgrnam\([^)]+\)'), - ('getlogin(', 'getlogin_r(', _UNSAFE_FUNC_PREFIX + r'getlogin\(\)'), - ('getpwnam(', 'getpwnam_r(', _UNSAFE_FUNC_PREFIX + r'getpwnam\([^)]+\)'), - ('getpwuid(', 'getpwuid_r(', _UNSAFE_FUNC_PREFIX + r'getpwuid\([^)]+\)'), - ('gmtime(', 'gmtime_r(', _UNSAFE_FUNC_PREFIX + r'gmtime\([^)]+\)'), - ('localtime(', 'localtime_r(', _UNSAFE_FUNC_PREFIX + r'localtime\([^)]+\)'), - ('rand(', 'rand_r(', _UNSAFE_FUNC_PREFIX + r'rand\(\)'), - ('strtok(', 'strtok_r(', - _UNSAFE_FUNC_PREFIX + r'strtok\([^)]+\)'), - ('ttyname(', 'ttyname_r(', _UNSAFE_FUNC_PREFIX + r'ttyname\([^)]+\)'), - ) + ("asctime(", "asctime_r(", _UNSAFE_FUNC_PREFIX + r"asctime\([^)]+\)"), + ("ctime(", "ctime_r(", _UNSAFE_FUNC_PREFIX + r"ctime\([^)]+\)"), + ("getgrgid(", "getgrgid_r(", _UNSAFE_FUNC_PREFIX + r"getgrgid\([^)]+\)"), + ("getgrnam(", "getgrnam_r(", _UNSAFE_FUNC_PREFIX + r"getgrnam\([^)]+\)"), + ("getlogin(", "getlogin_r(", _UNSAFE_FUNC_PREFIX + r"getlogin\(\)"), + ("getpwnam(", "getpwnam_r(", _UNSAFE_FUNC_PREFIX + r"getpwnam\([^)]+\)"), + ("getpwuid(", "getpwuid_r(", _UNSAFE_FUNC_PREFIX + r"getpwuid\([^)]+\)"), + ("gmtime(", "gmtime_r(", _UNSAFE_FUNC_PREFIX + r"gmtime\([^)]+\)"), + ("localtime(", "localtime_r(", _UNSAFE_FUNC_PREFIX + r"localtime\([^)]+\)"), + ("rand(", "rand_r(", _UNSAFE_FUNC_PREFIX + r"rand\(\)"), + ("strtok(", "strtok_r(", _UNSAFE_FUNC_PREFIX + r"strtok\([^)]+\)"), + ("ttyname(", "ttyname_r(", _UNSAFE_FUNC_PREFIX + r"ttyname\([^)]+\)"), +) def CheckPosixThreading(filename, clean_lines, linenum, error): - """Checks for calls to thread-unsafe functions. - - Much code has been originally written without consideration of - multi-threading. Also, engineers are relying on their old experience; - they have learned posix before threading extensions were added. These - tests guide the engineers to use thread-safe functions (when using - posix directly). - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - for single_thread_func, multithread_safe_func, pattern in _THREADING_LIST: - # Additional pattern matching check to confirm that this is the - # function we are looking for - if Search(pattern, line): - error(filename, linenum, 'runtime/threadsafe_fn', 2, - 'Consider using ' + multithread_safe_func + - '...) instead of ' + single_thread_func + - '...) for improved thread safety.') + """Checks for calls to thread-unsafe functions. + Much code has been originally written without consideration of + multi-threading. Also, engineers are relying on their old experience; + they have learned posix before threading extensions were added. These + tests guide the engineers to use thread-safe functions (when using + posix directly). -def CheckVlogArguments(filename, clean_lines, linenum, error): - """Checks that VLOG() is only used for defining a logging level. - - For example, VLOG(2) is correct. VLOG(INFO), VLOG(WARNING), VLOG(ERROR), and - VLOG(FATAL) are not. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - if Search(r'\bVLOG\((INFO|ERROR|WARNING|DFATAL|FATAL)\)', line): - error(filename, linenum, 'runtime/vlog', 5, - 'VLOG() should be used with numeric verbosity level. ' - 'Use LOG() if you want symbolic severity levels.') - -# Matches invalid increment: *count++, which moves pointer instead of -# incrementing a value. -_RE_PATTERN_INVALID_INCREMENT = re.compile( - r'^\s*\*\w+(\+\+|--);') - - -def CheckInvalidIncrement(filename, clean_lines, linenum, error): - """Checks for invalid increment *count++. - - For example following function: - void increment_counter(int* count) { - *count++; - } - is invalid, because it effectively does count++, moving pointer, and should - be replaced with ++*count, (*count)++ or *count += 1. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - if _RE_PATTERN_INVALID_INCREMENT.match(line): - error(filename, linenum, 'runtime/invalid_increment', 5, - 'Changing pointer instead of value (or unused value of operator*).') - - -def IsMacroDefinition(clean_lines, linenum): - if Search(r'^#define', clean_lines[linenum]): - return True - - if linenum > 0 and Search(r'\\$', clean_lines[linenum - 1]): - return True - - return False - - -def IsForwardClassDeclaration(clean_lines, linenum): - return Match(r'^\s*(\btemplate\b)*.*class\s+\w+;\s*$', clean_lines[linenum]) - - -class _BlockInfo(object): - """Stores information about a generic block of code.""" + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + for single_thread_func, multithread_safe_func, pattern in _THREADING_LIST: + # Additional pattern matching check to confirm that this is the + # function we are looking for + if Search(pattern, line): + error( + filename, + linenum, + "runtime/threadsafe_fn", + 2, + "Consider using " + + multithread_safe_func + + "...) instead of " + + single_thread_func + + "...) for improved thread safety.", + ) - def __init__(self, linenum, seen_open_brace): - self.starting_linenum = linenum - self.seen_open_brace = seen_open_brace - self.open_parentheses = 0 - self.inline_asm = _NO_ASM - self.check_namespace_indentation = False - def CheckBegin(self, filename, clean_lines, linenum, error): - """Run checks that applies to text up to the opening brace. +def CheckVlogArguments(filename, clean_lines, linenum, error): + """Checks that VLOG() is only used for defining a logging level. - This is mostly for checking the text after the class identifier - and the "{", usually where the base class is specified. For other - blocks, there isn't much to check, so we always pass. + For example, VLOG(2) is correct. VLOG(INFO), VLOG(WARNING), VLOG(ERROR), and + VLOG(FATAL) are not. Args: filename: The name of the current file. @@ -2225,12 +2303,32 @@ def CheckBegin(self, filename, clean_lines, linenum, error): linenum: The number of the line to check. error: The function to call with any errors found. """ - pass + line = clean_lines.elided[linenum] + if Search(r"\bVLOG\((INFO|ERROR|WARNING|DFATAL|FATAL)\)", line): + error( + filename, + linenum, + "runtime/vlog", + 5, + "VLOG() should be used with numeric verbosity level. " + "Use LOG() if you want symbolic severity levels.", + ) + + +# Matches invalid increment: *count++, which moves pointer instead of +# incrementing a value. +_RE_PATTERN_INVALID_INCREMENT = re.compile(r"^\s*\*\w+(\+\+|--);") - def CheckEnd(self, filename, clean_lines, linenum, error): - """Run checks that applies to text after the closing brace. - This is mostly used for checking end of namespace comments. +def CheckInvalidIncrement(filename, clean_lines, linenum, error): + """Checks for invalid increment *count++. + + For example following function: + void increment_counter(int* count) { + *count++; + } + is invalid, because it effectively does count++, moving pointer, and should + be replaced with ++*count, (*count)++ or *count += 1. Args: filename: The name of the current file. @@ -2238,2193 +2336,2659 @@ def CheckEnd(self, filename, clean_lines, linenum, error): linenum: The number of the line to check. error: The function to call with any errors found. """ - pass + line = clean_lines.elided[linenum] + if _RE_PATTERN_INVALID_INCREMENT.match(line): + error( + filename, + linenum, + "runtime/invalid_increment", + 5, + "Changing pointer instead of value (or unused value of operator*).", + ) - def IsBlockInfo(self): - """Returns true if this block is a _BlockInfo. - This is convenient for verifying that an object is an instance of - a _BlockInfo, but not an instance of any of the derived classes. +def IsMacroDefinition(clean_lines, linenum): + if Search(r"^#define", clean_lines[linenum]): + return True - Returns: - True for this class, False for derived classes. - """ - return self.__class__ == _BlockInfo + if linenum > 0 and Search(r"\\$", clean_lines[linenum - 1]): + return True + + return False + + +def IsForwardClassDeclaration(clean_lines, linenum): + return Match(r"^\s*(\btemplate\b)*.*class\s+\w+;\s*$", clean_lines[linenum]) + + +class _BlockInfo(object): + """Stores information about a generic block of code.""" + + def __init__(self, linenum, seen_open_brace): + self.starting_linenum = linenum + self.seen_open_brace = seen_open_brace + self.open_parentheses = 0 + self.inline_asm = _NO_ASM + self.check_namespace_indentation = False + + def CheckBegin(self, filename, clean_lines, linenum, error): + """Run checks that applies to text up to the opening brace. + + This is mostly for checking the text after the class identifier + and the "{", usually where the base class is specified. For other + blocks, there isn't much to check, so we always pass. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + pass + + def CheckEnd(self, filename, clean_lines, linenum, error): + """Run checks that applies to text after the closing brace. + + This is mostly used for checking end of namespace comments. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + pass + + def IsBlockInfo(self): + """Returns true if this block is a _BlockInfo. + + This is convenient for verifying that an object is an instance of + a _BlockInfo, but not an instance of any of the derived classes. + + Returns: + True for this class, False for derived classes. + """ + return self.__class__ == _BlockInfo class _ExternCInfo(_BlockInfo): - """Stores information about an 'extern "C"' block.""" + """Stores information about an 'extern "C"' block.""" - def __init__(self, linenum): - _BlockInfo.__init__(self, linenum, True) + def __init__(self, linenum): + _BlockInfo.__init__(self, linenum, True) class _ClassInfo(_BlockInfo): - """Stores information about a class.""" - - def __init__(self, name, class_or_struct, clean_lines, linenum): - _BlockInfo.__init__(self, linenum, False) - self.name = name - self.is_derived = False - self.check_namespace_indentation = True - if class_or_struct == 'struct': - self.access = 'public' - self.is_struct = True - else: - self.access = 'private' - self.is_struct = False + """Stores information about a class.""" + + def __init__(self, name, class_or_struct, clean_lines, linenum): + _BlockInfo.__init__(self, linenum, False) + self.name = name + self.is_derived = False + self.check_namespace_indentation = True + if class_or_struct == "struct": + self.access = "public" + self.is_struct = True + else: + self.access = "private" + self.is_struct = False - # Remember initial indentation level for this class. Using raw_lines here - # instead of elided to account for leading comments. - self.class_indent = GetIndentLevel(clean_lines.raw_lines[linenum]) + # Remember initial indentation level for this class. Using raw_lines here + # instead of elided to account for leading comments. + self.class_indent = GetIndentLevel(clean_lines.raw_lines[linenum]) - # Try to find the end of the class. This will be confused by things like: - # class A { - # } *x = { ... - # - # But it's still good enough for CheckSectionSpacing. - self.last_line = 0 - depth = 0 - for i in range(linenum, clean_lines.NumLines()): - line = clean_lines.elided[i] - depth += line.count('{') - line.count('}') - if not depth: - self.last_line = i - break - - def CheckBegin(self, filename, clean_lines, linenum, error): - # Look for a bare ':' - if Search('(^|[^:]):($|[^:])', clean_lines.elided[linenum]): - self.is_derived = True - - def CheckEnd(self, filename, clean_lines, linenum, error): - # If there is a DISALLOW macro, it should appear near the end of - # the class. - seen_last_thing_in_class = False - for i in xrange(linenum - 1, self.starting_linenum, -1): - match = Search( - r'\b(DISALLOW_COPY_AND_ASSIGN|DISALLOW_IMPLICIT_CONSTRUCTORS)\(' + - self.name + r'\)', - clean_lines.elided[i]) - if match: - if seen_last_thing_in_class: - error(filename, i, 'readability/constructors', 3, - match.group(1) + ' should be the last thing in the class') - break - - if not Match(r'^\s*$', clean_lines.elided[i]): - seen_last_thing_in_class = True - - # Check that closing brace is aligned with beginning of the class. - # Only do this if the closing brace is indented by only whitespaces. - # This means we will not check single-line class definitions. - indent = Match(r'^( *)\}', clean_lines.elided[linenum]) - if indent and len(indent.group(1)) != self.class_indent: - if self.is_struct: - parent = 'struct ' + self.name - else: - parent = 'class ' + self.name - error(filename, linenum, 'whitespace/indent', 3, - 'Closing brace should be aligned with beginning of %s' % parent) + # Try to find the end of the class. This will be confused by things like: + # class A { + # } *x = { ... + # + # But it's still good enough for CheckSectionSpacing. + self.last_line = 0 + depth = 0 + for i in range(linenum, clean_lines.NumLines()): + line = clean_lines.elided[i] + depth += line.count("{") - line.count("}") + if not depth: + self.last_line = i + break + + def CheckBegin(self, filename, clean_lines, linenum, error): + # Look for a bare ':' + if Search("(^|[^:]):($|[^:])", clean_lines.elided[linenum]): + self.is_derived = True + + def CheckEnd(self, filename, clean_lines, linenum, error): + # If there is a DISALLOW macro, it should appear near the end of + # the class. + seen_last_thing_in_class = False + for i in xrange(linenum - 1, self.starting_linenum, -1): + match = Search( + r"\b(DISALLOW_COPY_AND_ASSIGN|DISALLOW_IMPLICIT_CONSTRUCTORS)\(" + + self.name + + r"\)", + clean_lines.elided[i], + ) + if match: + if seen_last_thing_in_class: + error( + filename, + i, + "readability/constructors", + 3, + match.group(1) + " should be the last thing in the class", + ) + break + + if not Match(r"^\s*$", clean_lines.elided[i]): + seen_last_thing_in_class = True + + # Check that closing brace is aligned with beginning of the class. + # Only do this if the closing brace is indented by only whitespaces. + # This means we will not check single-line class definitions. + indent = Match(r"^( *)\}", clean_lines.elided[linenum]) + if indent and len(indent.group(1)) != self.class_indent: + if self.is_struct: + parent = "struct " + self.name + else: + parent = "class " + self.name + error( + filename, + linenum, + "whitespace/indent", + 3, + "Closing brace should be aligned with beginning of %s" % parent, + ) class _NamespaceInfo(_BlockInfo): - """Stores information about a namespace.""" + """Stores information about a namespace.""" - def __init__(self, name, linenum): - _BlockInfo.__init__(self, linenum, False) - self.name = name or '' - self.check_namespace_indentation = True + def __init__(self, name, linenum): + _BlockInfo.__init__(self, linenum, False) + self.name = name or "" + self.check_namespace_indentation = True - def CheckEnd(self, filename, clean_lines, linenum, error): - """Check end of namespace comments.""" - line = clean_lines.raw_lines[linenum] + def CheckEnd(self, filename, clean_lines, linenum, error): + """Check end of namespace comments.""" + line = clean_lines.raw_lines[linenum] - # Check how many lines is enclosed in this namespace. Don't issue - # warning for missing namespace comments if there aren't enough - # lines. However, do apply checks if there is already an end of - # namespace comment and it's incorrect. - # - # TODO(unknown): We always want to check end of namespace comments - # if a namespace is large, but sometimes we also want to apply the - # check if a short namespace contained nontrivial things (something - # other than forward declarations). There is currently no logic on - # deciding what these nontrivial things are, so this check is - # triggered by namespace size only, which works most of the time. - if (linenum - self.starting_linenum < 10 - and not Match(r'^\s*};*\s*(//|/\*).*\bnamespace\b', line)): - return - - # Look for matching comment at end of namespace. - # - # Note that we accept C style "/* */" comments for terminating - # namespaces, so that code that terminate namespaces inside - # preprocessor macros can be cpplint clean. - # - # We also accept stuff like "// end of namespace ." with the - # period at the end. - # - # Besides these, we don't accept anything else, otherwise we might - # get false negatives when existing comment is a substring of the - # expected namespace. - if self.name: - # Named namespace - if not Match((r'^\s*};*\s*(//|/\*).*\bnamespace\s+' + - re.escape(self.name) + r'[\*/\.\\\s]*$'), - line): - error(filename, linenum, 'readability/namespace', 5, - 'Namespace should be terminated with "// namespace %s"' % - self.name) - else: - # Anonymous namespace - if not Match(r'^\s*};*\s*(//|/\*).*\bnamespace[\*/\.\\\s]*$', line): - # If "// namespace anonymous" or "// anonymous namespace (more text)", - # mention "// anonymous namespace" as an acceptable form - if Match(r'^\s*}.*\b(namespace anonymous|anonymous namespace)\b', line): - error(filename, linenum, 'readability/namespace', 5, - 'Anonymous namespace should be terminated with "// namespace"' - ' or "// anonymous namespace"') + # Check how many lines is enclosed in this namespace. Don't issue + # warning for missing namespace comments if there aren't enough + # lines. However, do apply checks if there is already an end of + # namespace comment and it's incorrect. + # + # TODO(unknown): We always want to check end of namespace comments + # if a namespace is large, but sometimes we also want to apply the + # check if a short namespace contained nontrivial things (something + # other than forward declarations). There is currently no logic on + # deciding what these nontrivial things are, so this check is + # triggered by namespace size only, which works most of the time. + if linenum - self.starting_linenum < 10 and not Match( + r"^\s*};*\s*(//|/\*).*\bnamespace\b", line + ): + return + + # Look for matching comment at end of namespace. + # + # Note that we accept C style "/* */" comments for terminating + # namespaces, so that code that terminate namespaces inside + # preprocessor macros can be cpplint clean. + # + # We also accept stuff like "// end of namespace ." with the + # period at the end. + # + # Besides these, we don't accept anything else, otherwise we might + # get false negatives when existing comment is a substring of the + # expected namespace. + if self.name: + # Named namespace + if not Match( + ( + r"^\s*};*\s*(//|/\*).*\bnamespace\s+" + + re.escape(self.name) + + r"[\*/\.\\\s]*$" + ), + line, + ): + error( + filename, + linenum, + "readability/namespace", + 5, + 'Namespace should be terminated with "// namespace %s"' % self.name, + ) else: - error(filename, linenum, 'readability/namespace', 5, - 'Anonymous namespace should be terminated with "// namespace"') + # Anonymous namespace + if not Match(r"^\s*};*\s*(//|/\*).*\bnamespace[\*/\.\\\s]*$", line): + # If "// namespace anonymous" or "// anonymous namespace (more text)", + # mention "// anonymous namespace" as an acceptable form + if Match(r"^\s*}.*\b(namespace anonymous|anonymous namespace)\b", line): + error( + filename, + linenum, + "readability/namespace", + 5, + 'Anonymous namespace should be terminated with "// namespace"' + ' or "// anonymous namespace"', + ) + else: + error( + filename, + linenum, + "readability/namespace", + 5, + 'Anonymous namespace should be terminated with "// namespace"', + ) class _PreprocessorInfo(object): - """Stores checkpoints of nesting stacks when #if/#else is seen.""" + """Stores checkpoints of nesting stacks when #if/#else is seen.""" - def __init__(self, stack_before_if): - # The entire nesting stack before #if - self.stack_before_if = stack_before_if + def __init__(self, stack_before_if): + # The entire nesting stack before #if + self.stack_before_if = stack_before_if - # The entire nesting stack up to #else - self.stack_before_else = [] + # The entire nesting stack up to #else + self.stack_before_else = [] - # Whether we have already seen #else or #elif - self.seen_else = False + # Whether we have already seen #else or #elif + self.seen_else = False class NestingState(object): - """Holds states related to parsing braces.""" - - def __init__(self): - # Stack for tracking all braces. An object is pushed whenever we - # see a "{", and popped when we see a "}". Only 3 types of - # objects are possible: - # - _ClassInfo: a class or struct. - # - _NamespaceInfo: a namespace. - # - _BlockInfo: some other type of block. - self.stack = [] - - # Top of the previous stack before each Update(). - # - # Because the nesting_stack is updated at the end of each line, we - # had to do some convoluted checks to find out what is the current - # scope at the beginning of the line. This check is simplified by - # saving the previous top of nesting stack. - # - # We could save the full stack, but we only need the top. Copying - # the full nesting stack would slow down cpplint by ~10%. - self.previous_stack_top = [] + """Holds states related to parsing braces.""" + + def __init__(self): + # Stack for tracking all braces. An object is pushed whenever we + # see a "{", and popped when we see a "}". Only 3 types of + # objects are possible: + # - _ClassInfo: a class or struct. + # - _NamespaceInfo: a namespace. + # - _BlockInfo: some other type of block. + self.stack = [] + + # Top of the previous stack before each Update(). + # + # Because the nesting_stack is updated at the end of each line, we + # had to do some convoluted checks to find out what is the current + # scope at the beginning of the line. This check is simplified by + # saving the previous top of nesting stack. + # + # We could save the full stack, but we only need the top. Copying + # the full nesting stack would slow down cpplint by ~10%. + self.previous_stack_top = [] + + # Stack of _PreprocessorInfo objects. + self.pp_stack = [] + + def SeenOpenBrace(self): + """Check if we have seen the opening brace for the innermost block. + + Returns: + True if we have seen the opening brace, False if the innermost + block is still expecting an opening brace. + """ + return (not self.stack) or self.stack[-1].seen_open_brace + + def InNamespaceBody(self): + """Check if we are currently one level inside a namespace body. + + Returns: + True if top of the stack is a namespace block, False otherwise. + """ + return self.stack and isinstance(self.stack[-1], _NamespaceInfo) + + def InExternC(self): + """Check if we are currently one level inside an 'extern "C"' block. + + Returns: + True if top of the stack is an extern block, False otherwise. + """ + return self.stack and isinstance(self.stack[-1], _ExternCInfo) + + def InClassDeclaration(self): + """Check if we are currently one level inside a class or struct declaration. + + Returns: + True if top of the stack is a class/struct, False otherwise. + """ + return self.stack and isinstance(self.stack[-1], _ClassInfo) + + def InAsmBlock(self): + """Check if we are currently one level inside an inline ASM block. + + Returns: + True if the top of the stack is a block containing inline ASM. + """ + return self.stack and self.stack[-1].inline_asm != _NO_ASM + + def InTemplateArgumentList(self, clean_lines, linenum, pos): + """Check if current position is inside template argument list. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + pos: position just after the suspected template argument. + Returns: + True if (linenum, pos) is inside template arguments. + """ + while linenum < clean_lines.NumLines(): + # Find the earliest character that might indicate a template argument + line = clean_lines.elided[linenum] + match = Match(r"^[^{};=\[\]\.<>]*(.)", line[pos:]) + if not match: + linenum += 1 + pos = 0 + continue + token = match.group(1) + pos += len(match.group(0)) + + # These things do not look like template argument list: + # class Suspect { + # class Suspect x; } + if token in ("{", "}", ";"): + return False - # Stack of _PreprocessorInfo objects. - self.pp_stack = [] + # These things look like template argument list: + # template + # template + # template + # template + if token in (">", "=", "[", "]", "."): + return True + + # Check if token is an unmatched '<'. + # If not, move on to the next character. + if token != "<": + pos += 1 + if pos >= len(line): + linenum += 1 + pos = 0 + continue + + # We can't be sure if we just find a single '<', and need to + # find the matching '>'. + (_, end_line, end_pos) = CloseExpression(clean_lines, linenum, pos - 1) + if end_pos < 0: + # Not sure if template argument list or syntax error in file + return False + linenum = end_line + pos = end_pos + return False - def SeenOpenBrace(self): - """Check if we have seen the opening brace for the innermost block. + def UpdatePreprocessor(self, line): + """Update preprocessor stack. + + We need to handle preprocessors due to classes like this: + #ifdef SWIG + struct ResultDetailsPageElementExtensionPoint { + #else + struct ResultDetailsPageElementExtensionPoint : public Extension { + #endif + + We make the following assumptions (good enough for most files): + - Preprocessor condition evaluates to true from #if up to first + #else/#elif/#endif. + + - Preprocessor condition evaluates to false from #else/#elif up + to #endif. We still perform lint checks on these lines, but + these do not affect nesting stack. + + Args: + line: current line to check. + """ + if Match(r"^\s*#\s*(if|ifdef|ifndef)\b", line): + # Beginning of #if block, save the nesting stack here. The saved + # stack will allow us to restore the parsing state in the #else case. + self.pp_stack.append(_PreprocessorInfo(copy.deepcopy(self.stack))) + elif Match(r"^\s*#\s*(else|elif)\b", line): + # Beginning of #else block + if self.pp_stack: + if not self.pp_stack[-1].seen_else: + # This is the first #else or #elif block. Remember the + # whole nesting stack up to this point. This is what we + # keep after the #endif. + self.pp_stack[-1].seen_else = True + self.pp_stack[-1].stack_before_else = copy.deepcopy(self.stack) + + # Restore the stack to how it was before the #if + self.stack = copy.deepcopy(self.pp_stack[-1].stack_before_if) + else: + # TODO(unknown): unexpected #else, issue warning? + pass + elif Match(r"^\s*#\s*endif\b", line): + # End of #if or #else blocks. + if self.pp_stack: + # If we saw an #else, we will need to restore the nesting + # stack to its former state before the #else, otherwise we + # will just continue from where we left off. + if self.pp_stack[-1].seen_else: + # Here we can just use a shallow copy since we are the last + # reference to it. + self.stack = self.pp_stack[-1].stack_before_else + # Drop the corresponding #if + self.pp_stack.pop() + else: + # TODO(unknown): unexpected #endif, issue warning? + pass + + # TODO(unknown): Update() is too long, but we will refactor later. + def Update(self, filename, clean_lines, linenum, error): + """Update nesting state with current line. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Remember top of the previous nesting stack. + # + # The stack is always pushed/popped and not modified in place, so + # we can just do a shallow copy instead of copy.deepcopy. Using + # deepcopy would slow down cpplint by ~28%. + if self.stack: + self.previous_stack_top = self.stack[-1] + else: + self.previous_stack_top = None - Returns: - True if we have seen the opening brace, False if the innermost - block is still expecting an opening brace. - """ - return (not self.stack) or self.stack[-1].seen_open_brace + # Update pp_stack + self.UpdatePreprocessor(line) + + # Count parentheses. This is to avoid adding struct arguments to + # the nesting stack. + if self.stack: + inner_block = self.stack[-1] + depth_change = line.count("(") - line.count(")") + inner_block.open_parentheses += depth_change + + # Also check if we are starting or ending an inline assembly block. + if inner_block.inline_asm in (_NO_ASM, _END_ASM): + if ( + depth_change != 0 + and inner_block.open_parentheses == 1 + and _MATCH_ASM.match(line) + ): + # Enter assembly block + inner_block.inline_asm = _INSIDE_ASM + else: + # Not entering assembly block. If previous line was _END_ASM, + # we will now shift to _NO_ASM state. + inner_block.inline_asm = _NO_ASM + elif ( + inner_block.inline_asm == _INSIDE_ASM + and inner_block.open_parentheses == 0 + ): + # Exit assembly block + inner_block.inline_asm = _END_ASM + + # Consume namespace declaration at the beginning of the line. Do + # this in a loop so that we catch same line declarations like this: + # namespace proto2 { namespace bridge { class MessageSet; } } + while True: + # Match start of namespace. The "\b\s*" below catches namespace + # declarations even if it weren't followed by a whitespace, this + # is so that we don't confuse our namespace checker. The + # missing spaces will be flagged by CheckSpacing. + namespace_decl_match = Match(r"^\s*namespace\b\s*([:\w]+)?(.*)$", line) + if not namespace_decl_match: + break + + new_namespace = _NamespaceInfo(namespace_decl_match.group(1), linenum) + self.stack.append(new_namespace) + + line = namespace_decl_match.group(2) + if line.find("{") != -1: + new_namespace.seen_open_brace = True + line = line[line.find("{") + 1 :] + + # Look for a class declaration in whatever is left of the line + # after parsing namespaces. The regexp accounts for decorated classes + # such as in: + # class LOCKABLE API Object { + # }; + class_decl_match = Match( + r"^(\s*(?:template\s*<[\w\s<>,:]*>\s*)?" + r"(class|struct)\s+(?:[A-Z_]+\s+)*(\w+(?:::\w+)*))" + r"(.*)$", + line, + ) + if class_decl_match and ( + not self.stack or self.stack[-1].open_parentheses == 0 + ): + # We do not want to accept classes that are actually template arguments: + # template , + # template class Ignore3> + # void Function() {}; + # + # To avoid template argument cases, we scan forward and look for + # an unmatched '>'. If we see one, assume we are inside a + # template argument list. + end_declaration = len(class_decl_match.group(1)) + if not self.InTemplateArgumentList(clean_lines, linenum, end_declaration): + self.stack.append( + _ClassInfo( + class_decl_match.group(3), + class_decl_match.group(2), + clean_lines, + linenum, + ) + ) + line = class_decl_match.group(4) + + # If we have not yet seen the opening brace for the innermost block, + # run checks here. + if not self.SeenOpenBrace(): + self.stack[-1].CheckBegin(filename, clean_lines, linenum, error) + + # Update access control if we are inside a class/struct + if self.stack and isinstance(self.stack[-1], _ClassInfo): + classinfo = self.stack[-1] + access_match = Match( + r"^(.*)\b(public|private|protected|signals)(\s+(?:slots\s*)?)?" + r":(?:[^:]|$)", + line, + ) + if access_match: + classinfo.access = access_match.group(2) + + # Check that access keywords are indented +1 space. Skip this + # check if the keywords are not preceded by whitespaces. + indent = access_match.group(1) + if len(indent) != classinfo.class_indent + 1 and Match( + r"^\s*$", indent + ): + if classinfo.is_struct: + parent = "struct " + classinfo.name + else: + parent = "class " + classinfo.name + slots = "" + if access_match.group(3): + slots = access_match.group(3) + error( + filename, + linenum, + "whitespace/indent", + 3, + "%s%s: should be indented +1 space inside %s" + % (access_match.group(2), slots, parent), + ) + + # Consume braces or semicolons from what's left of the line + while True: + # Match first brace, semicolon, or closed parenthesis. + matched = Match(r"^[^{;)}]*([{;)}])(.*)$", line) + if not matched: + break + + token = matched.group(1) + if token == "{": + # If namespace or class hasn't seen a opening brace yet, mark + # namespace/class head as complete. Push a new block onto the + # stack otherwise. + if not self.SeenOpenBrace(): + self.stack[-1].seen_open_brace = True + elif Match(r'^extern\s*"[^"]*"\s*\{', line): + self.stack.append(_ExternCInfo(linenum)) + else: + self.stack.append(_BlockInfo(linenum, True)) + if _MATCH_ASM.match(line): + self.stack[-1].inline_asm = _BLOCK_ASM + + elif token == ";" or token == ")": + # If we haven't seen an opening brace yet, but we already saw + # a semicolon, this is probably a forward declaration. Pop + # the stack for these. + # + # Similarly, if we haven't seen an opening brace yet, but we + # already saw a closing parenthesis, then these are probably + # function arguments with extra "class" or "struct" keywords. + # Also pop these stack for these. + if not self.SeenOpenBrace(): + self.stack.pop() + else: # token == '}' + # Perform end of block checks and pop the stack. + if self.stack: + self.stack[-1].CheckEnd(filename, clean_lines, linenum, error) + self.stack.pop() + line = matched.group(2) + + def InnermostClass(self): + """Get class info on the top of the stack. + + Returns: + A _ClassInfo object if we are inside a class, or None otherwise. + """ + for i in range(len(self.stack), 0, -1): + classinfo = self.stack[i - 1] + if isinstance(classinfo, _ClassInfo): + return classinfo + return None - def InNamespaceBody(self): - """Check if we are currently one level inside a namespace body. + def CheckCompletedBlocks(self, filename, error): + """Checks that all classes and namespaces have been completely parsed. + + Call this when all lines in a file have been processed. + Args: + filename: The name of the current file. + error: The function to call with any errors found. + """ + # Note: This test can result in false positives if #ifdef constructs + # get in the way of brace matching. See the testBuildClass test in + # cpplint_unittest.py for an example of this. + for obj in self.stack: + if isinstance(obj, _ClassInfo): + error( + filename, + obj.starting_linenum, + "build/class", + 5, + "Failed to find complete declaration of class %s" % obj.name, + ) + elif isinstance(obj, _NamespaceInfo): + error( + filename, + obj.starting_linenum, + "build/namespaces", + 5, + "Failed to find complete declaration of namespace %s" % obj.name, + ) + + +def CheckForNonStandardConstructs(filename, clean_lines, linenum, nesting_state, error): + r"""Logs an error if we see certain non-ANSI constructs ignored by gcc-2. + + Complain about several constructs which gcc-2 accepts, but which are + not standard C++. Warning about these in lint is one way to ease the + transition to new compilers. + - put storage class first (e.g. "static const" instead of "const static"). + - "%lld" instead of %qd" in printf-type functions. + - "%1$d" is non-standard in printf-type functions. + - "\%" is an undefined character escape sequence. + - text after #endif is not allowed. + - invalid inner-style forward declaration. + - >? and ?= and )\?=?\s*(\w+|[+-]?\d+)(\.\d*)?", line): + error( + filename, + linenum, + "build/deprecated", + 3, + ">? and ))?' + # r'\s*const\s*' + type_name + '\s*&\s*\w+\s*;' + error( + filename, + linenum, + "runtime/member_string_references", + 2, + "const string& members are dangerous. It is much better to use " + "alternatives, such as pointers or simple constants.", + ) + + # Everything else in this function operates on class declarations. + # Return early if the top of the nesting stack is not a class, or if + # the class head is not completed yet. + classinfo = nesting_state.InnermostClass() + if not classinfo or not classinfo.seen_open_brace: + return - def InClassDeclaration(self): - """Check if we are currently one level inside a class or struct declaration. + # The class may have been declared with namespace or classname qualifiers. + # The constructor and destructor will not have those qualifiers. + base_classname = classinfo.name.split("::")[-1] + + # Look for single-argument constructors that aren't marked explicit. + # Technically a valid construct, but against style. + explicit_constructor_match = Match( + r"\s+(?:(?:inline|constexpr)\s+)*(explicit\s+)?" + r"(?:(?:inline|constexpr)\s+)*%s\s*" + r"\(((?:[^()]|\([^()]*\))*)\)" % re.escape(base_classname), + line, + ) - Returns: - True if top of the stack is a class/struct, False otherwise. - """ - return self.stack and isinstance(self.stack[-1], _ClassInfo) + if explicit_constructor_match: + is_marked_explicit = explicit_constructor_match.group(1) - def InAsmBlock(self): - """Check if we are currently one level inside an inline ASM block. + if not explicit_constructor_match.group(2): + constructor_args = [] + else: + constructor_args = explicit_constructor_match.group(2).split(",") + + # collapse arguments so that commas in template parameter lists and function + # argument parameter lists don't split arguments in two + i = 0 + while i < len(constructor_args): + constructor_arg = constructor_args[i] + while constructor_arg.count("<") > constructor_arg.count( + ">" + ) or constructor_arg.count("(") > constructor_arg.count(")"): + constructor_arg += "," + constructor_args[i + 1] + del constructor_args[i + 1] + constructor_args[i] = constructor_arg + i += 1 + + defaulted_args = [arg for arg in constructor_args if "=" in arg] + noarg_constructor = ( + not constructor_args + or # empty arg list + # 'void' arg specifier + (len(constructor_args) == 1 and constructor_args[0].strip() == "void") + ) + onearg_constructor = ( + (len(constructor_args) == 1 and not noarg_constructor) # exactly one arg + or + # all but at most one arg defaulted + ( + len(constructor_args) >= 1 + and not noarg_constructor + and len(defaulted_args) >= len(constructor_args) - 1 + ) + ) + initializer_list_constructor = bool( + onearg_constructor + and Search(r"\bstd\s*::\s*initializer_list\b", constructor_args[0]) + ) + copy_constructor = bool( + onearg_constructor + and Match( + r"(const\s+)?%s(\s*<[^>]*>)?(\s+const)?\s*(?:<\w+>\s*)?&" + % re.escape(base_classname), + constructor_args[0].strip(), + ) + ) + + if ( + not is_marked_explicit + and onearg_constructor + and not initializer_list_constructor + and not copy_constructor + ): + if defaulted_args: + error( + filename, + linenum, + "runtime/explicit", + 5, + "Constructors callable with one argument " + "should be marked explicit.", + ) + else: + error( + filename, + linenum, + "runtime/explicit", + 5, + "Single-parameter constructors should be marked explicit.", + ) + elif is_marked_explicit and not onearg_constructor: + if noarg_constructor: + error( + filename, + linenum, + "runtime/explicit", + 5, + "Zero-parameter constructors should not be marked explicit.", + ) - Returns: - True if the top of the stack is a block containing inline ASM. - """ - return self.stack and self.stack[-1].inline_asm != _NO_ASM - def InTemplateArgumentList(self, clean_lines, linenum, pos): - """Check if current position is inside template argument list. +def CheckSpacingForFunctionCall(filename, clean_lines, linenum, error): + """Checks for the correctness of various spacing around function calls. Args: + filename: The name of the current file. clean_lines: A CleansedLines instance containing the file. linenum: The number of the line to check. - pos: position just after the suspected template argument. - Returns: - True if (linenum, pos) is inside template arguments. + error: The function to call with any errors found. """ - while linenum < clean_lines.NumLines(): - # Find the earliest character that might indicate a template argument - line = clean_lines.elided[linenum] - match = Match(r'^[^{};=\[\]\.<>]*(.)', line[pos:]) - if not match: - linenum += 1 - pos = 0 - continue - token = match.group(1) - pos += len(match.group(0)) - - # These things do not look like template argument list: - # class Suspect { - # class Suspect x; } - if token in ('{', '}', ';'): return False - - # These things look like template argument list: - # template - # template - # template - # template - if token in ('>', '=', '[', ']', '.'): return True - - # Check if token is an unmatched '<'. - # If not, move on to the next character. - if token != '<': - pos += 1 - if pos >= len(line): - linenum += 1 - pos = 0 - continue - - # We can't be sure if we just find a single '<', and need to - # find the matching '>'. - (_, end_line, end_pos) = CloseExpression(clean_lines, linenum, pos - 1) - if end_pos < 0: - # Not sure if template argument list or syntax error in file - return False - linenum = end_line - pos = end_pos - return False + line = clean_lines.elided[linenum] - def UpdatePreprocessor(self, line): - """Update preprocessor stack. + # Since function calls often occur inside if/for/while/switch + # expressions - which have their own, more liberal conventions - we + # first see if we should be looking inside such an expression for a + # function call, to which we can apply more strict standards. + fncall = line # if there's no control flow construct, look at whole line + for pattern in ( + r"\bif\s*\((.*)\)\s*{", + r"\bfor\s*\((.*)\)\s*{", + r"\bwhile\s*\((.*)\)\s*[{;]", + r"\bswitch\s*\((.*)\)\s*{", + ): + match = Search(pattern, line) + if match: + fncall = match.group(1) # look inside the parens for function calls + break + + # Except in if/for/while/switch, there should never be space + # immediately inside parens (eg "f( 3, 4 )"). We make an exception + # for nested parens ( (a+b) + c ). Likewise, there should never be + # a space before a ( when it's a function argument. I assume it's a + # function argument when the char before the whitespace is legal in + # a function name (alnum + _) and we're not starting a macro. Also ignore + # pointers and references to arrays and functions coz they're too tricky: + # we use a very simple way to recognize these: + # " (something)(maybe-something)" or + # " (something)(maybe-something," or + # " (something)[something]" + # Note that we assume the contents of [] to be short enough that + # they'll never need to wrap. + if ( # Ignore control structures. + not Search(r"\b(if|for|while|switch|return|new|delete|catch|sizeof)\b", fncall) + and + # Ignore pointers/references to functions. + not Search(r" \([^)]+\)\([^)]*(\)|,$)", fncall) + and + # Ignore pointers/references to arrays. + not Search(r" \([^)]+\)\[[^\]]+\]", fncall) + ): + if Search(r"\w\s*\(\s(?!\s*\\$)", fncall): # a ( used for a fn call + error( + filename, + linenum, + "whitespace/parens", + 4, + "Extra space after ( in function call", + ) + elif Search(r"\(\s+(?!(\s*\\)|\()", fncall): + error(filename, linenum, "whitespace/parens", 2, "Extra space after (") + if ( + Search(r"\w\s+\(", fncall) + and not Search(r"_{0,2}asm_{0,2}\s+_{0,2}volatile_{0,2}\s+\(", fncall) + and not Search(r"#\s*define|typedef|using\s+\w+\s*=", fncall) + and not Search(r"\w\s+\((\w+::)*\*\w+\)\(", fncall) + and not Search(r"\bcase\s+\(", fncall) + ): + # TODO(unknown): Space after an operator function seem to be a common + # error, silence those for now by restricting them to highest verbosity. + if Search(r"\boperator_*\b", line): + error( + filename, + linenum, + "whitespace/parens", + 0, + "Extra space before ( in function call", + ) + else: + error( + filename, + linenum, + "whitespace/parens", + 4, + "Extra space before ( in function call", + ) + # If the ) is followed only by a newline or a { + newline, assume it's + # part of a control statement (if/while/etc), and don't complain + if Search(r"[^)]\s+\)\s*[^{\s]", fncall): + # If the closing parenthesis is preceded by only whitespaces, + # try to give a more descriptive error message. + if Search(r"^\s+\)", fncall): + error( + filename, + linenum, + "whitespace/parens", + 2, + "Closing ) should be moved to the previous line", + ) + else: + error(filename, linenum, "whitespace/parens", 2, "Extra space before )") - We need to handle preprocessors due to classes like this: - #ifdef SWIG - struct ResultDetailsPageElementExtensionPoint { - #else - struct ResultDetailsPageElementExtensionPoint : public Extension { - #endif - We make the following assumptions (good enough for most files): - - Preprocessor condition evaluates to true from #if up to first - #else/#elif/#endif. +def IsBlankLine(line): + """Returns true if the given line is blank. - - Preprocessor condition evaluates to false from #else/#elif up - to #endif. We still perform lint checks on these lines, but - these do not affect nesting stack. + We consider a line to be blank if the line is empty or consists of + only white spaces. Args: - line: current line to check. + line: A line of a string. + + Returns: + True, if the given line is blank. """ - if Match(r'^\s*#\s*(if|ifdef|ifndef)\b', line): - # Beginning of #if block, save the nesting stack here. The saved - # stack will allow us to restore the parsing state in the #else case. - self.pp_stack.append(_PreprocessorInfo(copy.deepcopy(self.stack))) - elif Match(r'^\s*#\s*(else|elif)\b', line): - # Beginning of #else block - if self.pp_stack: - if not self.pp_stack[-1].seen_else: - # This is the first #else or #elif block. Remember the - # whole nesting stack up to this point. This is what we - # keep after the #endif. - self.pp_stack[-1].seen_else = True - self.pp_stack[-1].stack_before_else = copy.deepcopy(self.stack) - - # Restore the stack to how it was before the #if - self.stack = copy.deepcopy(self.pp_stack[-1].stack_before_if) - else: - # TODO(unknown): unexpected #else, issue warning? - pass - elif Match(r'^\s*#\s*endif\b', line): - # End of #if or #else blocks. - if self.pp_stack: - # If we saw an #else, we will need to restore the nesting - # stack to its former state before the #else, otherwise we - # will just continue from where we left off. - if self.pp_stack[-1].seen_else: - # Here we can just use a shallow copy since we are the last - # reference to it. - self.stack = self.pp_stack[-1].stack_before_else - # Drop the corresponding #if - self.pp_stack.pop() - else: - # TODO(unknown): unexpected #endif, issue warning? - pass + return not line or line.isspace() + + +def CheckForNamespaceIndentation(filename, nesting_state, clean_lines, line, error): + is_namespace_indent_item = ( + len(nesting_state.stack) > 1 + and nesting_state.stack[-1].check_namespace_indentation + and isinstance(nesting_state.previous_stack_top, _NamespaceInfo) + and nesting_state.previous_stack_top == nesting_state.stack[-2] + ) + + if ShouldCheckNamespaceIndentation( + nesting_state, is_namespace_indent_item, clean_lines.elided, line + ): + CheckItemIndentationInNamespace(filename, clean_lines.elided, line, error) + - # TODO(unknown): Update() is too long, but we will refactor later. - def Update(self, filename, clean_lines, linenum, error): - """Update nesting state with current line. +def CheckForFunctionLengths(filename, clean_lines, linenum, function_state, error): + """Reports for long function bodies. + + For an overview why this is done, see: + https://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Write_Short_Functions + + Uses a simplistic algorithm assuming other style guidelines + (especially spacing) are followed. + Only checks unindented functions, so class members are unchecked. + Trivial bodies are unchecked, so constructors with huge initializer lists + may be missed. + Blank/comment lines are not counted so as to avoid encouraging the removal + of vertical space and comments just to get through a lint check. + NOLINT *on the last line of a function* disables this check. Args: filename: The name of the current file. clean_lines: A CleansedLines instance containing the file. linenum: The number of the line to check. + function_state: Current function name and lines in body so far. error: The function to call with any errors found. """ - line = clean_lines.elided[linenum] + lines = clean_lines.lines + line = lines[linenum] + joined_line = "" + + starting_func = False + regexp = r"(\w(\w|::|\*|\&|\s)*)\(" # decls * & space::name( ... + match_result = Match(regexp, line) + if match_result: + # If the name is all caps and underscores, figure it's a macro and + # ignore it, unless it's TEST or TEST_F. + function_name = match_result.group(1).split()[-1] + if ( + function_name == "TEST" + or function_name == "TEST_F" + or (not Match(r"[A-Z_]+$", function_name)) + ): + starting_func = True + + if starting_func: + body_found = False + for start_linenum in xrange(linenum, clean_lines.NumLines()): + start_line = lines[start_linenum] + joined_line += " " + start_line.lstrip() + if Search(r"(;|})", start_line): # Declarations and trivial functions + body_found = True + break # ... ignore + elif Search(r"{", start_line): + body_found = True + function = Search(r"((\w|:)*)\(", line).group(1) + if Match(r"TEST", function): # Handle TEST... macros + parameter_regexp = Search(r"(\(.*\))", joined_line) + if parameter_regexp: # Ignore bad syntax + function += parameter_regexp.group(1) + else: + function += "()" + function_state.Begin(function) + break + if not body_found: + # No body for the function (or evidence of a non-function) was found. + error( + filename, + linenum, + "readability/fn_size", + 5, + "Lint failed to find start of function body.", + ) + elif Match(r"^\}\s*$", line): # function end + function_state.Check(error, filename, linenum) + function_state.End() + elif not Match(r"^\s*$", line): + function_state.Count() # Count non-blank/non-comment lines. + + +_RE_PATTERN_TODO = re.compile(r"^//(\s*)TODO(\(.+?\))?:?(\s|$)?") - # Remember top of the previous nesting stack. - # - # The stack is always pushed/popped and not modified in place, so - # we can just do a shallow copy instead of copy.deepcopy. Using - # deepcopy would slow down cpplint by ~28%. - if self.stack: - self.previous_stack_top = self.stack[-1] - else: - self.previous_stack_top = None - - # Update pp_stack - self.UpdatePreprocessor(line) - - # Count parentheses. This is to avoid adding struct arguments to - # the nesting stack. - if self.stack: - inner_block = self.stack[-1] - depth_change = line.count('(') - line.count(')') - inner_block.open_parentheses += depth_change - - # Also check if we are starting or ending an inline assembly block. - if inner_block.inline_asm in (_NO_ASM, _END_ASM): - if (depth_change != 0 and - inner_block.open_parentheses == 1 and - _MATCH_ASM.match(line)): - # Enter assembly block - inner_block.inline_asm = _INSIDE_ASM - else: - # Not entering assembly block. If previous line was _END_ASM, - # we will now shift to _NO_ASM state. - inner_block.inline_asm = _NO_ASM - elif (inner_block.inline_asm == _INSIDE_ASM and - inner_block.open_parentheses == 0): - # Exit assembly block - inner_block.inline_asm = _END_ASM - - # Consume namespace declaration at the beginning of the line. Do - # this in a loop so that we catch same line declarations like this: - # namespace proto2 { namespace bridge { class MessageSet; } } - while True: - # Match start of namespace. The "\b\s*" below catches namespace - # declarations even if it weren't followed by a whitespace, this - # is so that we don't confuse our namespace checker. The - # missing spaces will be flagged by CheckSpacing. - namespace_decl_match = Match(r'^\s*namespace\b\s*([:\w]+)?(.*)$', line) - if not namespace_decl_match: - break - - new_namespace = _NamespaceInfo(namespace_decl_match.group(1), linenum) - self.stack.append(new_namespace) - - line = namespace_decl_match.group(2) - if line.find('{') != -1: - new_namespace.seen_open_brace = True - line = line[line.find('{') + 1:] - - # Look for a class declaration in whatever is left of the line - # after parsing namespaces. The regexp accounts for decorated classes - # such as in: - # class LOCKABLE API Object { - # }; - class_decl_match = Match( - r'^(\s*(?:template\s*<[\w\s<>,:]*>\s*)?' - r'(class|struct)\s+(?:[A-Z_]+\s+)*(\w+(?:::\w+)*))' - r'(.*)$', line) - if (class_decl_match and - (not self.stack or self.stack[-1].open_parentheses == 0)): - # We do not want to accept classes that are actually template arguments: - # template , - # template class Ignore3> - # void Function() {}; - # - # To avoid template argument cases, we scan forward and look for - # an unmatched '>'. If we see one, assume we are inside a - # template argument list. - end_declaration = len(class_decl_match.group(1)) - if not self.InTemplateArgumentList(clean_lines, linenum, end_declaration): - self.stack.append(_ClassInfo( - class_decl_match.group(3), class_decl_match.group(2), - clean_lines, linenum)) - line = class_decl_match.group(4) - - # If we have not yet seen the opening brace for the innermost block, - # run checks here. - if not self.SeenOpenBrace(): - self.stack[-1].CheckBegin(filename, clean_lines, linenum, error) - - # Update access control if we are inside a class/struct - if self.stack and isinstance(self.stack[-1], _ClassInfo): - classinfo = self.stack[-1] - access_match = Match( - r'^(.*)\b(public|private|protected|signals)(\s+(?:slots\s*)?)?' - r':(?:[^:]|$)', - line) - if access_match: - classinfo.access = access_match.group(2) - - # Check that access keywords are indented +1 space. Skip this - # check if the keywords are not preceded by whitespaces. - indent = access_match.group(1) - if (len(indent) != classinfo.class_indent + 1 and - Match(r'^\s*$', indent)): - if classinfo.is_struct: - parent = 'struct ' + classinfo.name - else: - parent = 'class ' + classinfo.name - slots = '' - if access_match.group(3): - slots = access_match.group(3) - error(filename, linenum, 'whitespace/indent', 3, - '%s%s: should be indented +1 space inside %s' % ( - access_match.group(2), slots, parent)) - - # Consume braces or semicolons from what's left of the line - while True: - # Match first brace, semicolon, or closed parenthesis. - matched = Match(r'^[^{;)}]*([{;)}])(.*)$', line) - if not matched: - break - - token = matched.group(1) - if token == '{': - # If namespace or class hasn't seen a opening brace yet, mark - # namespace/class head as complete. Push a new block onto the - # stack otherwise. - if not self.SeenOpenBrace(): - self.stack[-1].seen_open_brace = True - elif Match(r'^extern\s*"[^"]*"\s*\{', line): - self.stack.append(_ExternCInfo(linenum)) - else: - self.stack.append(_BlockInfo(linenum, True)) - if _MATCH_ASM.match(line): - self.stack[-1].inline_asm = _BLOCK_ASM - - elif token == ';' or token == ')': - # If we haven't seen an opening brace yet, but we already saw - # a semicolon, this is probably a forward declaration. Pop - # the stack for these. - # - # Similarly, if we haven't seen an opening brace yet, but we - # already saw a closing parenthesis, then these are probably - # function arguments with extra "class" or "struct" keywords. - # Also pop these stack for these. - if not self.SeenOpenBrace(): - self.stack.pop() - else: # token == '}' - # Perform end of block checks and pop the stack. - if self.stack: - self.stack[-1].CheckEnd(filename, clean_lines, linenum, error) - self.stack.pop() - line = matched.group(2) - def InnermostClass(self): - """Get class info on the top of the stack. +def CheckComment(line, filename, linenum, next_line_start, error): + """Checks for common mistakes in comments. - Returns: - A _ClassInfo object if we are inside a class, or None otherwise. + Args: + line: The line in question. + filename: The name of the current file. + linenum: The number of the line to check. + next_line_start: The first non-whitespace column of the next line. + error: The function to call with any errors found. """ - for i in range(len(self.stack), 0, -1): - classinfo = self.stack[i - 1] - if isinstance(classinfo, _ClassInfo): - return classinfo - return None + commentpos = line.find("//") + if commentpos != -1: + # Check if the // may be in quotes. If so, ignore it + if re.sub(r"\\.", "", line[0:commentpos]).count('"') % 2 == 0: + # Allow one space for new scopes, two spaces otherwise: + if not (Match(r"^.*{ *//", line) and next_line_start == commentpos) and ( + (commentpos >= 1 and line[commentpos - 1] not in string.whitespace) + or (commentpos >= 2 and line[commentpos - 2] not in string.whitespace) + ): + error( + filename, + linenum, + "whitespace/comments", + 2, + "At least two spaces is best between code and comments", + ) + + # Checks for common mistakes in TODO comments. + comment = line[commentpos:] + match = _RE_PATTERN_TODO.match(comment) + if match: + # One whitespace is correct; zero whitespace is handled elsewhere. + leading_whitespace = match.group(1) + if len(leading_whitespace) > 1: + error( + filename, + linenum, + "whitespace/todo", + 2, + "Too many spaces before TODO", + ) + + username = match.group(2) + if not username: + error( + filename, + linenum, + "readability/todo", + 2, + "Missing username in TODO; it should look like " + '"// TODO(my_username): Stuff."', + ) + + middle_whitespace = match.group(3) + # Comparisons made explicit for correctness -- pylint: disable=g-explicit-bool-comparison + if middle_whitespace != " " and middle_whitespace != "": + error( + filename, + linenum, + "whitespace/todo", + 2, + "TODO(my_username) should be followed by a space", + ) + + # If the comment contains an alphanumeric character, there + # should be a space somewhere between it and the // unless + # it's a /// or //! Doxygen comment. + if Match(r"//[^ ]*\w", comment) and not Match( + r"(///|//\!)(\s+|$)", comment + ): + error( + filename, + linenum, + "whitespace/comments", + 4, + "Should have a space between // and comment", + ) + - def CheckCompletedBlocks(self, filename, error): - """Checks that all classes and namespaces have been completely parsed. +def CheckSpacing(filename, clean_lines, linenum, nesting_state, error): + """Checks for the correctness of various spacing issues in the code. + + Things we check for: spaces around operators, spaces after + if/for/while/switch, no spaces around parens in function calls, two + spaces between code and comment, don't start a block with a blank + line, don't end a function with a blank line, don't add a blank line + after public/protected/private, don't have too many blank lines in a row. - Call this when all lines in a file have been processed. Args: filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. error: The function to call with any errors found. """ - # Note: This test can result in false positives if #ifdef constructs - # get in the way of brace matching. See the testBuildClass test in - # cpplint_unittest.py for an example of this. - for obj in self.stack: - if isinstance(obj, _ClassInfo): - error(filename, obj.starting_linenum, 'build/class', 5, - 'Failed to find complete declaration of class %s' % - obj.name) - elif isinstance(obj, _NamespaceInfo): - error(filename, obj.starting_linenum, 'build/namespaces', 5, - 'Failed to find complete declaration of namespace %s' % - obj.name) - - -def CheckForNonStandardConstructs(filename, clean_lines, linenum, - nesting_state, error): - r"""Logs an error if we see certain non-ANSI constructs ignored by gcc-2. - - Complain about several constructs which gcc-2 accepts, but which are - not standard C++. Warning about these in lint is one way to ease the - transition to new compilers. - - put storage class first (e.g. "static const" instead of "const static"). - - "%lld" instead of %qd" in printf-type functions. - - "%1$d" is non-standard in printf-type functions. - - "\%" is an undefined character escape sequence. - - text after #endif is not allowed. - - invalid inner-style forward declaration. - - >? and ?= and )\?=?\s*(\w+|[+-]?\d+)(\.\d*)?', - line): - error(filename, linenum, 'build/deprecated', 3, - '>? and ))?' - # r'\s*const\s*' + type_name + '\s*&\s*\w+\s*;' - error(filename, linenum, 'runtime/member_string_references', 2, - 'const string& members are dangerous. It is much better to use ' - 'alternatives, such as pointers or simple constants.') - - # Everything else in this function operates on class declarations. - # Return early if the top of the nesting stack is not a class, or if - # the class head is not completed yet. - classinfo = nesting_state.InnermostClass() - if not classinfo or not classinfo.seen_open_brace: - return - - # The class may have been declared with namespace or classname qualifiers. - # The constructor and destructor will not have those qualifiers. - base_classname = classinfo.name.split('::')[-1] - - # Look for single-argument constructors that aren't marked explicit. - # Technically a valid construct, but against style. - explicit_constructor_match = Match( - r'\s+(?:(?:inline|constexpr)\s+)*(explicit\s+)?' - r'(?:(?:inline|constexpr)\s+)*%s\s*' - r'\(((?:[^()]|\([^()]*\))*)\)' - % re.escape(base_classname), - line) - - if explicit_constructor_match: - is_marked_explicit = explicit_constructor_match.group(1) - - if not explicit_constructor_match.group(2): - constructor_args = [] - else: - constructor_args = explicit_constructor_match.group(2).split(',') - - # collapse arguments so that commas in template parameter lists and function - # argument parameter lists don't split arguments in two - i = 0 - while i < len(constructor_args): - constructor_arg = constructor_args[i] - while (constructor_arg.count('<') > constructor_arg.count('>') or - constructor_arg.count('(') > constructor_arg.count(')')): - constructor_arg += ',' + constructor_args[i + 1] - del constructor_args[i + 1] - constructor_args[i] = constructor_arg - i += 1 - - defaulted_args = [arg for arg in constructor_args if '=' in arg] - noarg_constructor = (not constructor_args or # empty arg list - # 'void' arg specifier - (len(constructor_args) == 1 and - constructor_args[0].strip() == 'void')) - onearg_constructor = ((len(constructor_args) == 1 and # exactly one arg - not noarg_constructor) or - # all but at most one arg defaulted - (len(constructor_args) >= 1 and - not noarg_constructor and - len(defaulted_args) >= len(constructor_args) - 1)) - initializer_list_constructor = bool( - onearg_constructor and - Search(r'\bstd\s*::\s*initializer_list\b', constructor_args[0])) - copy_constructor = bool( - onearg_constructor and - Match(r'(const\s+)?%s(\s*<[^>]*>)?(\s+const)?\s*(?:<\w+>\s*)?&' - % re.escape(base_classname), constructor_args[0].strip())) - - if (not is_marked_explicit and - onearg_constructor and - not initializer_list_constructor and - not copy_constructor): - if defaulted_args: - error(filename, linenum, 'runtime/explicit', 5, - 'Constructors callable with one argument ' - 'should be marked explicit.') - else: - error(filename, linenum, 'runtime/explicit', 5, - 'Single-parameter constructors should be marked explicit.') - elif is_marked_explicit and not onearg_constructor: - if noarg_constructor: - error(filename, linenum, 'runtime/explicit', 5, - 'Zero-parameter constructors should not be marked explicit.') - -def CheckSpacingForFunctionCall(filename, clean_lines, linenum, error): - """Checks for the correctness of various spacing around function calls. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - # Since function calls often occur inside if/for/while/switch - # expressions - which have their own, more liberal conventions - we - # first see if we should be looking inside such an expression for a - # function call, to which we can apply more strict standards. - fncall = line # if there's no control flow construct, look at whole line - for pattern in (r'\bif\s*\((.*)\)\s*{', - r'\bfor\s*\((.*)\)\s*{', - r'\bwhile\s*\((.*)\)\s*[{;]', - r'\bswitch\s*\((.*)\)\s*{'): - match = Search(pattern, line) - if match: - fncall = match.group(1) # look inside the parens for function calls - break - - # Except in if/for/while/switch, there should never be space - # immediately inside parens (eg "f( 3, 4 )"). We make an exception - # for nested parens ( (a+b) + c ). Likewise, there should never be - # a space before a ( when it's a function argument. I assume it's a - # function argument when the char before the whitespace is legal in - # a function name (alnum + _) and we're not starting a macro. Also ignore - # pointers and references to arrays and functions coz they're too tricky: - # we use a very simple way to recognize these: - # " (something)(maybe-something)" or - # " (something)(maybe-something," or - # " (something)[something]" - # Note that we assume the contents of [] to be short enough that - # they'll never need to wrap. - if ( # Ignore control structures. - not Search(r'\b(if|for|while|switch|return|new|delete|catch|sizeof)\b', - fncall) and - # Ignore pointers/references to functions. - not Search(r' \([^)]+\)\([^)]*(\)|,$)', fncall) and - # Ignore pointers/references to arrays. - not Search(r' \([^)]+\)\[[^\]]+\]', fncall)): - if Search(r'\w\s*\(\s(?!\s*\\$)', fncall): # a ( used for a fn call - error(filename, linenum, 'whitespace/parens', 4, - 'Extra space after ( in function call') - elif Search(r'\(\s+(?!(\s*\\)|\()', fncall): - error(filename, linenum, 'whitespace/parens', 2, - 'Extra space after (') - if (Search(r'\w\s+\(', fncall) and - not Search(r'_{0,2}asm_{0,2}\s+_{0,2}volatile_{0,2}\s+\(', fncall) and - not Search(r'#\s*define|typedef|using\s+\w+\s*=', fncall) and - not Search(r'\w\s+\((\w+::)*\*\w+\)\(', fncall) and - not Search(r'\bcase\s+\(', fncall)): - # TODO(unknown): Space after an operator function seem to be a common - # error, silence those for now by restricting them to highest verbosity. - if Search(r'\boperator_*\b', line): - error(filename, linenum, 'whitespace/parens', 0, - 'Extra space before ( in function call') - else: - error(filename, linenum, 'whitespace/parens', 4, - 'Extra space before ( in function call') - # If the ) is followed only by a newline or a { + newline, assume it's - # part of a control statement (if/while/etc), and don't complain - if Search(r'[^)]\s+\)\s*[^{\s]', fncall): - # If the closing parenthesis is preceded by only whitespaces, - # try to give a more descriptive error message. - if Search(r'^\s+\)', fncall): - error(filename, linenum, 'whitespace/parens', 2, - 'Closing ) should be moved to the previous line') - else: - error(filename, linenum, 'whitespace/parens', 2, - 'Extra space before )') + # Don't use "elided" lines here, otherwise we can't check commented lines. + # Don't want to use "raw" either, because we don't want to check inside C++11 + # raw strings, + raw = clean_lines.lines_without_raw_strings + line = raw[linenum] + # Before nixing comments, check if the line is blank for no good + # reason. This includes the first line after a block is opened, and + # blank lines at the end of a function (ie, right before a line like '}' + # + # Skip all the blank line checks if we are immediately inside a + # namespace body. In other words, don't issue blank line warnings + # for this block: + # namespace { + # + # } + # + # A warning about missing end of namespace comments will be issued instead. + # + # Also skip blank line checks for 'extern "C"' blocks, which are formatted + # like namespaces. + if ( + IsBlankLine(line) + and not nesting_state.InNamespaceBody() + and not nesting_state.InExternC() + ): + elided = clean_lines.elided + prev_line = elided[linenum - 1] + prevbrace = prev_line.rfind("{") + # TODO(unknown): Don't complain if line before blank line, and line after, + # both start with alnums and are indented the same amount. + # This ignores whitespace at the start of a namespace block + # because those are not usually indented. + if prevbrace != -1 and prev_line[prevbrace:].find("}") == -1: + # OK, we have a blank line at the start of a code block. Before we + # complain, we check if it is an exception to the rule: The previous + # non-empty line has the parameters of a function header that are indented + # 4 spaces (because they did not fit in a 80 column line when placed on + # the same line as the function name). We also check for the case where + # the previous line is indented 6 spaces, which may happen when the + # initializers of a constructor do not fit into a 80 column line. + exception = False + if Match(r" {6}\w", prev_line): # Initializer list? + # We are looking for the opening column of initializer list, which + # should be indented 4 spaces to cause 6 space indentation afterwards. + search_position = linenum - 2 + while search_position >= 0 and Match( + r" {6}\w", elided[search_position] + ): + search_position -= 1 + exception = ( + search_position >= 0 and elided[search_position][:5] == " :" + ) + else: + # Search for the function arguments or an initializer list. We use a + # simple heuristic here: If the line is indented 4 spaces; and we have a + # closing paren, without the opening paren, followed by an opening brace + # or colon (for initializer lists) we assume that it is the last line of + # a function header. If we have a colon indented 4 spaces, it is an + # initializer list. + exception = Match( + r" {4}\w[^\(]*\)\s*(const\s*)?(\{\s*$|:)", prev_line + ) or Match(r" {4}:", prev_line) + + if not exception: + error( + filename, + linenum, + "whitespace/blank_line", + 2, + "Redundant blank line at the start of a code block " + "should be deleted.", + ) + # Ignore blank lines at the end of a block in a long if-else + # chain, like this: + # if (condition1) { + # // Something followed by a blank line + # + # } else if (condition2) { + # // Something else + # } + if linenum + 1 < clean_lines.NumLines(): + next_line = raw[linenum + 1] + if ( + next_line + and Match(r"\s*}", next_line) + and next_line.find("} else ") == -1 + ): + error( + filename, + linenum, + "whitespace/blank_line", + 3, + "Redundant blank line at the end of a code block " + "should be deleted.", + ) + + matched = Match(r"\s*(public|protected|private):", prev_line) + if matched: + error( + filename, + linenum, + "whitespace/blank_line", + 3, + 'Do not leave a blank line after "%s:"' % matched.group(1), + ) + + # Next, check comments + next_line_start = 0 + if linenum + 1 < clean_lines.NumLines(): + next_line = raw[linenum + 1] + next_line_start = len(next_line) - len(next_line.lstrip()) + CheckComment(line, filename, linenum, next_line_start, error) -def IsBlankLine(line): - """Returns true if the given line is blank. - - We consider a line to be blank if the line is empty or consists of - only white spaces. - - Args: - line: A line of a string. - - Returns: - True, if the given line is blank. - """ - return not line or line.isspace() - - -def CheckForNamespaceIndentation(filename, nesting_state, clean_lines, line, - error): - is_namespace_indent_item = ( - len(nesting_state.stack) > 1 and - nesting_state.stack[-1].check_namespace_indentation and - isinstance(nesting_state.previous_stack_top, _NamespaceInfo) and - nesting_state.previous_stack_top == nesting_state.stack[-2]) - - if ShouldCheckNamespaceIndentation(nesting_state, is_namespace_indent_item, - clean_lines.elided, line): - CheckItemIndentationInNamespace(filename, clean_lines.elided, - line, error) - - -def CheckForFunctionLengths(filename, clean_lines, linenum, - function_state, error): - """Reports for long function bodies. - - For an overview why this is done, see: - https://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Write_Short_Functions - - Uses a simplistic algorithm assuming other style guidelines - (especially spacing) are followed. - Only checks unindented functions, so class members are unchecked. - Trivial bodies are unchecked, so constructors with huge initializer lists - may be missed. - Blank/comment lines are not counted so as to avoid encouraging the removal - of vertical space and comments just to get through a lint check. - NOLINT *on the last line of a function* disables this check. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - function_state: Current function name and lines in body so far. - error: The function to call with any errors found. - """ - lines = clean_lines.lines - line = lines[linenum] - joined_line = '' - - starting_func = False - regexp = r'(\w(\w|::|\*|\&|\s)*)\(' # decls * & space::name( ... - match_result = Match(regexp, line) - if match_result: - # If the name is all caps and underscores, figure it's a macro and - # ignore it, unless it's TEST or TEST_F. - function_name = match_result.group(1).split()[-1] - if function_name == 'TEST' or function_name == 'TEST_F' or ( - not Match(r'[A-Z_]+$', function_name)): - starting_func = True - - if starting_func: - body_found = False - for start_linenum in xrange(linenum, clean_lines.NumLines()): - start_line = lines[start_linenum] - joined_line += ' ' + start_line.lstrip() - if Search(r'(;|})', start_line): # Declarations and trivial functions - body_found = True - break # ... ignore - elif Search(r'{', start_line): - body_found = True - function = Search(r'((\w|:)*)\(', line).group(1) - if Match(r'TEST', function): # Handle TEST... macros - parameter_regexp = Search(r'(\(.*\))', joined_line) - if parameter_regexp: # Ignore bad syntax - function += parameter_regexp.group(1) - else: - function += '()' - function_state.Begin(function) - break - if not body_found: - # No body for the function (or evidence of a non-function) was found. - error(filename, linenum, 'readability/fn_size', 5, - 'Lint failed to find start of function body.') - elif Match(r'^\}\s*$', line): # function end - function_state.Check(error, filename, linenum) - function_state.End() - elif not Match(r'^\s*$', line): - function_state.Count() # Count non-blank/non-comment lines. + # get rid of comments and strings + line = clean_lines.elided[linenum] + # You shouldn't have spaces before your brackets, except maybe after + # 'delete []' or 'return []() {};' + if Search(r"\w\s+\[", line) and not Search(r"(?:delete|return)\s+\[", line): + error(filename, linenum, "whitespace/braces", 5, "Extra space before [") -_RE_PATTERN_TODO = re.compile(r'^//(\s*)TODO(\(.+?\))?:?(\s|$)?') + # In range-based for, we wanted spaces before and after the colon, but + # not around "::" tokens that might appear. + if Search(r"for *\(.*[^:]:[^: ]", line) or Search(r"for *\(.*[^: ]:[^:]", line): + error( + filename, + linenum, + "whitespace/forcolon", + 2, + "Missing space around colon in range-based for loop", + ) -def CheckComment(line, filename, linenum, next_line_start, error): - """Checks for common mistakes in comments. - - Args: - line: The line in question. - filename: The name of the current file. - linenum: The number of the line to check. - next_line_start: The first non-whitespace column of the next line. - error: The function to call with any errors found. - """ - commentpos = line.find('//') - if commentpos != -1: - # Check if the // may be in quotes. If so, ignore it - if re.sub(r'\\.', '', line[0:commentpos]).count('"') % 2 == 0: - # Allow one space for new scopes, two spaces otherwise: - if (not (Match(r'^.*{ *//', line) and next_line_start == commentpos) and - ((commentpos >= 1 and - line[commentpos-1] not in string.whitespace) or - (commentpos >= 2 and - line[commentpos-2] not in string.whitespace))): - error(filename, linenum, 'whitespace/comments', 2, - 'At least two spaces is best between code and comments') - - # Checks for common mistakes in TODO comments. - comment = line[commentpos:] - match = _RE_PATTERN_TODO.match(comment) - if match: - # One whitespace is correct; zero whitespace is handled elsewhere. - leading_whitespace = match.group(1) - if len(leading_whitespace) > 1: - error(filename, linenum, 'whitespace/todo', 2, - 'Too many spaces before TODO') - - username = match.group(2) - if not username: - error(filename, linenum, 'readability/todo', 2, - 'Missing username in TODO; it should look like ' - '"// TODO(my_username): Stuff."') - - middle_whitespace = match.group(3) - # Comparisons made explicit for correctness -- pylint: disable=g-explicit-bool-comparison - if middle_whitespace != ' ' and middle_whitespace != '': - error(filename, linenum, 'whitespace/todo', 2, - 'TODO(my_username) should be followed by a space') - - # If the comment contains an alphanumeric character, there - # should be a space somewhere between it and the // unless - # it's a /// or //! Doxygen comment. - if (Match(r'//[^ ]*\w', comment) and - not Match(r'(///|//\!)(\s+|$)', comment)): - error(filename, linenum, 'whitespace/comments', 4, - 'Should have a space between // and comment') +def CheckOperatorSpacing(filename, clean_lines, linenum, error): + """Checks for horizontal spacing around operators. + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] -def CheckSpacing(filename, clean_lines, linenum, nesting_state, error): - """Checks for the correctness of various spacing issues in the code. - - Things we check for: spaces around operators, spaces after - if/for/while/switch, no spaces around parens in function calls, two - spaces between code and comment, don't start a block with a blank - line, don't end a function with a blank line, don't add a blank line - after public/protected/private, don't have too many blank lines in a row. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - nesting_state: A NestingState instance which maintains information about - the current stack of nested blocks being parsed. - error: The function to call with any errors found. - """ - - # Don't use "elided" lines here, otherwise we can't check commented lines. - # Don't want to use "raw" either, because we don't want to check inside C++11 - # raw strings, - raw = clean_lines.lines_without_raw_strings - line = raw[linenum] - - # Before nixing comments, check if the line is blank for no good - # reason. This includes the first line after a block is opened, and - # blank lines at the end of a function (ie, right before a line like '}' - # - # Skip all the blank line checks if we are immediately inside a - # namespace body. In other words, don't issue blank line warnings - # for this block: - # namespace { - # - # } - # - # A warning about missing end of namespace comments will be issued instead. - # - # Also skip blank line checks for 'extern "C"' blocks, which are formatted - # like namespaces. - if (IsBlankLine(line) and - not nesting_state.InNamespaceBody() and - not nesting_state.InExternC()): - elided = clean_lines.elided - prev_line = elided[linenum - 1] - prevbrace = prev_line.rfind('{') - # TODO(unknown): Don't complain if line before blank line, and line after, - # both start with alnums and are indented the same amount. - # This ignores whitespace at the start of a namespace block - # because those are not usually indented. - if prevbrace != -1 and prev_line[prevbrace:].find('}') == -1: - # OK, we have a blank line at the start of a code block. Before we - # complain, we check if it is an exception to the rule: The previous - # non-empty line has the parameters of a function header that are indented - # 4 spaces (because they did not fit in a 80 column line when placed on - # the same line as the function name). We also check for the case where - # the previous line is indented 6 spaces, which may happen when the - # initializers of a constructor do not fit into a 80 column line. - exception = False - if Match(r' {6}\w', prev_line): # Initializer list? - # We are looking for the opening column of initializer list, which - # should be indented 4 spaces to cause 6 space indentation afterwards. - search_position = linenum-2 - while (search_position >= 0 - and Match(r' {6}\w', elided[search_position])): - search_position -= 1 - exception = (search_position >= 0 - and elided[search_position][:5] == ' :') - else: - # Search for the function arguments or an initializer list. We use a - # simple heuristic here: If the line is indented 4 spaces; and we have a - # closing paren, without the opening paren, followed by an opening brace - # or colon (for initializer lists) we assume that it is the last line of - # a function header. If we have a colon indented 4 spaces, it is an - # initializer list. - exception = (Match(r' {4}\w[^\(]*\)\s*(const\s*)?(\{\s*$|:)', - prev_line) - or Match(r' {4}:', prev_line)) - - if not exception: - error(filename, linenum, 'whitespace/blank_line', 2, - 'Redundant blank line at the start of a code block ' - 'should be deleted.') - # Ignore blank lines at the end of a block in a long if-else - # chain, like this: - # if (condition1) { - # // Something followed by a blank line + # Don't try to do spacing checks for operator methods. Do this by + # replacing the troublesome characters with something else, + # preserving column position for all other characters. # - # } else if (condition2) { - # // Something else - # } - if linenum + 1 < clean_lines.NumLines(): - next_line = raw[linenum + 1] - if (next_line - and Match(r'\s*}', next_line) - and next_line.find('} else ') == -1): - error(filename, linenum, 'whitespace/blank_line', 3, - 'Redundant blank line at the end of a code block ' - 'should be deleted.') - - matched = Match(r'\s*(public|protected|private):', prev_line) - if matched: - error(filename, linenum, 'whitespace/blank_line', 3, - 'Do not leave a blank line after "%s:"' % matched.group(1)) + # The replacement is done repeatedly to avoid false positives from + # operators that call operators. + while True: + match = Match(r"^(.*\boperator\b)(\S+)(\s*\(.*)$", line) + if match: + line = match.group(1) + ("_" * len(match.group(2))) + match.group(3) + else: + break - # Next, check comments - next_line_start = 0 - if linenum + 1 < clean_lines.NumLines(): - next_line = raw[linenum + 1] - next_line_start = len(next_line) - len(next_line.lstrip()) - CheckComment(line, filename, linenum, next_line_start, error) + # We allow no-spaces around = within an if: "if ( (a=Foo()) == 0 )". + # Otherwise not. Note we only check for non-spaces on *both* sides; + # sometimes people put non-spaces on one side when aligning ='s among + # many lines (not that this is behavior that I approve of...) + if ( + (Search(r"[\w.]=", line) or Search(r"=[\w.]", line)) + and not Search(r"\b(if|while|for) ", line) + # Operators taken from [lex.operators] in C++11 standard. + and not Search(r"(>=|<=|==|!=|&=|\^=|\|=|\+=|\*=|\/=|\%=)", line) + and not Search(r"operator=", line) + ): + error(filename, linenum, "whitespace/operators", 4, "Missing spaces around =") + + # It's ok not to have spaces around binary operators like + - * /, but if + # there's too little whitespace, we get concerned. It's hard to tell, + # though, so we punt on this one for now. TODO. + + # You should always have whitespace around binary operators. + # + # Check <= and >= first to avoid false positives with < and >, then + # check non-include lines for spacing around < and >. + # + # If the operator is followed by a comma, assume it's be used in a + # macro context and don't do any checks. This avoids false + # positives. + # + # Note that && is not included here. This is because there are too + # many false positives due to RValue references. + match = Search(r"[^<>=!\s](==|!=|<=|>=|\|\|)[^<>=!\s,;\)]", line) + if match: + error( + filename, + linenum, + "whitespace/operators", + 3, + "Missing spaces around %s" % match.group(1), + ) + elif not Match(r"#.*include", line): + # Look for < that is not surrounded by spaces. This is only + # triggered if both sides are missing spaces, even though + # technically should should flag if at least one side is missing a + # space. This is done to avoid some false positives with shifts. + match = Match(r"^(.*[^\s<])<[^\s=<,]", line) + if match: + (_, _, end_pos) = CloseExpression(clean_lines, linenum, len(match.group(1))) + if end_pos <= -1: + error( + filename, + linenum, + "whitespace/operators", + 3, + "Missing spaces around <", + ) + + # Look for > that is not surrounded by spaces. Similar to the + # above, we only trigger if both sides are missing spaces to avoid + # false positives with shifts. + match = Match(r"^(.*[^-\s>])>[^\s=>,]", line) + if match: + (_, _, start_pos) = ReverseCloseExpression( + clean_lines, linenum, len(match.group(1)) + ) + if start_pos <= -1: + error( + filename, + linenum, + "whitespace/operators", + 3, + "Missing spaces around >", + ) + + # We allow no-spaces around << when used like this: 10<<20, but + # not otherwise (particularly, not when used as streams) + # + # We also allow operators following an opening parenthesis, since + # those tend to be macros that deal with operators. + match = Search(r"(operator|[^\s(<])(?:L|UL|LL|ULL|l|ul|ll|ull)?<<([^\s,=<])", line) + if ( + match + and not (match.group(1).isdigit() and match.group(2).isdigit()) + and not (match.group(1) == "operator" and match.group(2) == ";") + ): + error(filename, linenum, "whitespace/operators", 3, "Missing spaces around <<") + + # We allow no-spaces around >> for almost anything. This is because + # C++11 allows ">>" to close nested templates, which accounts for + # most cases when ">>" is not followed by a space. + # + # We still warn on ">>" followed by alpha character, because that is + # likely due to ">>" being used for right shifts, e.g.: + # value >> alpha + # + # When ">>" is used to close templates, the alphanumeric letter that + # follows would be part of an identifier, and there should still be + # a space separating the template type and the identifier. + # type> alpha + match = Search(r">>[a-zA-Z_]", line) + if match: + error(filename, linenum, "whitespace/operators", 3, "Missing spaces around >>") - # get rid of comments and strings - line = clean_lines.elided[linenum] + # There shouldn't be space around unary operators + match = Search(r"(!\s|~\s|[\s]--[\s;]|[\s]\+\+[\s;])", line) + if match: + error( + filename, + linenum, + "whitespace/operators", + 4, + "Extra space for operator %s" % match.group(1), + ) - # You shouldn't have spaces before your brackets, except maybe after - # 'delete []' or 'return []() {};' - if Search(r'\w\s+\[', line) and not Search(r'(?:delete|return)\s+\[', line): - error(filename, linenum, 'whitespace/braces', 5, - 'Extra space before [') - # In range-based for, we wanted spaces before and after the colon, but - # not around "::" tokens that might appear. - if (Search(r'for *\(.*[^:]:[^: ]', line) or - Search(r'for *\(.*[^: ]:[^:]', line)): - error(filename, linenum, 'whitespace/forcolon', 2, - 'Missing space around colon in range-based for loop') +def CheckParenthesisSpacing(filename, clean_lines, linenum, error): + """Checks for horizontal spacing around parentheses. + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] -def CheckOperatorSpacing(filename, clean_lines, linenum, error): - """Checks for horizontal spacing around operators. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - # Don't try to do spacing checks for operator methods. Do this by - # replacing the troublesome characters with something else, - # preserving column position for all other characters. - # - # The replacement is done repeatedly to avoid false positives from - # operators that call operators. - while True: - match = Match(r'^(.*\boperator\b)(\S+)(\s*\(.*)$', line) - if match: - line = match.group(1) + ('_' * len(match.group(2))) + match.group(3) - else: - break - - # We allow no-spaces around = within an if: "if ( (a=Foo()) == 0 )". - # Otherwise not. Note we only check for non-spaces on *both* sides; - # sometimes people put non-spaces on one side when aligning ='s among - # many lines (not that this is behavior that I approve of...) - if ((Search(r'[\w.]=', line) or - Search(r'=[\w.]', line)) - and not Search(r'\b(if|while|for) ', line) - # Operators taken from [lex.operators] in C++11 standard. - and not Search(r'(>=|<=|==|!=|&=|\^=|\|=|\+=|\*=|\/=|\%=)', line) - and not Search(r'operator=', line)): - error(filename, linenum, 'whitespace/operators', 4, - 'Missing spaces around =') - - # It's ok not to have spaces around binary operators like + - * /, but if - # there's too little whitespace, we get concerned. It's hard to tell, - # though, so we punt on this one for now. TODO. - - # You should always have whitespace around binary operators. - # - # Check <= and >= first to avoid false positives with < and >, then - # check non-include lines for spacing around < and >. - # - # If the operator is followed by a comma, assume it's be used in a - # macro context and don't do any checks. This avoids false - # positives. - # - # Note that && is not included here. This is because there are too - # many false positives due to RValue references. - match = Search(r'[^<>=!\s](==|!=|<=|>=|\|\|)[^<>=!\s,;\)]', line) - if match: - error(filename, linenum, 'whitespace/operators', 3, - 'Missing spaces around %s' % match.group(1)) - elif not Match(r'#.*include', line): - # Look for < that is not surrounded by spaces. This is only - # triggered if both sides are missing spaces, even though - # technically should should flag if at least one side is missing a - # space. This is done to avoid some false positives with shifts. - match = Match(r'^(.*[^\s<])<[^\s=<,]', line) + # No spaces after an if, while, switch, or for + match = Search(r" (if\(|for\(|while\(|switch\()", line) if match: - (_, _, end_pos) = CloseExpression( - clean_lines, linenum, len(match.group(1))) - if end_pos <= -1: - error(filename, linenum, 'whitespace/operators', 3, - 'Missing spaces around <') - - # Look for > that is not surrounded by spaces. Similar to the - # above, we only trigger if both sides are missing spaces to avoid - # false positives with shifts. - match = Match(r'^(.*[^-\s>])>[^\s=>,]', line) + error( + filename, + linenum, + "whitespace/parens", + 5, + "Missing space before ( in %s" % match.group(1), + ) + + # For if/for/while/switch, the left and right parens should be + # consistent about how many spaces are inside the parens, and + # there should either be zero or one spaces inside the parens. + # We don't want: "if ( foo)" or "if ( foo )". + # Exception: "for ( ; foo; bar)" and "for (foo; bar; )" are allowed. + match = Search( + r"\b(if|for|while|switch)\s*" r"\(([ ]*)(.).*[^ ]+([ ]*)\)\s*{\s*$", line + ) if match: - (_, _, start_pos) = ReverseCloseExpression( - clean_lines, linenum, len(match.group(1))) - if start_pos <= -1: - error(filename, linenum, 'whitespace/operators', 3, - 'Missing spaces around >') - - # We allow no-spaces around << when used like this: 10<<20, but - # not otherwise (particularly, not when used as streams) - # - # We also allow operators following an opening parenthesis, since - # those tend to be macros that deal with operators. - match = Search(r'(operator|[^\s(<])(?:L|UL|LL|ULL|l|ul|ll|ull)?<<([^\s,=<])', line) - if (match and not (match.group(1).isdigit() and match.group(2).isdigit()) and - not (match.group(1) == 'operator' and match.group(2) == ';')): - error(filename, linenum, 'whitespace/operators', 3, - 'Missing spaces around <<') - - # We allow no-spaces around >> for almost anything. This is because - # C++11 allows ">>" to close nested templates, which accounts for - # most cases when ">>" is not followed by a space. - # - # We still warn on ">>" followed by alpha character, because that is - # likely due to ">>" being used for right shifts, e.g.: - # value >> alpha - # - # When ">>" is used to close templates, the alphanumeric letter that - # follows would be part of an identifier, and there should still be - # a space separating the template type and the identifier. - # type> alpha - match = Search(r'>>[a-zA-Z_]', line) - if match: - error(filename, linenum, 'whitespace/operators', 3, - 'Missing spaces around >>') - - # There shouldn't be space around unary operators - match = Search(r'(!\s|~\s|[\s]--[\s;]|[\s]\+\+[\s;])', line) - if match: - error(filename, linenum, 'whitespace/operators', 4, - 'Extra space for operator %s' % match.group(1)) + if len(match.group(2)) != len(match.group(4)): + if not ( + match.group(3) == ";" + and len(match.group(2)) == 1 + len(match.group(4)) + or not match.group(2) + and Search(r"\bfor\s*\(.*; \)", line) + ): + error( + filename, + linenum, + "whitespace/parens", + 5, + "Mismatching spaces inside () in %s" % match.group(1), + ) + if len(match.group(2)) not in [0, 1]: + error( + filename, + linenum, + "whitespace/parens", + 5, + "Should have zero or one spaces inside ( and ) in %s" % match.group(1), + ) -def CheckParenthesisSpacing(filename, clean_lines, linenum, error): - """Checks for horizontal spacing around parentheses. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - # No spaces after an if, while, switch, or for - match = Search(r' (if\(|for\(|while\(|switch\()', line) - if match: - error(filename, linenum, 'whitespace/parens', 5, - 'Missing space before ( in %s' % match.group(1)) - - # For if/for/while/switch, the left and right parens should be - # consistent about how many spaces are inside the parens, and - # there should either be zero or one spaces inside the parens. - # We don't want: "if ( foo)" or "if ( foo )". - # Exception: "for ( ; foo; bar)" and "for (foo; bar; )" are allowed. - match = Search(r'\b(if|for|while|switch)\s*' - r'\(([ ]*)(.).*[^ ]+([ ]*)\)\s*{\s*$', - line) - if match: - if len(match.group(2)) != len(match.group(4)): - if not (match.group(3) == ';' and - len(match.group(2)) == 1 + len(match.group(4)) or - not match.group(2) and Search(r'\bfor\s*\(.*; \)', line)): - error(filename, linenum, 'whitespace/parens', 5, - 'Mismatching spaces inside () in %s' % match.group(1)) - if len(match.group(2)) not in [0, 1]: - error(filename, linenum, 'whitespace/parens', 5, - 'Should have zero or one spaces inside ( and ) in %s' % - match.group(1)) +def CheckCommaSpacing(filename, clean_lines, linenum, error): + """Checks for horizontal spacing near commas and semicolons. + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + raw = clean_lines.lines_without_raw_strings + line = clean_lines.elided[linenum] -def CheckCommaSpacing(filename, clean_lines, linenum, error): - """Checks for horizontal spacing near commas and semicolons. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - raw = clean_lines.lines_without_raw_strings - line = clean_lines.elided[linenum] - - # You should always have a space after a comma (either as fn arg or operator) - # - # This does not apply when the non-space character following the - # comma is another comma, since the only time when that happens is - # for empty macro arguments. - # - # We run this check in two passes: first pass on elided lines to - # verify that lines contain missing whitespaces, second pass on raw - # lines to confirm that those missing whitespaces are not due to - # elided comments. - if (Search(r',[^,\s]', ReplaceAll(r'\boperator\s*,\s*\(', 'F(', line)) and - Search(r',[^,\s]', raw[linenum])): - error(filename, linenum, 'whitespace/comma', 3, - 'Missing space after ,') - - # You should always have a space after a semicolon - # except for few corner cases - # TODO(unknown): clarify if 'if (1) { return 1;}' is requires one more - # space after ; - if Search(r';[^\s};\\)/]', line): - error(filename, linenum, 'whitespace/semicolon', 3, - 'Missing space after ;') + # You should always have a space after a comma (either as fn arg or operator) + # + # This does not apply when the non-space character following the + # comma is another comma, since the only time when that happens is + # for empty macro arguments. + # + # We run this check in two passes: first pass on elided lines to + # verify that lines contain missing whitespaces, second pass on raw + # lines to confirm that those missing whitespaces are not due to + # elided comments. + if Search(r",[^,\s]", ReplaceAll(r"\boperator\s*,\s*\(", "F(", line)) and Search( + r",[^,\s]", raw[linenum] + ): + error(filename, linenum, "whitespace/comma", 3, "Missing space after ,") + + # You should always have a space after a semicolon + # except for few corner cases + # TODO(unknown): clarify if 'if (1) { return 1;}' is requires one more + # space after ; + if Search(r";[^\s};\\)/]", line): + error(filename, linenum, "whitespace/semicolon", 3, "Missing space after ;") def _IsType(clean_lines, nesting_state, expr): - """Check if expression looks like a type name, returns true if so. - - Args: - clean_lines: A CleansedLines instance containing the file. - nesting_state: A NestingState instance which maintains information about - the current stack of nested blocks being parsed. - expr: The expression to check. - Returns: - True, if token looks like a type. - """ - # Keep only the last token in the expression - last_word = Match(r'^.*(\b\S+)$', expr) - if last_word: - token = last_word.group(1) - else: - token = expr - - # Match native types and stdint types - if _TYPES.match(token): - return True - - # Try a bit harder to match templated types. Walk up the nesting - # stack until we find something that resembles a typename - # declaration for what we are looking for. - typename_pattern = (r'\b(?:typename|class|struct)\s+' + re.escape(token) + - r'\b') - block_index = len(nesting_state.stack) - 1 - while block_index >= 0: - if isinstance(nesting_state.stack[block_index], _NamespaceInfo): - return False - - # Found where the opening brace is. We want to scan from this - # line up to the beginning of the function, minus a few lines. - # template - # class C - # : public ... { // start scanning here - last_line = nesting_state.stack[block_index].starting_linenum - - next_block_start = 0 - if block_index > 0: - next_block_start = nesting_state.stack[block_index - 1].starting_linenum - first_line = last_line - while first_line >= next_block_start: - if clean_lines.elided[first_line].find('template') >= 0: - break - first_line -= 1 - if first_line < next_block_start: - # Didn't find any "template" keyword before reaching the next block, - # there are probably no template things to check for this block - block_index -= 1 - continue - - # Look for typename in the specified range - for i in xrange(first_line, last_line + 1, 1): - if Search(typename_pattern, clean_lines.elided[i]): - return True - block_index -= 1 + """Check if expression looks like a type name, returns true if so. - return False + Args: + clean_lines: A CleansedLines instance containing the file. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + expr: The expression to check. + Returns: + True, if token looks like a type. + """ + # Keep only the last token in the expression + last_word = Match(r"^.*(\b\S+)$", expr) + if last_word: + token = last_word.group(1) + else: + token = expr + # Match native types and stdint types + if _TYPES.match(token): + return True -def CheckBracesSpacing(filename, clean_lines, linenum, nesting_state, error): - """Checks for horizontal spacing near commas. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - nesting_state: A NestingState instance which maintains information about - the current stack of nested blocks being parsed. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - # Except after an opening paren, or after another opening brace (in case of - # an initializer list, for instance), you should have spaces before your - # braces when they are delimiting blocks, classes, namespaces etc. - # And since you should never have braces at the beginning of a line, - # this is an easy test. Except that braces used for initialization don't - # follow the same rule; we often don't want spaces before those. - match = Match(r'^(.*[^ ({>]){', line) - - if match: - # Try a bit harder to check for brace initialization. This - # happens in one of the following forms: - # Constructor() : initializer_list_{} { ... } - # Constructor{}.MemberFunction() - # Type variable{}; - # FunctionCall(type{}, ...); - # LastArgument(..., type{}); - # LOG(INFO) << type{} << " ..."; - # map_of_type[{...}] = ...; - # ternary = expr ? new type{} : nullptr; - # OuterTemplate{}> - # - # We check for the character following the closing brace, and - # silence the warning if it's one of those listed above, i.e. - # "{.;,)<>]:". - # - # To account for nested initializer list, we allow any number of - # closing braces up to "{;,)<". We can't simply silence the - # warning on first sight of closing brace, because that would - # cause false negatives for things that are not initializer lists. - # Silence this: But not this: - # Outer{ if (...) { - # Inner{...} if (...){ // Missing space before { - # }; } - # - # There is a false negative with this approach if people inserted - # spurious semicolons, e.g. "if (cond){};", but we will catch the - # spurious semicolon with a separate check. - leading_text = match.group(1) - (endline, endlinenum, endpos) = CloseExpression( - clean_lines, linenum, len(match.group(1))) - trailing_text = '' - if endpos > -1: - trailing_text = endline[endpos:] - for offset in xrange(endlinenum + 1, - min(endlinenum + 3, clean_lines.NumLines() - 1)): - trailing_text += clean_lines.elided[offset] - # We also suppress warnings for `uint64_t{expression}` etc., as the style - # guide recommends brace initialization for integral types to avoid - # overflow/truncation. - if (not Match(r'^[\s}]*[{.;,)<>\]:]', trailing_text) - and not _IsType(clean_lines, nesting_state, leading_text)): - error(filename, linenum, 'whitespace/braces', 5, - 'Missing space before {') - - # Make sure '} else {' has spaces. - if Search(r'}else', line): - error(filename, linenum, 'whitespace/braces', 5, - 'Missing space before else') - - # You shouldn't have a space before a semicolon at the end of the line. - # There's a special case for "for" since the style guide allows space before - # the semicolon there. - if Search(r':\s*;\s*$', line): - error(filename, linenum, 'whitespace/semicolon', 5, - 'Semicolon defining empty statement. Use {} instead.') - elif Search(r'^\s*;\s*$', line): - error(filename, linenum, 'whitespace/semicolon', 5, - 'Line contains only semicolon. If this should be an empty statement, ' - 'use {} instead.') - elif (Search(r'\s+;\s*$', line) and - not Search(r'\bfor\b', line)): - error(filename, linenum, 'whitespace/semicolon', 5, - 'Extra space before last semicolon. If this should be an empty ' - 'statement, use {} instead.') + # Try a bit harder to match templated types. Walk up the nesting + # stack until we find something that resembles a typename + # declaration for what we are looking for. + typename_pattern = r"\b(?:typename|class|struct)\s+" + re.escape(token) + r"\b" + block_index = len(nesting_state.stack) - 1 + while block_index >= 0: + if isinstance(nesting_state.stack[block_index], _NamespaceInfo): + return False + + # Found where the opening brace is. We want to scan from this + # line up to the beginning of the function, minus a few lines. + # template + # class C + # : public ... { // start scanning here + last_line = nesting_state.stack[block_index].starting_linenum + + next_block_start = 0 + if block_index > 0: + next_block_start = nesting_state.stack[block_index - 1].starting_linenum + first_line = last_line + while first_line >= next_block_start: + if clean_lines.elided[first_line].find("template") >= 0: + break + first_line -= 1 + if first_line < next_block_start: + # Didn't find any "template" keyword before reaching the next block, + # there are probably no template things to check for this block + block_index -= 1 + continue + # Look for typename in the specified range + for i in xrange(first_line, last_line + 1, 1): + if Search(typename_pattern, clean_lines.elided[i]): + return True + block_index -= 1 -def IsDecltype(clean_lines, linenum, column): - """Check if the token ending on (linenum, column) is decltype(). - - Args: - clean_lines: A CleansedLines instance containing the file. - linenum: the number of the line to check. - column: end column of the token to check. - Returns: - True if this token is decltype() expression, False otherwise. - """ - (text, _, start_col) = ReverseCloseExpression(clean_lines, linenum, column) - if start_col < 0: return False - if Search(r'\bdecltype\s*$', text[0:start_col]): - return True - return False -def CheckSectionSpacing(filename, clean_lines, class_info, linenum, error): - """Checks for additional blank line issues related to sections. - - Currently the only thing checked here is blank line before protected/private. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - class_info: A _ClassInfo objects. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - # Skip checks if the class is small, where small means 25 lines or less. - # 25 lines seems like a good cutoff since that's the usual height of - # terminals, and any class that can't fit in one screen can't really - # be considered "small". - # - # Also skip checks if we are on the first line. This accounts for - # classes that look like - # class Foo { public: ... }; - # - # If we didn't find the end of the class, last_line would be zero, - # and the check will be skipped by the first condition. - if (class_info.last_line - class_info.starting_linenum <= 24 or - linenum <= class_info.starting_linenum): - return - - matched = Match(r'\s*(public|protected|private):', clean_lines.lines[linenum]) - if matched: - # Issue warning if the line before public/protected/private was - # not a blank line, but don't do this if the previous line contains - # "class" or "struct". This can happen two ways: - # - We are at the beginning of the class. - # - We are forward-declaring an inner class that is semantically - # private, but needed to be public for implementation reasons. - # Also ignores cases where the previous line ends with a backslash as can be - # common when defining classes in C macros. - prev_line = clean_lines.lines[linenum - 1] - if (not IsBlankLine(prev_line) and - not Search(r'\b(class|struct)\b', prev_line) and - not Search(r'\\$', prev_line)): - # Try a bit harder to find the beginning of the class. This is to - # account for multi-line base-specifier lists, e.g.: - # class Derived - # : public Base { - end_class_head = class_info.starting_linenum - for i in range(class_info.starting_linenum, linenum): - if Search(r'\{\s*$', clean_lines.lines[i]): - end_class_head = i - break - if end_class_head < linenum - 1: - error(filename, linenum, 'whitespace/blank_line', 3, - '"%s:" should be preceded by a blank line' % matched.group(1)) +def CheckBracesSpacing(filename, clean_lines, linenum, nesting_state, error): + """Checks for horizontal spacing near commas. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + # Except after an opening paren, or after another opening brace (in case of + # an initializer list, for instance), you should have spaces before your + # braces when they are delimiting blocks, classes, namespaces etc. + # And since you should never have braces at the beginning of a line, + # this is an easy test. Except that braces used for initialization don't + # follow the same rule; we often don't want spaces before those. + match = Match(r"^(.*[^ ({>]){", line) -def GetPreviousNonBlankLine(clean_lines, linenum): - """Return the most recent non-blank line and its line number. + if match: + # Try a bit harder to check for brace initialization. This + # happens in one of the following forms: + # Constructor() : initializer_list_{} { ... } + # Constructor{}.MemberFunction() + # Type variable{}; + # FunctionCall(type{}, ...); + # LastArgument(..., type{}); + # LOG(INFO) << type{} << " ..."; + # map_of_type[{...}] = ...; + # ternary = expr ? new type{} : nullptr; + # OuterTemplate{}> + # + # We check for the character following the closing brace, and + # silence the warning if it's one of those listed above, i.e. + # "{.;,)<>]:". + # + # To account for nested initializer list, we allow any number of + # closing braces up to "{;,)<". We can't simply silence the + # warning on first sight of closing brace, because that would + # cause false negatives for things that are not initializer lists. + # Silence this: But not this: + # Outer{ if (...) { + # Inner{...} if (...){ // Missing space before { + # }; } + # + # There is a false negative with this approach if people inserted + # spurious semicolons, e.g. "if (cond){};", but we will catch the + # spurious semicolon with a separate check. + leading_text = match.group(1) + (endline, endlinenum, endpos) = CloseExpression( + clean_lines, linenum, len(match.group(1)) + ) + trailing_text = "" + if endpos > -1: + trailing_text = endline[endpos:] + for offset in xrange( + endlinenum + 1, min(endlinenum + 3, clean_lines.NumLines() - 1) + ): + trailing_text += clean_lines.elided[offset] + # We also suppress warnings for `uint64_t{expression}` etc., as the style + # guide recommends brace initialization for integral types to avoid + # overflow/truncation. + if not Match(r"^[\s}]*[{.;,)<>\]:]", trailing_text) and not _IsType( + clean_lines, nesting_state, leading_text + ): + error(filename, linenum, "whitespace/braces", 5, "Missing space before {") + + # Make sure '} else {' has spaces. + if Search(r"}else", line): + error(filename, linenum, "whitespace/braces", 5, "Missing space before else") + + # You shouldn't have a space before a semicolon at the end of the line. + # There's a special case for "for" since the style guide allows space before + # the semicolon there. + if Search(r":\s*;\s*$", line): + error( + filename, + linenum, + "whitespace/semicolon", + 5, + "Semicolon defining empty statement. Use {} instead.", + ) + elif Search(r"^\s*;\s*$", line): + error( + filename, + linenum, + "whitespace/semicolon", + 5, + "Line contains only semicolon. If this should be an empty statement, " + "use {} instead.", + ) + elif Search(r"\s+;\s*$", line) and not Search(r"\bfor\b", line): + error( + filename, + linenum, + "whitespace/semicolon", + 5, + "Extra space before last semicolon. If this should be an empty " + "statement, use {} instead.", + ) - Args: - clean_lines: A CleansedLines instance containing the file contents. - linenum: The number of the line to check. - Returns: - A tuple with two elements. The first element is the contents of the last - non-blank line before the current line, or the empty string if this is the - first non-blank line. The second is the line number of that line, or -1 - if this is the first non-blank line. - """ +def IsDecltype(clean_lines, linenum, column): + """Check if the token ending on (linenum, column) is decltype(). - prevlinenum = linenum - 1 - while prevlinenum >= 0: - prevline = clean_lines.elided[prevlinenum] - if not IsBlankLine(prevline): # if not a blank line... - return (prevline, prevlinenum) - prevlinenum -= 1 - return ('', -1) + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: the number of the line to check. + column: end column of the token to check. + Returns: + True if this token is decltype() expression, False otherwise. + """ + (text, _, start_col) = ReverseCloseExpression(clean_lines, linenum, column) + if start_col < 0: + return False + if Search(r"\bdecltype\s*$", text[0:start_col]): + return True + return False -def CheckBraces(filename, clean_lines, linenum, error): - """Looks for misplaced braces (e.g. at the end of line). - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - - line = clean_lines.elided[linenum] # get rid of comments and strings - - if Match(r'\s*{\s*$', line): - # We allow an open brace to start a line in the case where someone is using - # braces in a block to explicitly create a new scope, which is commonly used - # to control the lifetime of stack-allocated variables. Braces are also - # used for brace initializers inside function calls. We don't detect this - # perfectly: we just don't complain if the last non-whitespace character on - # the previous non-blank line is ',', ';', ':', '(', '{', or '}', or if the - # previous line starts a preprocessor block. We also allow a brace on the - # following line if it is part of an array initialization and would not fit - # within the 80 character limit of the preceding line. - prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0] - if (not Search(r'[,;:}{(]\s*$', prevline) and - not Match(r'\s*#', prevline) and - not (GetLineWidth(prevline) > _line_length - 2 and '[]' in prevline)): - error(filename, linenum, 'whitespace/braces', 4, - '{ should almost always be at the end of the previous line') - - # An else clause should be on the same line as the preceding closing brace. - if Match(r'\s*else\b\s*(?:if\b|\{|$)', line): - prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0] - if Match(r'\s*}\s*$', prevline): - error(filename, linenum, 'whitespace/newline', 4, - 'An else should appear on the same line as the preceding }') - - # If braces come on one side of an else, they should be on both. - # However, we have to worry about "else if" that spans multiple lines! - if Search(r'else if\s*\(', line): # could be multi-line if - brace_on_left = bool(Search(r'}\s*else if\s*\(', line)) - # find the ( after the if - pos = line.find('else if') - pos = line.find('(', pos) - if pos > 0: - (endline, _, endpos) = CloseExpression(clean_lines, linenum, pos) - brace_on_right = endline[endpos:].find('{') != -1 - if brace_on_left != brace_on_right: # must be brace after if - error(filename, linenum, 'readability/braces', 5, - 'If an else has a brace on one side, it should have it on both') - elif Search(r'}\s*else[^{]*$', line) or Match(r'[^}]*else\s*{', line): - error(filename, linenum, 'readability/braces', 5, - 'If an else has a brace on one side, it should have it on both') - - # Likewise, an else should never have the else clause on the same line - if Search(r'\belse [^\s{]', line) and not Search(r'\belse if\b', line): - error(filename, linenum, 'whitespace/newline', 4, - 'Else clause should never be on same line as else (use 2 lines)') - - # In the same way, a do/while should never be on one line - if Match(r'\s*do [^\s{]', line): - error(filename, linenum, 'whitespace/newline', 4, - 'do/while clauses should not be on a single line') - - # Check single-line if/else bodies. The style guide says 'curly braces are not - # required for single-line statements'. We additionally allow multi-line, - # single statements, but we reject anything with more than one semicolon in - # it. This means that the first semicolon after the if should be at the end of - # its line, and the line after that should have an indent level equal to or - # lower than the if. We also check for ambiguous if/else nesting without - # braces. - if_else_match = Search(r'\b(if\s*\(|else\b)', line) - if if_else_match and not Match(r'\s*#', line): - if_indent = GetIndentLevel(line) - endline, endlinenum, endpos = line, linenum, if_else_match.end() - if_match = Search(r'\bif\s*\(', line) - if if_match: - # This could be a multiline if condition, so find the end first. - pos = if_match.end() - 1 - (endline, endlinenum, endpos) = CloseExpression(clean_lines, linenum, pos) - # Check for an opening brace, either directly after the if or on the next - # line. If found, this isn't a single-statement conditional. - if (not Match(r'\s*{', endline[endpos:]) - and not (Match(r'\s*$', endline[endpos:]) - and endlinenum < (len(clean_lines.elided) - 1) - and Match(r'\s*{', clean_lines.elided[endlinenum + 1]))): - while (endlinenum < len(clean_lines.elided) - and ';' not in clean_lines.elided[endlinenum][endpos:]): - endlinenum += 1 - endpos = 0 - if endlinenum < len(clean_lines.elided): - endline = clean_lines.elided[endlinenum] - # We allow a mix of whitespace and closing braces (e.g. for one-liner - # methods) and a single \ after the semicolon (for macros) - endpos = endline.find(';') - if not Match(r';[\s}]*(\\?)$', endline[endpos:]): - # Semicolon isn't the last character, there's something trailing. - # Output a warning if the semicolon is not contained inside - # a lambda expression. - if not Match(r'^[^{};]*\[[^\[\]]*\][^{}]*\{[^{}]*\}\s*\)*[;,]\s*$', - endline): - error(filename, linenum, 'readability/braces', 4, - 'If/else bodies with multiple statements require braces') - elif endlinenum < len(clean_lines.elided) - 1: - # Make sure the next line is dedented - next_line = clean_lines.elided[endlinenum + 1] - next_indent = GetIndentLevel(next_line) - # With ambiguous nested if statements, this will error out on the - # if that *doesn't* match the else, regardless of whether it's the - # inner one or outer one. - if (if_match and Match(r'\s*else\b', next_line) - and next_indent != if_indent): - error(filename, linenum, 'readability/braces', 4, - 'Else clause should be indented at the same level as if. ' - 'Ambiguous nested if/else chains require braces.') - elif next_indent > if_indent: - error(filename, linenum, 'readability/braces', 4, - 'If/else bodies with multiple statements require braces') +def CheckSectionSpacing(filename, clean_lines, class_info, linenum, error): + """Checks for additional blank line issues related to sections. + Currently the only thing checked here is blank line before protected/private. -def CheckTrailingSemicolon(filename, clean_lines, linenum, error): - """Looks for redundant trailing semicolon. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - - line = clean_lines.elided[linenum] - - # Block bodies should not be followed by a semicolon. Due to C++11 - # brace initialization, there are more places where semicolons are - # required than not, so we use an allowed list approach to check these - # rather than an exclusion list. These are the places where "};" should - # be replaced by just "}": - # 1. Some flavor of block following closing parenthesis: - # for (;;) {}; - # while (...) {}; - # switch (...) {}; - # Function(...) {}; - # if (...) {}; - # if (...) else if (...) {}; - # - # 2. else block: - # if (...) else {}; - # - # 3. const member function: - # Function(...) const {}; - # - # 4. Block following some statement: - # x = 42; - # {}; - # - # 5. Block at the beginning of a function: - # Function(...) { - # {}; - # } - # - # Note that naively checking for the preceding "{" will also match - # braces inside multi-dimensional arrays, but this is fine since - # that expression will not contain semicolons. - # - # 6. Block following another block: - # while (true) {} - # {}; - # - # 7. End of namespaces: - # namespace {}; - # - # These semicolons seems far more common than other kinds of - # redundant semicolons, possibly due to people converting classes - # to namespaces. For now we do not warn for this case. - # - # Try matching case 1 first. - match = Match(r'^(.*\)\s*)\{', line) - if match: - # Matched closing parenthesis (case 1). Check the token before the - # matching opening parenthesis, and don't warn if it looks like a - # macro. This avoids these false positives: - # - macro that defines a base class - # - multi-line macro that defines a base class - # - macro that defines the whole class-head + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + class_info: A _ClassInfo objects. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + # Skip checks if the class is small, where small means 25 lines or less. + # 25 lines seems like a good cutoff since that's the usual height of + # terminals, and any class that can't fit in one screen can't really + # be considered "small". # - # But we still issue warnings for macros that we know are safe to - # warn, specifically: - # - TEST, TEST_F, TEST_P, MATCHER, MATCHER_P - # - TYPED_TEST - # - INTERFACE_DEF - # - EXCLUSIVE_LOCKS_REQUIRED, SHARED_LOCKS_REQUIRED, LOCKS_EXCLUDED: + # Also skip checks if we are on the first line. This accounts for + # classes that look like + # class Foo { public: ... }; # - # We implement a list of allowed safe macros instead of a list of - # unsafe macros, even though the latter appears less frequently in - # google code and would have been easier to implement. This is because - # the downside for getting the allowed list wrong means some extra - # semicolons, while the downside for getting the exclusion list wrong - # would result in compile errors. + # If we didn't find the end of the class, last_line would be zero, + # and the check will be skipped by the first condition. + if ( + class_info.last_line - class_info.starting_linenum <= 24 + or linenum <= class_info.starting_linenum + ): + return + + matched = Match(r"\s*(public|protected|private):", clean_lines.lines[linenum]) + if matched: + # Issue warning if the line before public/protected/private was + # not a blank line, but don't do this if the previous line contains + # "class" or "struct". This can happen two ways: + # - We are at the beginning of the class. + # - We are forward-declaring an inner class that is semantically + # private, but needed to be public for implementation reasons. + # Also ignores cases where the previous line ends with a backslash as can be + # common when defining classes in C macros. + prev_line = clean_lines.lines[linenum - 1] + if ( + not IsBlankLine(prev_line) + and not Search(r"\b(class|struct)\b", prev_line) + and not Search(r"\\$", prev_line) + ): + # Try a bit harder to find the beginning of the class. This is to + # account for multi-line base-specifier lists, e.g.: + # class Derived + # : public Base { + end_class_head = class_info.starting_linenum + for i in range(class_info.starting_linenum, linenum): + if Search(r"\{\s*$", clean_lines.lines[i]): + end_class_head = i + break + if end_class_head < linenum - 1: + error( + filename, + linenum, + "whitespace/blank_line", + 3, + '"%s:" should be preceded by a blank line' % matched.group(1), + ) + + +def GetPreviousNonBlankLine(clean_lines, linenum): + """Return the most recent non-blank line and its line number. + + Args: + clean_lines: A CleansedLines instance containing the file contents. + linenum: The number of the line to check. + + Returns: + A tuple with two elements. The first element is the contents of the last + non-blank line before the current line, or the empty string if this is the + first non-blank line. The second is the line number of that line, or -1 + if this is the first non-blank line. + """ + + prevlinenum = linenum - 1 + while prevlinenum >= 0: + prevline = clean_lines.elided[prevlinenum] + if not IsBlankLine(prevline): # if not a blank line... + return (prevline, prevlinenum) + prevlinenum -= 1 + return ("", -1) + + +def CheckBraces(filename, clean_lines, linenum, error): + """Looks for misplaced braces (e.g. at the end of line). + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + + line = clean_lines.elided[linenum] # get rid of comments and strings + + if Match(r"\s*{\s*$", line): + # We allow an open brace to start a line in the case where someone is using + # braces in a block to explicitly create a new scope, which is commonly used + # to control the lifetime of stack-allocated variables. Braces are also + # used for brace initializers inside function calls. We don't detect this + # perfectly: we just don't complain if the last non-whitespace character on + # the previous non-blank line is ',', ';', ':', '(', '{', or '}', or if the + # previous line starts a preprocessor block. We also allow a brace on the + # following line if it is part of an array initialization and would not fit + # within the 80 character limit of the preceding line. + prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0] + if ( + not Search(r"[,;:}{(]\s*$", prevline) + and not Match(r"\s*#", prevline) + and not (GetLineWidth(prevline) > _line_length - 2 and "[]" in prevline) + ): + error( + filename, + linenum, + "whitespace/braces", + 4, + "{ should almost always be at the end of the previous line", + ) + + # An else clause should be on the same line as the preceding closing brace. + if Match(r"\s*else\b\s*(?:if\b|\{|$)", line): + prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0] + if Match(r"\s*}\s*$", prevline): + error( + filename, + linenum, + "whitespace/newline", + 4, + "An else should appear on the same line as the preceding }", + ) + + # If braces come on one side of an else, they should be on both. + # However, we have to worry about "else if" that spans multiple lines! + if Search(r"else if\s*\(", line): # could be multi-line if + brace_on_left = bool(Search(r"}\s*else if\s*\(", line)) + # find the ( after the if + pos = line.find("else if") + pos = line.find("(", pos) + if pos > 0: + (endline, _, endpos) = CloseExpression(clean_lines, linenum, pos) + brace_on_right = endline[endpos:].find("{") != -1 + if brace_on_left != brace_on_right: # must be brace after if + error( + filename, + linenum, + "readability/braces", + 5, + "If an else has a brace on one side, it should have it on both", + ) + elif Search(r"}\s*else[^{]*$", line) or Match(r"[^}]*else\s*{", line): + error( + filename, + linenum, + "readability/braces", + 5, + "If an else has a brace on one side, it should have it on both", + ) + + # Likewise, an else should never have the else clause on the same line + if Search(r"\belse [^\s{]", line) and not Search(r"\belse if\b", line): + error( + filename, + linenum, + "whitespace/newline", + 4, + "Else clause should never be on same line as else (use 2 lines)", + ) + + # In the same way, a do/while should never be on one line + if Match(r"\s*do [^\s{]", line): + error( + filename, + linenum, + "whitespace/newline", + 4, + "do/while clauses should not be on a single line", + ) + + # Check single-line if/else bodies. The style guide says 'curly braces are not + # required for single-line statements'. We additionally allow multi-line, + # single statements, but we reject anything with more than one semicolon in + # it. This means that the first semicolon after the if should be at the end of + # its line, and the line after that should have an indent level equal to or + # lower than the if. We also check for ambiguous if/else nesting without + # braces. + if_else_match = Search(r"\b(if\s*\(|else\b)", line) + if if_else_match and not Match(r"\s*#", line): + if_indent = GetIndentLevel(line) + endline, endlinenum, endpos = line, linenum, if_else_match.end() + if_match = Search(r"\bif\s*\(", line) + if if_match: + # This could be a multiline if condition, so find the end first. + pos = if_match.end() - 1 + (endline, endlinenum, endpos) = CloseExpression(clean_lines, linenum, pos) + # Check for an opening brace, either directly after the if or on the next + # line. If found, this isn't a single-statement conditional. + if not Match(r"\s*{", endline[endpos:]) and not ( + Match(r"\s*$", endline[endpos:]) + and endlinenum < (len(clean_lines.elided) - 1) + and Match(r"\s*{", clean_lines.elided[endlinenum + 1]) + ): + while ( + endlinenum < len(clean_lines.elided) + and ";" not in clean_lines.elided[endlinenum][endpos:] + ): + endlinenum += 1 + endpos = 0 + if endlinenum < len(clean_lines.elided): + endline = clean_lines.elided[endlinenum] + # We allow a mix of whitespace and closing braces (e.g. for one-liner + # methods) and a single \ after the semicolon (for macros) + endpos = endline.find(";") + if not Match(r";[\s}]*(\\?)$", endline[endpos:]): + # Semicolon isn't the last character, there's something trailing. + # Output a warning if the semicolon is not contained inside + # a lambda expression. + if not Match( + r"^[^{};]*\[[^\[\]]*\][^{}]*\{[^{}]*\}\s*\)*[;,]\s*$", endline + ): + error( + filename, + linenum, + "readability/braces", + 4, + "If/else bodies with multiple statements require braces", + ) + elif endlinenum < len(clean_lines.elided) - 1: + # Make sure the next line is dedented + next_line = clean_lines.elided[endlinenum + 1] + next_indent = GetIndentLevel(next_line) + # With ambiguous nested if statements, this will error out on the + # if that *doesn't* match the else, regardless of whether it's the + # inner one or outer one. + if ( + if_match + and Match(r"\s*else\b", next_line) + and next_indent != if_indent + ): + error( + filename, + linenum, + "readability/braces", + 4, + "Else clause should be indented at the same level as if. " + "Ambiguous nested if/else chains require braces.", + ) + elif next_indent > if_indent: + error( + filename, + linenum, + "readability/braces", + 4, + "If/else bodies with multiple statements require braces", + ) + + +def CheckTrailingSemicolon(filename, clean_lines, linenum, error): + """Looks for redundant trailing semicolon. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + + line = clean_lines.elided[linenum] + + # Block bodies should not be followed by a semicolon. Due to C++11 + # brace initialization, there are more places where semicolons are + # required than not, so we use an allowed list approach to check these + # rather than an exclusion list. These are the places where "};" should + # be replaced by just "}": + # 1. Some flavor of block following closing parenthesis: + # for (;;) {}; + # while (...) {}; + # switch (...) {}; + # Function(...) {}; + # if (...) {}; + # if (...) else if (...) {}; # - # In addition to macros, we also don't want to warn on - # - Compound literals - # - Lambdas - # - alignas specifier with anonymous structs - # - decltype - closing_brace_pos = match.group(1).rfind(')') - opening_parenthesis = ReverseCloseExpression( - clean_lines, linenum, closing_brace_pos) - if opening_parenthesis[2] > -1: - line_prefix = opening_parenthesis[0][0:opening_parenthesis[2]] - macro = Search(r'\b([A-Z_][A-Z0-9_]*)\s*$', line_prefix) - func = Match(r'^(.*\])\s*$', line_prefix) - if ((macro and - macro.group(1) not in ( - 'TEST', 'TEST_F', 'MATCHER', 'MATCHER_P', 'TYPED_TEST', - 'EXCLUSIVE_LOCKS_REQUIRED', 'SHARED_LOCKS_REQUIRED', - 'LOCKS_EXCLUDED', 'INTERFACE_DEF')) or - (func and not Search(r'\boperator\s*\[\s*\]', func.group(1))) or - Search(r'\b(?:struct|union)\s+alignas\s*$', line_prefix) or - Search(r'\bdecltype$', line_prefix) or - Search(r'\s+=\s*$', line_prefix)): - match = None - if (match and - opening_parenthesis[1] > 1 and - Search(r'\]\s*$', clean_lines.elided[opening_parenthesis[1] - 1])): - # Multi-line lambda-expression - match = None - - else: - # Try matching cases 2-3. - match = Match(r'^(.*(?:else|\)\s*const)\s*)\{', line) - if not match: - # Try matching cases 4-6. These are always matched on separate lines. - # - # Note that we can't simply concatenate the previous line to the - # current line and do a single match, otherwise we may output - # duplicate warnings for the blank line case: - # if (cond) { - # // blank line - # } - prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0] - if prevline and Search(r'[;{}]\s*$', prevline): - match = Match(r'^(\s*)\{', line) - - # Check matching closing brace - if match: - (endline, endlinenum, endpos) = CloseExpression( - clean_lines, linenum, len(match.group(1))) - if endpos > -1 and Match(r'^\s*;', endline[endpos:]): - # Current {} pair is eligible for semicolon check, and we have found - # the redundant semicolon, output warning here. - # - # Note: because we are scanning forward for opening braces, and - # outputting warnings for the matching closing brace, if there are - # nested blocks with trailing semicolons, we will get the error - # messages in reversed order. - - # We need to check the line forward for NOLINT - raw_lines = clean_lines.raw_lines - ParseNolintSuppressions(filename, raw_lines[endlinenum-1], endlinenum-1, - error) - ParseNolintSuppressions(filename, raw_lines[endlinenum], endlinenum, - error) - - error(filename, endlinenum, 'readability/braces', 4, - "You don't need a ; after a }") + # 2. else block: + # if (...) else {}; + # + # 3. const member function: + # Function(...) const {}; + # + # 4. Block following some statement: + # x = 42; + # {}; + # + # 5. Block at the beginning of a function: + # Function(...) { + # {}; + # } + # + # Note that naively checking for the preceding "{" will also match + # braces inside multi-dimensional arrays, but this is fine since + # that expression will not contain semicolons. + # + # 6. Block following another block: + # while (true) {} + # {}; + # + # 7. End of namespaces: + # namespace {}; + # + # These semicolons seems far more common than other kinds of + # redundant semicolons, possibly due to people converting classes + # to namespaces. For now we do not warn for this case. + # + # Try matching case 1 first. + match = Match(r"^(.*\)\s*)\{", line) + if match: + # Matched closing parenthesis (case 1). Check the token before the + # matching opening parenthesis, and don't warn if it looks like a + # macro. This avoids these false positives: + # - macro that defines a base class + # - multi-line macro that defines a base class + # - macro that defines the whole class-head + # + # But we still issue warnings for macros that we know are safe to + # warn, specifically: + # - TEST, TEST_F, TEST_P, MATCHER, MATCHER_P + # - TYPED_TEST + # - INTERFACE_DEF + # - EXCLUSIVE_LOCKS_REQUIRED, SHARED_LOCKS_REQUIRED, LOCKS_EXCLUDED: + # + # We implement a list of allowed safe macros instead of a list of + # unsafe macros, even though the latter appears less frequently in + # google code and would have been easier to implement. This is because + # the downside for getting the allowed list wrong means some extra + # semicolons, while the downside for getting the exclusion list wrong + # would result in compile errors. + # + # In addition to macros, we also don't want to warn on + # - Compound literals + # - Lambdas + # - alignas specifier with anonymous structs + # - decltype + closing_brace_pos = match.group(1).rfind(")") + opening_parenthesis = ReverseCloseExpression( + clean_lines, linenum, closing_brace_pos + ) + if opening_parenthesis[2] > -1: + line_prefix = opening_parenthesis[0][0 : opening_parenthesis[2]] + macro = Search(r"\b([A-Z_][A-Z0-9_]*)\s*$", line_prefix) + func = Match(r"^(.*\])\s*$", line_prefix) + if ( + ( + macro + and macro.group(1) + not in ( + "TEST", + "TEST_F", + "MATCHER", + "MATCHER_P", + "TYPED_TEST", + "EXCLUSIVE_LOCKS_REQUIRED", + "SHARED_LOCKS_REQUIRED", + "LOCKS_EXCLUDED", + "INTERFACE_DEF", + ) + ) + or (func and not Search(r"\boperator\s*\[\s*\]", func.group(1))) + or Search(r"\b(?:struct|union)\s+alignas\s*$", line_prefix) + or Search(r"\bdecltype$", line_prefix) + or Search(r"\s+=\s*$", line_prefix) + ): + match = None + if ( + match + and opening_parenthesis[1] > 1 + and Search(r"\]\s*$", clean_lines.elided[opening_parenthesis[1] - 1]) + ): + # Multi-line lambda-expression + match = None + + else: + # Try matching cases 2-3. + match = Match(r"^(.*(?:else|\)\s*const)\s*)\{", line) + if not match: + # Try matching cases 4-6. These are always matched on separate lines. + # + # Note that we can't simply concatenate the previous line to the + # current line and do a single match, otherwise we may output + # duplicate warnings for the blank line case: + # if (cond) { + # // blank line + # } + prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0] + if prevline and Search(r"[;{}]\s*$", prevline): + match = Match(r"^(\s*)\{", line) + + # Check matching closing brace + if match: + (endline, endlinenum, endpos) = CloseExpression( + clean_lines, linenum, len(match.group(1)) + ) + if endpos > -1 and Match(r"^\s*;", endline[endpos:]): + # Current {} pair is eligible for semicolon check, and we have found + # the redundant semicolon, output warning here. + # + # Note: because we are scanning forward for opening braces, and + # outputting warnings for the matching closing brace, if there are + # nested blocks with trailing semicolons, we will get the error + # messages in reversed order. + + # We need to check the line forward for NOLINT + raw_lines = clean_lines.raw_lines + ParseNolintSuppressions( + filename, raw_lines[endlinenum - 1], endlinenum - 1, error + ) + ParseNolintSuppressions(filename, raw_lines[endlinenum], endlinenum, error) + + error( + filename, + endlinenum, + "readability/braces", + 4, + "You don't need a ; after a }", + ) def CheckEmptyBlockBody(filename, clean_lines, linenum, error): - """Look for empty loop/conditional body with only a single semicolon. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - - # Search for loop keywords at the beginning of the line. Because only - # whitespaces are allowed before the keywords, this will also ignore most - # do-while-loops, since those lines should start with closing brace. - # - # We also check "if" blocks here, since an empty conditional block - # is likely an error. - line = clean_lines.elided[linenum] - matched = Match(r'\s*(for|while|if)\s*\(', line) - if matched: - # Find the end of the conditional expression. - (end_line, end_linenum, end_pos) = CloseExpression( - clean_lines, linenum, line.find('(')) - - # Output warning if what follows the condition expression is a semicolon. - # No warning for all other cases, including whitespace or newline, since we - # have a separate check for semicolons preceded by whitespace. - if end_pos >= 0 and Match(r';', end_line[end_pos:]): - if matched.group(1) == 'if': - error(filename, end_linenum, 'whitespace/empty_conditional_body', 5, - 'Empty conditional bodies should use {}') - else: - error(filename, end_linenum, 'whitespace/empty_loop_body', 5, - 'Empty loop bodies should use {} or continue') - - # Check for if statements that have completely empty bodies (no comments) - # and no else clauses. - if end_pos >= 0 and matched.group(1) == 'if': - # Find the position of the opening { for the if statement. - # Return without logging an error if it has no brackets. - opening_linenum = end_linenum - opening_line_fragment = end_line[end_pos:] - # Loop until EOF or find anything that's not whitespace or opening {. - while not Search(r'^\s*\{', opening_line_fragment): - if Search(r'^(?!\s*$)', opening_line_fragment): - # Conditional has no brackets. - return - opening_linenum += 1 - if opening_linenum == len(clean_lines.elided): - # Couldn't find conditional's opening { or any code before EOF. - return - opening_line_fragment = clean_lines.elided[opening_linenum] - # Set opening_line (opening_line_fragment may not be entire opening line). - opening_line = clean_lines.elided[opening_linenum] - - # Find the position of the closing }. - opening_pos = opening_line_fragment.find('{') - if opening_linenum == end_linenum: - # We need to make opening_pos relative to the start of the entire line. - opening_pos += end_pos - (closing_line, closing_linenum, closing_pos) = CloseExpression( - clean_lines, opening_linenum, opening_pos) - if closing_pos < 0: - return + """Look for empty loop/conditional body with only a single semicolon. - # Now construct the body of the conditional. This consists of the portion - # of the opening line after the {, all lines until the closing line, - # and the portion of the closing line before the }. - if (clean_lines.raw_lines[opening_linenum] != - CleanseComments(clean_lines.raw_lines[opening_linenum])): - # Opening line ends with a comment, so conditional isn't empty. - return - if closing_linenum > opening_linenum: - # Opening line after the {. Ignore comments here since we checked above. - body = list(opening_line[opening_pos+1:]) - # All lines until closing line, excluding closing line, with comments. - body.extend(clean_lines.raw_lines[opening_linenum+1:closing_linenum]) - # Closing line before the }. Won't (and can't) have comments. - body.append(clean_lines.elided[closing_linenum][:closing_pos-1]) - body = '\n'.join(body) - else: - # If statement has brackets and fits on a single line. - body = opening_line[opening_pos+1:closing_pos-1] - - # Check if the body is empty - if not _EMPTY_CONDITIONAL_BODY_PATTERN.search(body): - return - # The body is empty. Now make sure there's not an else clause. - current_linenum = closing_linenum - current_line_fragment = closing_line[closing_pos:] - # Loop until EOF or find anything that's not whitespace or else clause. - while Search(r'^\s*$|^(?=\s*else)', current_line_fragment): - if Search(r'^(?=\s*else)', current_line_fragment): - # Found an else clause, so don't log an error. - return - current_linenum += 1 - if current_linenum == len(clean_lines.elided): - break - current_line_fragment = clean_lines.elided[current_linenum] - - # The body is empty and there's no else clause until EOF or other code. - error(filename, end_linenum, 'whitespace/empty_if_body', 4, - ('If statement had no body and no else clause')) + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + + # Search for loop keywords at the beginning of the line. Because only + # whitespaces are allowed before the keywords, this will also ignore most + # do-while-loops, since those lines should start with closing brace. + # + # We also check "if" blocks here, since an empty conditional block + # is likely an error. + line = clean_lines.elided[linenum] + matched = Match(r"\s*(for|while|if)\s*\(", line) + if matched: + # Find the end of the conditional expression. + (end_line, end_linenum, end_pos) = CloseExpression( + clean_lines, linenum, line.find("(") + ) + + # Output warning if what follows the condition expression is a semicolon. + # No warning for all other cases, including whitespace or newline, since we + # have a separate check for semicolons preceded by whitespace. + if end_pos >= 0 and Match(r";", end_line[end_pos:]): + if matched.group(1) == "if": + error( + filename, + end_linenum, + "whitespace/empty_conditional_body", + 5, + "Empty conditional bodies should use {}", + ) + else: + error( + filename, + end_linenum, + "whitespace/empty_loop_body", + 5, + "Empty loop bodies should use {} or continue", + ) + + # Check for if statements that have completely empty bodies (no comments) + # and no else clauses. + if end_pos >= 0 and matched.group(1) == "if": + # Find the position of the opening { for the if statement. + # Return without logging an error if it has no brackets. + opening_linenum = end_linenum + opening_line_fragment = end_line[end_pos:] + # Loop until EOF or find anything that's not whitespace or opening {. + while not Search(r"^\s*\{", opening_line_fragment): + if Search(r"^(?!\s*$)", opening_line_fragment): + # Conditional has no brackets. + return + opening_linenum += 1 + if opening_linenum == len(clean_lines.elided): + # Couldn't find conditional's opening { or any code before EOF. + return + opening_line_fragment = clean_lines.elided[opening_linenum] + # Set opening_line (opening_line_fragment may not be entire opening line). + opening_line = clean_lines.elided[opening_linenum] + + # Find the position of the closing }. + opening_pos = opening_line_fragment.find("{") + if opening_linenum == end_linenum: + # We need to make opening_pos relative to the start of the entire line. + opening_pos += end_pos + (closing_line, closing_linenum, closing_pos) = CloseExpression( + clean_lines, opening_linenum, opening_pos + ) + if closing_pos < 0: + return + + # Now construct the body of the conditional. This consists of the portion + # of the opening line after the {, all lines until the closing line, + # and the portion of the closing line before the }. + if clean_lines.raw_lines[opening_linenum] != CleanseComments( + clean_lines.raw_lines[opening_linenum] + ): + # Opening line ends with a comment, so conditional isn't empty. + return + if closing_linenum > opening_linenum: + # Opening line after the {. Ignore comments here since we checked above. + body = list(opening_line[opening_pos + 1 :]) + # All lines until closing line, excluding closing line, with comments. + body.extend( + clean_lines.raw_lines[opening_linenum + 1 : closing_linenum] + ) + # Closing line before the }. Won't (and can't) have comments. + body.append(clean_lines.elided[closing_linenum][: closing_pos - 1]) + body = "\n".join(body) + else: + # If statement has brackets and fits on a single line. + body = opening_line[opening_pos + 1 : closing_pos - 1] + + # Check if the body is empty + if not _EMPTY_CONDITIONAL_BODY_PATTERN.search(body): + return + # The body is empty. Now make sure there's not an else clause. + current_linenum = closing_linenum + current_line_fragment = closing_line[closing_pos:] + # Loop until EOF or find anything that's not whitespace or else clause. + while Search(r"^\s*$|^(?=\s*else)", current_line_fragment): + if Search(r"^(?=\s*else)", current_line_fragment): + # Found an else clause, so don't log an error. + return + current_linenum += 1 + if current_linenum == len(clean_lines.elided): + break + current_line_fragment = clean_lines.elided[current_linenum] + + # The body is empty and there's no else clause until EOF or other code. + error( + filename, + end_linenum, + "whitespace/empty_if_body", + 4, + ("If statement had no body and no else clause"), + ) def FindCheckMacro(line): - """Find a replaceable CHECK-like macro. - - Args: - line: line to search on. - Returns: - (macro name, start position), or (None, -1) if no replaceable - macro is found. - """ - for macro in _CHECK_MACROS: - i = line.find(macro) - if i >= 0: - # Find opening parenthesis. Do a regular expression match here - # to make sure that we are matching the expected CHECK macro, as - # opposed to some other macro that happens to contain the CHECK - # substring. - matched = Match(r'^(.*\b' + macro + r'\s*)\(', line) - if not matched: - continue - return (macro, len(matched.group(1))) - return (None, -1) + """Find a replaceable CHECK-like macro. + + Args: + line: line to search on. + Returns: + (macro name, start position), or (None, -1) if no replaceable + macro is found. + """ + for macro in _CHECK_MACROS: + i = line.find(macro) + if i >= 0: + # Find opening parenthesis. Do a regular expression match here + # to make sure that we are matching the expected CHECK macro, as + # opposed to some other macro that happens to contain the CHECK + # substring. + matched = Match(r"^(.*\b" + macro + r"\s*)\(", line) + if not matched: + continue + return (macro, len(matched.group(1))) + return (None, -1) def CheckCheck(filename, clean_lines, linenum, error): - """Checks the use of CHECK and EXPECT macros. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - - # Decide the set of replacement macros that should be suggested - lines = clean_lines.elided - (check_macro, start_pos) = FindCheckMacro(lines[linenum]) - if not check_macro: - return - - # Find end of the boolean expression by matching parentheses - (last_line, end_line, end_pos) = CloseExpression( - clean_lines, linenum, start_pos) - if end_pos < 0: - return - - # If the check macro is followed by something other than a - # semicolon, assume users will log their own custom error messages - # and don't suggest any replacements. - if not Match(r'\s*;', last_line[end_pos:]): - return - - if linenum == end_line: - expression = lines[linenum][start_pos + 1:end_pos - 1] - else: - expression = lines[linenum][start_pos + 1:] - for i in xrange(linenum + 1, end_line): - expression += lines[i] - expression += last_line[0:end_pos - 1] - - # Parse expression so that we can take parentheses into account. - # This avoids false positives for inputs like "CHECK((a < 4) == b)", - # which is not replaceable by CHECK_LE. - lhs = '' - rhs = '' - operator = None - while expression: - matched = Match(r'^\s*(<<|<<=|>>|>>=|->\*|->|&&|\|\||' - r'==|!=|>=|>|<=|<|\()(.*)$', expression) - if matched: - token = matched.group(1) - if token == '(': - # Parenthesized operand - expression = matched.group(2) - (end, _) = FindEndOfExpressionInLine(expression, 0, ['(']) - if end < 0: - return # Unmatched parenthesis - lhs += '(' + expression[0:end] - expression = expression[end:] - elif token in ('&&', '||'): - # Logical and/or operators. This means the expression - # contains more than one term, for example: - # CHECK(42 < a && a < b); - # - # These are not replaceable with CHECK_LE, so bail out early. + """Checks the use of CHECK and EXPECT macros. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + + # Decide the set of replacement macros that should be suggested + lines = clean_lines.elided + (check_macro, start_pos) = FindCheckMacro(lines[linenum]) + if not check_macro: + return + + # Find end of the boolean expression by matching parentheses + (last_line, end_line, end_pos) = CloseExpression(clean_lines, linenum, start_pos) + if end_pos < 0: + return + + # If the check macro is followed by something other than a + # semicolon, assume users will log their own custom error messages + # and don't suggest any replacements. + if not Match(r"\s*;", last_line[end_pos:]): return - elif token in ('<<', '<<=', '>>', '>>=', '->*', '->'): - # Non-relational operator - lhs += token - expression = matched.group(2) - else: - # Relational operator - operator = token - rhs = matched.group(2) - break + + if linenum == end_line: + expression = lines[linenum][start_pos + 1 : end_pos - 1] else: - # Unparenthesized operand. Instead of appending to lhs one character - # at a time, we do another regular expression match to consume several - # characters at once if possible. Trivial benchmark shows that this - # is more efficient when the operands are longer than a single - # character, which is generally the case. - matched = Match(r'^([^-=!<>()&|]+)(.*)$', expression) - if not matched: - matched = Match(r'^(\s*\S)(.*)$', expression) - if not matched: - break - lhs += matched.group(1) - expression = matched.group(2) - - # Only apply checks if we got all parts of the boolean expression - if not (lhs and operator and rhs): - return - - # Check that rhs do not contain logical operators. We already know - # that lhs is fine since the loop above parses out && and ||. - if rhs.find('&&') > -1 or rhs.find('||') > -1: - return - - # At least one of the operands must be a constant literal. This is - # to avoid suggesting replacements for unprintable things like - # CHECK(variable != iterator) - # - # The following pattern matches decimal, hex integers, strings, and - # characters (in that order). - lhs = lhs.strip() - rhs = rhs.strip() - match_constant = r'^([-+]?(\d+|0[xX][0-9a-fA-F]+)[lLuU]{0,3}|".*"|\'.*\')$' - if Match(match_constant, lhs) or Match(match_constant, rhs): - # Note: since we know both lhs and rhs, we can provide a more - # descriptive error message like: - # Consider using CHECK_EQ(x, 42) instead of CHECK(x == 42) - # Instead of: - # Consider using CHECK_EQ instead of CHECK(a == b) + expression = lines[linenum][start_pos + 1 :] + for i in xrange(linenum + 1, end_line): + expression += lines[i] + expression += last_line[0 : end_pos - 1] + + # Parse expression so that we can take parentheses into account. + # This avoids false positives for inputs like "CHECK((a < 4) == b)", + # which is not replaceable by CHECK_LE. + lhs = "" + rhs = "" + operator = None + while expression: + matched = Match( + r"^\s*(<<|<<=|>>|>>=|->\*|->|&&|\|\||" r"==|!=|>=|>|<=|<|\()(.*)$", + expression, + ) + if matched: + token = matched.group(1) + if token == "(": + # Parenthesized operand + expression = matched.group(2) + (end, _) = FindEndOfExpressionInLine(expression, 0, ["("]) + if end < 0: + return # Unmatched parenthesis + lhs += "(" + expression[0:end] + expression = expression[end:] + elif token in ("&&", "||"): + # Logical and/or operators. This means the expression + # contains more than one term, for example: + # CHECK(42 < a && a < b); + # + # These are not replaceable with CHECK_LE, so bail out early. + return + elif token in ("<<", "<<=", ">>", ">>=", "->*", "->"): + # Non-relational operator + lhs += token + expression = matched.group(2) + else: + # Relational operator + operator = token + rhs = matched.group(2) + break + else: + # Unparenthesized operand. Instead of appending to lhs one character + # at a time, we do another regular expression match to consume several + # characters at once if possible. Trivial benchmark shows that this + # is more efficient when the operands are longer than a single + # character, which is generally the case. + matched = Match(r"^([^-=!<>()&|]+)(.*)$", expression) + if not matched: + matched = Match(r"^(\s*\S)(.*)$", expression) + if not matched: + break + lhs += matched.group(1) + expression = matched.group(2) + + # Only apply checks if we got all parts of the boolean expression + if not (lhs and operator and rhs): + return + + # Check that rhs do not contain logical operators. We already know + # that lhs is fine since the loop above parses out && and ||. + if rhs.find("&&") > -1 or rhs.find("||") > -1: + return + + # At least one of the operands must be a constant literal. This is + # to avoid suggesting replacements for unprintable things like + # CHECK(variable != iterator) # - # We are still keeping the less descriptive message because if lhs - # or rhs gets long, the error message might become unreadable. - error(filename, linenum, 'readability/check', 2, - 'Consider using %s instead of %s(a %s b)' % ( - _CHECK_REPLACEMENT[check_macro][operator], - check_macro, operator)) + # The following pattern matches decimal, hex integers, strings, and + # characters (in that order). + lhs = lhs.strip() + rhs = rhs.strip() + match_constant = r'^([-+]?(\d+|0[xX][0-9a-fA-F]+)[lLuU]{0,3}|".*"|\'.*\')$' + if Match(match_constant, lhs) or Match(match_constant, rhs): + # Note: since we know both lhs and rhs, we can provide a more + # descriptive error message like: + # Consider using CHECK_EQ(x, 42) instead of CHECK(x == 42) + # Instead of: + # Consider using CHECK_EQ instead of CHECK(a == b) + # + # We are still keeping the less descriptive message because if lhs + # or rhs gets long, the error message might become unreadable. + error( + filename, + linenum, + "readability/check", + 2, + "Consider using %s instead of %s(a %s b)" + % (_CHECK_REPLACEMENT[check_macro][operator], check_macro, operator), + ) def CheckAltTokens(filename, clean_lines, linenum, error): - """Check alternative keywords being used in boolean expressions. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - # Avoid preprocessor lines - if Match(r'^\s*#', line): - return - - # Last ditch effort to avoid multi-line comments. This will not help - # if the comment started before the current line or ended after the - # current line, but it catches most of the false positives. At least, - # it provides a way to workaround this warning for people who use - # multi-line comments in preprocessor macros. - # - # TODO(unknown): remove this once cpplint has better support for - # multi-line comments. - if line.find('/*') >= 0 or line.find('*/') >= 0: - return - - for match in _ALT_TOKEN_REPLACEMENT_PATTERN.finditer(line): - error(filename, linenum, 'readability/alt_tokens', 2, - 'Use operator %s instead of %s' % ( - _ALT_TOKEN_REPLACEMENT[match.group(1)], match.group(1))) + """Check alternative keywords being used in boolean expressions. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Avoid preprocessor lines + if Match(r"^\s*#", line): + return + + # Last ditch effort to avoid multi-line comments. This will not help + # if the comment started before the current line or ended after the + # current line, but it catches most of the false positives. At least, + # it provides a way to workaround this warning for people who use + # multi-line comments in preprocessor macros. + # + # TODO(unknown): remove this once cpplint has better support for + # multi-line comments. + if line.find("/*") >= 0 or line.find("*/") >= 0: + return + + for match in _ALT_TOKEN_REPLACEMENT_PATTERN.finditer(line): + error( + filename, + linenum, + "readability/alt_tokens", + 2, + "Use operator %s instead of %s" + % (_ALT_TOKEN_REPLACEMENT[match.group(1)], match.group(1)), + ) def GetLineWidth(line): - """Determines the width of the line in column positions. - - Args: - line: A string, which may be a Unicode string. - - Returns: - The width of the line in column positions, accounting for Unicode - combining characters and wide characters. - """ - if isinstance(line, unicode): - width = 0 - for uc in unicodedata.normalize('NFC', line): - if unicodedata.east_asian_width(uc) in ('W', 'F'): - width += 2 - elif not unicodedata.combining(uc): - # Issue 337 - # https://mail.python.org/pipermail/python-list/2012-August/628809.html - if (sys.version_info.major, sys.version_info.minor) <= (3, 2): - # https://github.com/python/cpython/blob/2.7/Include/unicodeobject.h#L81 - is_wide_build = sysconfig.get_config_var("Py_UNICODE_SIZE") >= 4 - # https://github.com/python/cpython/blob/2.7/Objects/unicodeobject.c#L564 - is_low_surrogate = 0xDC00 <= ord(uc) <= 0xDFFF - if not is_wide_build and is_low_surrogate: - width -= 1 - - width += 1 - return width - else: - return len(line) - - -def CheckStyle(filename, clean_lines, linenum, file_extension, nesting_state, - error): - """Checks rules from the 'C++ style rules' section of cppguide.html. - - Most of these rules are hard to test (naming, comment style), but we - do what we can. In particular we check for 2-space indents, line lengths, - tab usage, spaces inside code, etc. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - file_extension: The extension (without the dot) of the filename. - nesting_state: A NestingState instance which maintains information about - the current stack of nested blocks being parsed. - error: The function to call with any errors found. - """ - - # Don't use "elided" lines here, otherwise we can't check commented lines. - # Don't want to use "raw" either, because we don't want to check inside C++11 - # raw strings, - raw_lines = clean_lines.lines_without_raw_strings - line = raw_lines[linenum] - prev = raw_lines[linenum - 1] if linenum > 0 else '' - - if line.find('\t') != -1: - error(filename, linenum, 'whitespace/tab', 1, - 'Tab found; better to use spaces') - - # One or three blank spaces at the beginning of the line is weird; it's - # hard to reconcile that with 2-space indents. - # NOTE: here are the conditions rob pike used for his tests. Mine aren't - # as sophisticated, but it may be worth becoming so: RLENGTH==initial_spaces - # if(RLENGTH > 20) complain = 0; - # if(match($0, " +(error|private|public|protected):")) complain = 0; - # if(match(prev, "&& *$")) complain = 0; - # if(match(prev, "\\|\\| *$")) complain = 0; - # if(match(prev, "[\",=><] *$")) complain = 0; - # if(match($0, " <<")) complain = 0; - # if(match(prev, " +for \\(")) complain = 0; - # if(prevodd && match(prevprev, " +for \\(")) complain = 0; - scope_or_label_pattern = r'\s*\w+\s*:\s*\\?$' - classinfo = nesting_state.InnermostClass() - initial_spaces = 0 - cleansed_line = clean_lines.elided[linenum] - while initial_spaces < len(line) and line[initial_spaces] == ' ': - initial_spaces += 1 - # There are certain situations we allow one space, notably for - # section labels, and also lines containing multi-line raw strings. - # We also don't check for lines that look like continuation lines - # (of lines ending in double quotes, commas, equals, or angle brackets) - # because the rules for how to indent those are non-trivial. - if (not Search(r'[",=><] *$', prev) and - (initial_spaces == 1 or initial_spaces == 3) and - not Match(scope_or_label_pattern, cleansed_line) and - not (clean_lines.raw_lines[linenum] != line and - Match(r'^\s*""', line))): - error(filename, linenum, 'whitespace/indent', 3, - 'Weird number of spaces at line-start. ' - 'Are you using a 2-space indent?') - - if line and line[-1].isspace(): - error(filename, linenum, 'whitespace/end_of_line', 4, - 'Line ends in whitespace. Consider deleting these extra spaces.') - - # Check if the line is a header guard. - is_header_guard = False - if IsHeaderExtension(file_extension): - cppvar = GetHeaderGuardCPPVariable(filename) - if (line.startswith('#ifndef %s' % cppvar) or - line.startswith('#define %s' % cppvar) or - line.startswith('#endif // %s' % cppvar)): - is_header_guard = True - # #include lines and header guards can be long, since there's no clean way to - # split them. - # - # URLs can be long too. It's possible to split these, but it makes them - # harder to cut&paste. - # - # The "$Id:...$" comment may also get very long without it being the - # developers fault. - if (not line.startswith('#include') and not is_header_guard and - not Match(r'^\s*//.*http(s?)://\S*$', line) and - not Match(r'^\s*//\s*[^\s]*$', line) and - not Match(r'^// \$Id:.*#[0-9]+ \$$', line)): - line_width = GetLineWidth(line) - if line_width > _line_length: - error(filename, linenum, 'whitespace/line_length', 2, - 'Lines should be <= %i characters long' % _line_length) - - if (cleansed_line.count(';') > 1 and - # for loops are allowed two ;'s (and may run over two lines). - cleansed_line.find('for') == -1 and - (GetPreviousNonBlankLine(clean_lines, linenum)[0].find('for') == -1 or - GetPreviousNonBlankLine(clean_lines, linenum)[0].find(';') != -1) and - # It's ok to have many commands in a switch case that fits in 1 line - not ((cleansed_line.find('case ') != -1 or - cleansed_line.find('default:') != -1) and - cleansed_line.find('break;') != -1)): - error(filename, linenum, 'whitespace/newline', 0, - 'More than one command on the same line') - - # Some more style checks - CheckBraces(filename, clean_lines, linenum, error) - CheckTrailingSemicolon(filename, clean_lines, linenum, error) - CheckEmptyBlockBody(filename, clean_lines, linenum, error) - CheckSpacing(filename, clean_lines, linenum, nesting_state, error) - CheckOperatorSpacing(filename, clean_lines, linenum, error) - CheckParenthesisSpacing(filename, clean_lines, linenum, error) - CheckCommaSpacing(filename, clean_lines, linenum, error) - CheckBracesSpacing(filename, clean_lines, linenum, nesting_state, error) - CheckSpacingForFunctionCall(filename, clean_lines, linenum, error) - CheckCheck(filename, clean_lines, linenum, error) - CheckAltTokens(filename, clean_lines, linenum, error) - classinfo = nesting_state.InnermostClass() - if classinfo: - CheckSectionSpacing(filename, clean_lines, classinfo, linenum, error) + """Determines the width of the line in column positions. + + Args: + line: A string, which may be a Unicode string. + + Returns: + The width of the line in column positions, accounting for Unicode + combining characters and wide characters. + """ + if isinstance(line, unicode): + width = 0 + for uc in unicodedata.normalize("NFC", line): + if unicodedata.east_asian_width(uc) in ("W", "F"): + width += 2 + elif not unicodedata.combining(uc): + # Issue 337 + # https://mail.python.org/pipermail/python-list/2012-August/628809.html + if (sys.version_info.major, sys.version_info.minor) <= (3, 2): + # https://github.com/python/cpython/blob/2.7/Include/unicodeobject.h#L81 + is_wide_build = sysconfig.get_config_var("Py_UNICODE_SIZE") >= 4 + # https://github.com/python/cpython/blob/2.7/Objects/unicodeobject.c#L564 + is_low_surrogate = 0xDC00 <= ord(uc) <= 0xDFFF + if not is_wide_build and is_low_surrogate: + width -= 1 + + width += 1 + return width + else: + return len(line) + + +def CheckStyle(filename, clean_lines, linenum, file_extension, nesting_state, error): + """Checks rules from the 'C++ style rules' section of cppguide.html. + + Most of these rules are hard to test (naming, comment style), but we + do what we can. In particular we check for 2-space indents, line lengths, + tab usage, spaces inside code, etc. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + file_extension: The extension (without the dot) of the filename. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: The function to call with any errors found. + """ + + # Don't use "elided" lines here, otherwise we can't check commented lines. + # Don't want to use "raw" either, because we don't want to check inside C++11 + # raw strings, + raw_lines = clean_lines.lines_without_raw_strings + line = raw_lines[linenum] + prev = raw_lines[linenum - 1] if linenum > 0 else "" + + if line.find("\t") != -1: + error(filename, linenum, "whitespace/tab", 1, "Tab found; better to use spaces") + + # One or three blank spaces at the beginning of the line is weird; it's + # hard to reconcile that with 2-space indents. + # NOTE: here are the conditions rob pike used for his tests. Mine aren't + # as sophisticated, but it may be worth becoming so: RLENGTH==initial_spaces + # if(RLENGTH > 20) complain = 0; + # if(match($0, " +(error|private|public|protected):")) complain = 0; + # if(match(prev, "&& *$")) complain = 0; + # if(match(prev, "\\|\\| *$")) complain = 0; + # if(match(prev, "[\",=><] *$")) complain = 0; + # if(match($0, " <<")) complain = 0; + # if(match(prev, " +for \\(")) complain = 0; + # if(prevodd && match(prevprev, " +for \\(")) complain = 0; + scope_or_label_pattern = r"\s*\w+\s*:\s*\\?$" + classinfo = nesting_state.InnermostClass() + initial_spaces = 0 + cleansed_line = clean_lines.elided[linenum] + while initial_spaces < len(line) and line[initial_spaces] == " ": + initial_spaces += 1 + # There are certain situations we allow one space, notably for + # section labels, and also lines containing multi-line raw strings. + # We also don't check for lines that look like continuation lines + # (of lines ending in double quotes, commas, equals, or angle brackets) + # because the rules for how to indent those are non-trivial. + if ( + not Search(r'[",=><] *$', prev) + and (initial_spaces == 1 or initial_spaces == 3) + and not Match(scope_or_label_pattern, cleansed_line) + and not (clean_lines.raw_lines[linenum] != line and Match(r'^\s*""', line)) + ): + error( + filename, + linenum, + "whitespace/indent", + 3, + "Weird number of spaces at line-start. " "Are you using a 2-space indent?", + ) + + if line and line[-1].isspace(): + error( + filename, + linenum, + "whitespace/end_of_line", + 4, + "Line ends in whitespace. Consider deleting these extra spaces.", + ) + + # Check if the line is a header guard. + is_header_guard = False + if IsHeaderExtension(file_extension): + cppvar = GetHeaderGuardCPPVariable(filename) + if ( + line.startswith("#ifndef %s" % cppvar) + or line.startswith("#define %s" % cppvar) + or line.startswith("#endif // %s" % cppvar) + ): + is_header_guard = True + # #include lines and header guards can be long, since there's no clean way to + # split them. + # + # URLs can be long too. It's possible to split these, but it makes them + # harder to cut&paste. + # + # The "$Id:...$" comment may also get very long without it being the + # developers fault. + if ( + not line.startswith("#include") + and not is_header_guard + and not Match(r"^\s*//.*http(s?)://\S*$", line) + and not Match(r"^\s*//\s*[^\s]*$", line) + and not Match(r"^// \$Id:.*#[0-9]+ \$$", line) + ): + line_width = GetLineWidth(line) + if line_width > _line_length: + error( + filename, + linenum, + "whitespace/line_length", + 2, + "Lines should be <= %i characters long" % _line_length, + ) + + if ( + cleansed_line.count(";") > 1 + and + # for loops are allowed two ;'s (and may run over two lines). + cleansed_line.find("for") == -1 + and ( + GetPreviousNonBlankLine(clean_lines, linenum)[0].find("for") == -1 + or GetPreviousNonBlankLine(clean_lines, linenum)[0].find(";") != -1 + ) + and + # It's ok to have many commands in a switch case that fits in 1 line + not ( + (cleansed_line.find("case ") != -1 or cleansed_line.find("default:") != -1) + and cleansed_line.find("break;") != -1 + ) + ): + error( + filename, + linenum, + "whitespace/newline", + 0, + "More than one command on the same line", + ) + + # Some more style checks + CheckBraces(filename, clean_lines, linenum, error) + CheckTrailingSemicolon(filename, clean_lines, linenum, error) + CheckEmptyBlockBody(filename, clean_lines, linenum, error) + CheckSpacing(filename, clean_lines, linenum, nesting_state, error) + CheckOperatorSpacing(filename, clean_lines, linenum, error) + CheckParenthesisSpacing(filename, clean_lines, linenum, error) + CheckCommaSpacing(filename, clean_lines, linenum, error) + CheckBracesSpacing(filename, clean_lines, linenum, nesting_state, error) + CheckSpacingForFunctionCall(filename, clean_lines, linenum, error) + CheckCheck(filename, clean_lines, linenum, error) + CheckAltTokens(filename, clean_lines, linenum, error) + classinfo = nesting_state.InnermostClass() + if classinfo: + CheckSectionSpacing(filename, clean_lines, classinfo, linenum, error) _RE_PATTERN_INCLUDE = re.compile(r'^\s*#\s*include\s*([<"])([^>"]*)[>"].*$') @@ -4433,224 +4997,262 @@ def CheckStyle(filename, clean_lines, linenum, file_extension, nesting_state, # _RE_FIRST_COMPONENT.match('foo.cc').group(0) == 'foo' # _RE_FIRST_COMPONENT.match('foo-bar_baz.cc').group(0) == 'foo' # _RE_FIRST_COMPONENT.match('foo_bar-baz.cc').group(0) == 'foo' -_RE_FIRST_COMPONENT = re.compile(r'^[^-_.]+') +_RE_FIRST_COMPONENT = re.compile(r"^[^-_.]+") def _DropCommonSuffixes(filename): - """Drops common suffixes like _test.cc or -inl.h from filename. - - For example: - >>> _DropCommonSuffixes('foo/foo-inl.h') - 'foo/foo' - >>> _DropCommonSuffixes('foo/bar/foo.cc') - 'foo/bar/foo' - >>> _DropCommonSuffixes('foo/foo_internal.h') - 'foo/foo' - >>> _DropCommonSuffixes('foo/foo_unusualinternal.h') - 'foo/foo_unusualinternal' - - Args: - filename: The input filename. - - Returns: - The filename with the common suffix removed. - """ - for suffix in ('test.cc', 'regtest.cc', 'unittest.cc', - 'inl.h', 'impl.h', 'internal.h'): - if (filename.endswith(suffix) and len(filename) > len(suffix) and - filename[-len(suffix) - 1] in ('-', '_')): - return filename[:-len(suffix) - 1] - return os.path.splitext(filename)[0] + """Drops common suffixes like _test.cc or -inl.h from filename. + + For example: + >>> _DropCommonSuffixes('foo/foo-inl.h') + 'foo/foo' + >>> _DropCommonSuffixes('foo/bar/foo.cc') + 'foo/bar/foo' + >>> _DropCommonSuffixes('foo/foo_internal.h') + 'foo/foo' + >>> _DropCommonSuffixes('foo/foo_unusualinternal.h') + 'foo/foo_unusualinternal' + + Args: + filename: The input filename. + + Returns: + The filename with the common suffix removed. + """ + for suffix in ( + "test.cc", + "regtest.cc", + "unittest.cc", + "inl.h", + "impl.h", + "internal.h", + ): + if ( + filename.endswith(suffix) + and len(filename) > len(suffix) + and filename[-len(suffix) - 1] in ("-", "_") + ): + return filename[: -len(suffix) - 1] + return os.path.splitext(filename)[0] def _ClassifyInclude(fileinfo, include, is_system): - """Figures out what kind of header 'include' is. - - Args: - fileinfo: The current file cpplint is running over. A FileInfo instance. - include: The path to a #included file. - is_system: True if the #include used <> rather than "". - - Returns: - One of the _XXX_HEADER constants. - - For example: - >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'stdio.h', True) - _C_SYS_HEADER - >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'string', True) - _CPP_SYS_HEADER - >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'foo/foo.h', False) - _LIKELY_MY_HEADER - >>> _ClassifyInclude(FileInfo('foo/foo_unknown_extension.cc'), - ... 'bar/foo_other_ext.h', False) - _POSSIBLE_MY_HEADER - >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'foo/bar.h', False) - _OTHER_HEADER - """ - # This is a list of all standard c++ header files, except - # those already checked for above. - is_cpp_h = include in _CPP_HEADERS - - if is_system: - if is_cpp_h: - return _CPP_SYS_HEADER - else: - return _C_SYS_HEADER - - # If the target file and the include we're checking share a - # basename when we drop common extensions, and the include - # lives in . , then it's likely to be owned by the target file. - target_dir, target_base = ( - os.path.split(_DropCommonSuffixes(fileinfo.RepositoryName()))) - include_dir, include_base = os.path.split(_DropCommonSuffixes(include)) - if target_base == include_base and ( - include_dir == target_dir or - include_dir == os.path.normpath(target_dir + '/../public')): - return _LIKELY_MY_HEADER - - # If the target and include share some initial basename - # component, it's possible the target is implementing the - # include, so it's allowed to be first, but we'll never - # complain if it's not there. - target_first_component = _RE_FIRST_COMPONENT.match(target_base) - include_first_component = _RE_FIRST_COMPONENT.match(include_base) - if (target_first_component and include_first_component and - target_first_component.group(0) == - include_first_component.group(0)): - return _POSSIBLE_MY_HEADER - - return _OTHER_HEADER + """Figures out what kind of header 'include' is. + + Args: + fileinfo: The current file cpplint is running over. A FileInfo instance. + include: The path to a #included file. + is_system: True if the #include used <> rather than "". + Returns: + One of the _XXX_HEADER constants. + + For example: + >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'stdio.h', True) + _C_SYS_HEADER + >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'string', True) + _CPP_SYS_HEADER + >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'foo/foo.h', False) + _LIKELY_MY_HEADER + >>> _ClassifyInclude(FileInfo('foo/foo_unknown_extension.cc'), + ... 'bar/foo_other_ext.h', False) + _POSSIBLE_MY_HEADER + >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'foo/bar.h', False) + _OTHER_HEADER + """ + # This is a list of all standard c++ header files, except + # those already checked for above. + is_cpp_h = include in _CPP_HEADERS + + if is_system: + if is_cpp_h: + return _CPP_SYS_HEADER + else: + return _C_SYS_HEADER + + # If the target file and the include we're checking share a + # basename when we drop common extensions, and the include + # lives in . , then it's likely to be owned by the target file. + target_dir, target_base = os.path.split( + _DropCommonSuffixes(fileinfo.RepositoryName()) + ) + include_dir, include_base = os.path.split(_DropCommonSuffixes(include)) + if target_base == include_base and ( + include_dir == target_dir + or include_dir == os.path.normpath(target_dir + "/../public") + ): + return _LIKELY_MY_HEADER + + # If the target and include share some initial basename + # component, it's possible the target is implementing the + # include, so it's allowed to be first, but we'll never + # complain if it's not there. + target_first_component = _RE_FIRST_COMPONENT.match(target_base) + include_first_component = _RE_FIRST_COMPONENT.match(include_base) + if ( + target_first_component + and include_first_component + and target_first_component.group(0) == include_first_component.group(0) + ): + return _POSSIBLE_MY_HEADER + + return _OTHER_HEADER def CheckIncludeLine(filename, clean_lines, linenum, include_state, error): - """Check rules that are applicable to #include lines. - - Strings on #include lines are NOT removed from elided line, to make - certain tasks easier. However, to prevent false positives, checks - applicable to #include lines in CheckLanguage must be put here. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - include_state: An _IncludeState instance in which the headers are inserted. - error: The function to call with any errors found. - """ - fileinfo = FileInfo(filename) - line = clean_lines.lines[linenum] - - # "include" should use the new style "foo/bar.h" instead of just "bar.h" - # Only do this check if the included header follows google naming - # conventions. If not, assume that it's a 3rd party API that - # requires special include conventions. - # - # We also make an exception for Lua headers, which follow google - # naming convention but not the include convention. - match = Match(r'#include\s*"([^/]+\.h)"', line) - if match and not _THIRD_PARTY_HEADERS_PATTERN.match(match.group(1)): - error(filename, linenum, 'build/include', 4, - 'Include the directory when naming .h files') - - # we shouldn't include a file more than once. actually, there are a - # handful of instances where doing so is okay, but in general it's - # not. - match = _RE_PATTERN_INCLUDE.search(line) - if match: - include = match.group(2) - is_system = (match.group(1) == '<') - duplicate_line = include_state.FindHeader(include) - if duplicate_line >= 0: - error(filename, linenum, 'build/include', 4, - '"%s" already included at %s:%s' % - (include, filename, duplicate_line)) - elif (include.endswith('.cc') and - os.path.dirname(fileinfo.RepositoryName()) != os.path.dirname(include)): - error(filename, linenum, 'build/include', 4, - 'Do not include .cc files from other packages') - elif not _THIRD_PARTY_HEADERS_PATTERN.match(include): - include_state.include_list[-1].append((include, linenum)) - - # We want to ensure that headers appear in the right order: - # 1) for foo.cc, foo.h (preferred location) - # 2) c system files - # 3) cpp system files - # 4) for foo.cc, foo.h (deprecated location) - # 5) other google headers - # - # We classify each include statement as one of those 5 types - # using a number of techniques. The include_state object keeps - # track of the highest type seen, and complains if we see a - # lower type after that. - error_message = include_state.CheckNextIncludeOrder( - _ClassifyInclude(fileinfo, include, is_system)) - if error_message: - error(filename, linenum, 'build/include_order', 4, - '%s. Should be: %s.h, c system, c++ system, other.' % - (error_message, fileinfo.BaseName())) - canonical_include = include_state.CanonicalizeAlphabeticalOrder(include) - if not include_state.IsInAlphabeticalOrder( - clean_lines, linenum, canonical_include): - error(filename, linenum, 'build/include_alpha', 4, - 'Include "%s" not in alphabetical order' % include) - include_state.SetLastHeader(canonical_include) + """Check rules that are applicable to #include lines. + + Strings on #include lines are NOT removed from elided line, to make + certain tasks easier. However, to prevent false positives, checks + applicable to #include lines in CheckLanguage must be put here. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + include_state: An _IncludeState instance in which the headers are inserted. + error: The function to call with any errors found. + """ + fileinfo = FileInfo(filename) + line = clean_lines.lines[linenum] + # "include" should use the new style "foo/bar.h" instead of just "bar.h" + # Only do this check if the included header follows google naming + # conventions. If not, assume that it's a 3rd party API that + # requires special include conventions. + # + # We also make an exception for Lua headers, which follow google + # naming convention but not the include convention. + match = Match(r'#include\s*"([^/]+\.h)"', line) + if match and not _THIRD_PARTY_HEADERS_PATTERN.match(match.group(1)): + error( + filename, + linenum, + "build/include", + 4, + "Include the directory when naming .h files", + ) + + # we shouldn't include a file more than once. actually, there are a + # handful of instances where doing so is okay, but in general it's + # not. + match = _RE_PATTERN_INCLUDE.search(line) + if match: + include = match.group(2) + is_system = match.group(1) == "<" + duplicate_line = include_state.FindHeader(include) + if duplicate_line >= 0: + error( + filename, + linenum, + "build/include", + 4, + '"%s" already included at %s:%s' % (include, filename, duplicate_line), + ) + elif include.endswith(".cc") and os.path.dirname( + fileinfo.RepositoryName() + ) != os.path.dirname(include): + error( + filename, + linenum, + "build/include", + 4, + "Do not include .cc files from other packages", + ) + elif not _THIRD_PARTY_HEADERS_PATTERN.match(include): + include_state.include_list[-1].append((include, linenum)) + + # We want to ensure that headers appear in the right order: + # 1) for foo.cc, foo.h (preferred location) + # 2) c system files + # 3) cpp system files + # 4) for foo.cc, foo.h (deprecated location) + # 5) other google headers + # + # We classify each include statement as one of those 5 types + # using a number of techniques. The include_state object keeps + # track of the highest type seen, and complains if we see a + # lower type after that. + error_message = include_state.CheckNextIncludeOrder( + _ClassifyInclude(fileinfo, include, is_system) + ) + if error_message: + error( + filename, + linenum, + "build/include_order", + 4, + "%s. Should be: %s.h, c system, c++ system, other." + % (error_message, fileinfo.BaseName()), + ) + canonical_include = include_state.CanonicalizeAlphabeticalOrder(include) + if not include_state.IsInAlphabeticalOrder( + clean_lines, linenum, canonical_include + ): + error( + filename, + linenum, + "build/include_alpha", + 4, + 'Include "%s" not in alphabetical order' % include, + ) + include_state.SetLastHeader(canonical_include) def _GetTextInside(text, start_pattern): - r"""Retrieves all the text between matching open and close parentheses. - - Given a string of lines and a regular expression string, retrieve all the text - following the expression and between opening punctuation symbols like - (, [, or {, and the matching close-punctuation symbol. This properly nested - occurrences of the punctuations, so for the text like - printf(a(), b(c())); - a call to _GetTextInside(text, r'printf\(') will return 'a(), b(c())'. - start_pattern must match string having an open punctuation symbol at the end. - - Args: - text: The lines to extract text. Its comments and strings must be elided. - It can be single line and can span multiple lines. - start_pattern: The regexp string indicating where to start extracting - the text. - Returns: - The extracted text. - None if either the opening string or ending punctuation could not be found. - """ - # TODO(unknown): Audit cpplint.py to see what places could be profitably - # rewritten to use _GetTextInside (and use inferior regexp matching today). - - # Give opening punctuations to get the matching close-punctuations. - matching_punctuation = {'(': ')', '{': '}', '[': ']'} - closing_punctuation = set(matching_punctuation.itervalues()) - - # Find the position to start extracting text. - match = re.search(start_pattern, text, re.M) - if not match: # start_pattern not found in text. - return None - start_position = match.end(0) - - assert start_position > 0, ( - 'start_pattern must ends with an opening punctuation.') - assert text[start_position - 1] in matching_punctuation, ( - 'start_pattern must ends with an opening punctuation.') - # Stack of closing punctuations we expect to have in text after position. - punctuation_stack = [matching_punctuation[text[start_position - 1]]] - position = start_position - while punctuation_stack and position < len(text): - if text[position] == punctuation_stack[-1]: - punctuation_stack.pop() - elif text[position] in closing_punctuation: - # A closing punctuation without matching opening punctuations. - return None - elif text[position] in matching_punctuation: - punctuation_stack.append(matching_punctuation[text[position]]) - position += 1 - if punctuation_stack: - # Opening punctuations left without matching close-punctuations. - return None - # punctuations match. - return text[start_position:position - 1] + r"""Retrieves all the text between matching open and close parentheses. + + Given a string of lines and a regular expression string, retrieve all the text + following the expression and between opening punctuation symbols like + (, [, or {, and the matching close-punctuation symbol. This properly nested + occurrences of the punctuations, so for the text like + printf(a(), b(c())); + a call to _GetTextInside(text, r'printf\(') will return 'a(), b(c())'. + start_pattern must match string having an open punctuation symbol at the end. + + Args: + text: The lines to extract text. Its comments and strings must be elided. + It can be single line and can span multiple lines. + start_pattern: The regexp string indicating where to start extracting + the text. + Returns: + The extracted text. + None if either the opening string or ending punctuation could not be found. + """ + # TODO(unknown): Audit cpplint.py to see what places could be profitably + # rewritten to use _GetTextInside (and use inferior regexp matching today). + + # Give opening punctuations to get the matching close-punctuations. + matching_punctuation = {"(": ")", "{": "}", "[": "]"} + closing_punctuation = set(matching_punctuation.itervalues()) + + # Find the position to start extracting text. + match = re.search(start_pattern, text, re.M) + if not match: # start_pattern not found in text. + return None + start_position = match.end(0) + + assert start_position > 0, "start_pattern must ends with an opening punctuation." + assert ( + text[start_position - 1] in matching_punctuation + ), "start_pattern must ends with an opening punctuation." + # Stack of closing punctuations we expect to have in text after position. + punctuation_stack = [matching_punctuation[text[start_position - 1]]] + position = start_position + while punctuation_stack and position < len(text): + if text[position] == punctuation_stack[-1]: + punctuation_stack.pop() + elif text[position] in closing_punctuation: + # A closing punctuation without matching opening punctuations. + return None + elif text[position] in matching_punctuation: + punctuation_stack.append(matching_punctuation[text[position]]) + position += 1 + if punctuation_stack: + # Opening punctuations left without matching close-punctuations. + return None + # punctuations match. + return text[start_position : position - 1] # Patterns for matching call-by-reference parameters. @@ -4662,1583 +5264,1925 @@ def _GetTextInside(text, start_pattern): # > # | [^<>] )* # > -_RE_PATTERN_IDENT = r'[_a-zA-Z]\w*' # =~ [[:alpha:]][[:alnum:]]* +_RE_PATTERN_IDENT = r"[_a-zA-Z]\w*" # =~ [[:alpha:]][[:alnum:]]* _RE_PATTERN_TYPE = ( - r'(?:const\s+)?(?:typename\s+|class\s+|struct\s+|union\s+|enum\s+)?' - r'(?:\w|' - r'\s*<(?:<(?:<[^<>]*>|[^<>])*>|[^<>])*>|' - r'::)+') + r"(?:const\s+)?(?:typename\s+|class\s+|struct\s+|union\s+|enum\s+)?" + r"(?:\w|" + r"\s*<(?:<(?:<[^<>]*>|[^<>])*>|[^<>])*>|" + r"::)+" +) # A call-by-reference parameter ends with '& identifier'. _RE_PATTERN_REF_PARAM = re.compile( - r'(' + _RE_PATTERN_TYPE + r'(?:\s*(?:\bconst\b|[*]))*\s*' - r'&\s*' + _RE_PATTERN_IDENT + r')\s*(?:=[^,()]+)?[,)]') + r"(" + _RE_PATTERN_TYPE + r"(?:\s*(?:\bconst\b|[*]))*\s*" + r"&\s*" + _RE_PATTERN_IDENT + r")\s*(?:=[^,()]+)?[,)]" +) # A call-by-const-reference parameter either ends with 'const& identifier' # or looks like 'const type& identifier' when 'type' is atomic. _RE_PATTERN_CONST_REF_PARAM = ( - r'(?:.*\s*\bconst\s*&\s*' + _RE_PATTERN_IDENT + - r'|const\s+' + _RE_PATTERN_TYPE + r'\s*&\s*' + _RE_PATTERN_IDENT + r')') + r"(?:.*\s*\bconst\s*&\s*" + + _RE_PATTERN_IDENT + + r"|const\s+" + + _RE_PATTERN_TYPE + + r"\s*&\s*" + + _RE_PATTERN_IDENT + + r")" +) # Stream types. -_RE_PATTERN_REF_STREAM_PARAM = ( - r'(?:.*stream\s*&\s*' + _RE_PATTERN_IDENT + r')') - - -def CheckLanguage(filename, clean_lines, linenum, file_extension, - include_state, nesting_state, error): - """Checks rules from the 'C++ language rules' section of cppguide.html. - - Some of these rules are hard to test (function overloading, using - uint32 inappropriately), but we do the best we can. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - file_extension: The extension (without the dot) of the filename. - include_state: An _IncludeState instance in which the headers are inserted. - nesting_state: A NestingState instance which maintains information about - the current stack of nested blocks being parsed. - error: The function to call with any errors found. - """ - # If the line is empty or consists of entirely a comment, no need to - # check it. - line = clean_lines.elided[linenum] - if not line: - return - - match = _RE_PATTERN_INCLUDE.search(line) - if match: - CheckIncludeLine(filename, clean_lines, linenum, include_state, error) - return - - # Reset include state across preprocessor directives. This is meant - # to silence warnings for conditional includes. - match = Match(r'^\s*#\s*(if|ifdef|ifndef|elif|else|endif)\b', line) - if match: - include_state.ResetSection(match.group(1)) - - # Make Windows paths like Unix. - fullname = os.path.abspath(filename).replace('\\', '/') - - # Perform other checks now that we are sure that this is not an include line - CheckCasts(filename, clean_lines, linenum, error) - CheckGlobalStatic(filename, clean_lines, linenum, error) - CheckPrintf(filename, clean_lines, linenum, error) - - if IsHeaderExtension(file_extension): - # TODO(unknown): check that 1-arg constructors are explicit. - # How to tell it's a constructor? - # (handled in CheckForNonStandardConstructs for now) - # TODO(unknown): check that classes declare or disable copy/assign - # (level 1 error) - pass +_RE_PATTERN_REF_STREAM_PARAM = r"(?:.*stream\s*&\s*" + _RE_PATTERN_IDENT + r")" + + +def CheckLanguage( + filename, clean_lines, linenum, file_extension, include_state, nesting_state, error +): + """Checks rules from the 'C++ language rules' section of cppguide.html. + + Some of these rules are hard to test (function overloading, using + uint32 inappropriately), but we do the best we can. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + file_extension: The extension (without the dot) of the filename. + include_state: An _IncludeState instance in which the headers are inserted. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: The function to call with any errors found. + """ + # If the line is empty or consists of entirely a comment, no need to + # check it. + line = clean_lines.elided[linenum] + if not line: + return - # Check if people are using the verboten C basic types. The only exception - # we regularly allow is "unsigned short port" for port. - if Search(r'\bshort port\b', line): - if not Search(r'\bunsigned short port\b', line): - error(filename, linenum, 'runtime/int', 4, - 'Use "unsigned short" for ports, not "short"') - else: - match = Search(r'\b(short|long(?! +double)|long long)\b', line) + match = _RE_PATTERN_INCLUDE.search(line) if match: - error(filename, linenum, 'runtime/int', 4, - 'Use int16/int64/etc, rather than the C type %s' % match.group(1)) - - # Check if some verboten operator overloading is going on - # TODO(unknown): catch out-of-line unary operator&: - # class X {}; - # int operator&(const X& x) { return 42; } // unary operator& - # The trick is it's hard to tell apart from binary operator&: - # class Y { int operator&(const Y& x) { return 23; } }; // binary operator& - if Search(r'\boperator\s*&\s*\(\s*\)', line): - error(filename, linenum, 'runtime/operator', 4, - 'Unary operator& is dangerous. Do not use it.') - - # Check for suspicious usage of "if" like - # } if (a == b) { - if Search(r'\}\s*if\s*\(', line): - error(filename, linenum, 'readability/braces', 4, - 'Did you mean "else if"? If not, start a new line for "if".') - - # Check for potential format string bugs like printf(foo). - # We constrain the pattern not to pick things like DocidForPrintf(foo). - # Not perfect but it can catch printf(foo.c_str()) and printf(foo->c_str()) - # TODO(unknown): Catch the following case. Need to change the calling - # convention of the whole function to process multiple line to handle it. - # printf( - # boy_this_is_a_really_long_variable_that_cannot_fit_on_the_prev_line); - printf_args = _GetTextInside(line, r'(?i)\b(string)?printf\s*\(') - if printf_args: - match = Match(r'([\w.\->()]+)$', printf_args) - if match and match.group(1) != '__VA_ARGS__': - function_name = re.search(r'\b((?:string)?printf)\s*\(', - line, re.I).group(1) - error(filename, linenum, 'runtime/printf', 4, - 'Potential format string bug. Do %s("%%s", %s) instead.' - % (function_name, match.group(1))) - - # Check for potential memset bugs like memset(buf, sizeof(buf), 0). - match = Search(r'memset\s*\(([^,]*),\s*([^,]*),\s*0\s*\)', line) - if match and not Match(r"^''|-?[0-9]+|0x[0-9A-Fa-f]$", match.group(2)): - error(filename, linenum, 'runtime/memset', 4, - 'Did you mean "memset(%s, 0, %s)"?' - % (match.group(1), match.group(2))) - - if Search(r'\busing namespace\b', line): - error(filename, linenum, 'build/namespaces', 5, - 'Do not use namespace using-directives. ' - 'Use using-declarations instead.') - - # Detect variable-length arrays. - match = Match(r'\s*(.+::)?(\w+) [a-z]\w*\[(.+)];', line) - if (match and match.group(2) != 'return' and match.group(2) != 'delete' and - match.group(3).find(']') == -1): - # Split the size using space and arithmetic operators as delimiters. - # If any of the resulting tokens are not compile time constants then - # report the error. - tokens = re.split(r'\s|\+|\-|\*|\/|<<|>>]', match.group(3)) - is_const = True - skip_next = False - for tok in tokens: - if skip_next: + CheckIncludeLine(filename, clean_lines, linenum, include_state, error) + return + + # Reset include state across preprocessor directives. This is meant + # to silence warnings for conditional includes. + match = Match(r"^\s*#\s*(if|ifdef|ifndef|elif|else|endif)\b", line) + if match: + include_state.ResetSection(match.group(1)) + + # Make Windows paths like Unix. + fullname = os.path.abspath(filename).replace("\\", "/") + + # Perform other checks now that we are sure that this is not an include line + CheckCasts(filename, clean_lines, linenum, error) + CheckGlobalStatic(filename, clean_lines, linenum, error) + CheckPrintf(filename, clean_lines, linenum, error) + + if IsHeaderExtension(file_extension): + # TODO(unknown): check that 1-arg constructors are explicit. + # How to tell it's a constructor? + # (handled in CheckForNonStandardConstructs for now) + # TODO(unknown): check that classes declare or disable copy/assign + # (level 1 error) + pass + + # Check if people are using the verboten C basic types. The only exception + # we regularly allow is "unsigned short port" for port. + if Search(r"\bshort port\b", line): + if not Search(r"\bunsigned short port\b", line): + error( + filename, + linenum, + "runtime/int", + 4, + 'Use "unsigned short" for ports, not "short"', + ) + else: + match = Search(r"\b(short|long(?! +double)|long long)\b", line) + if match: + error( + filename, + linenum, + "runtime/int", + 4, + "Use int16/int64/etc, rather than the C type %s" % match.group(1), + ) + + # Check if some verboten operator overloading is going on + # TODO(unknown): catch out-of-line unary operator&: + # class X {}; + # int operator&(const X& x) { return 42; } // unary operator& + # The trick is it's hard to tell apart from binary operator&: + # class Y { int operator&(const Y& x) { return 23; } }; // binary operator& + if Search(r"\boperator\s*&\s*\(\s*\)", line): + error( + filename, + linenum, + "runtime/operator", + 4, + "Unary operator& is dangerous. Do not use it.", + ) + + # Check for suspicious usage of "if" like + # } if (a == b) { + if Search(r"\}\s*if\s*\(", line): + error( + filename, + linenum, + "readability/braces", + 4, + 'Did you mean "else if"? If not, start a new line for "if".', + ) + + # Check for potential format string bugs like printf(foo). + # We constrain the pattern not to pick things like DocidForPrintf(foo). + # Not perfect but it can catch printf(foo.c_str()) and printf(foo->c_str()) + # TODO(unknown): Catch the following case. Need to change the calling + # convention of the whole function to process multiple line to handle it. + # printf( + # boy_this_is_a_really_long_variable_that_cannot_fit_on_the_prev_line); + printf_args = _GetTextInside(line, r"(?i)\b(string)?printf\s*\(") + if printf_args: + match = Match(r"([\w.\->()]+)$", printf_args) + if match and match.group(1) != "__VA_ARGS__": + function_name = re.search(r"\b((?:string)?printf)\s*\(", line, re.I).group( + 1 + ) + error( + filename, + linenum, + "runtime/printf", + 4, + 'Potential format string bug. Do %s("%%s", %s) instead.' + % (function_name, match.group(1)), + ) + + # Check for potential memset bugs like memset(buf, sizeof(buf), 0). + match = Search(r"memset\s*\(([^,]*),\s*([^,]*),\s*0\s*\)", line) + if match and not Match(r"^''|-?[0-9]+|0x[0-9A-Fa-f]$", match.group(2)): + error( + filename, + linenum, + "runtime/memset", + 4, + 'Did you mean "memset(%s, 0, %s)"?' % (match.group(1), match.group(2)), + ) + + if Search(r"\busing namespace\b", line): + error( + filename, + linenum, + "build/namespaces", + 5, + "Do not use namespace using-directives. " + "Use using-declarations instead.", + ) + + # Detect variable-length arrays. + match = Match(r"\s*(.+::)?(\w+) [a-z]\w*\[(.+)];", line) + if ( + match + and match.group(2) != "return" + and match.group(2) != "delete" + and match.group(3).find("]") == -1 + ): + # Split the size using space and arithmetic operators as delimiters. + # If any of the resulting tokens are not compile time constants then + # report the error. + tokens = re.split(r"\s|\+|\-|\*|\/|<<|>>]", match.group(3)) + is_const = True skip_next = False - continue - - if Search(r'sizeof\(.+\)', tok): continue - if Search(r'arraysize\(\w+\)', tok): continue - - tok = tok.lstrip('(') - tok = tok.rstrip(')') - if not tok: continue - if Match(r'\d+', tok): continue - if Match(r'0[xX][0-9a-fA-F]+', tok): continue - if Match(r'k[A-Z0-9]\w*', tok): continue - if Match(r'(.+::)?k[A-Z0-9]\w*', tok): continue - if Match(r'(.+::)?[A-Z][A-Z0-9_]*', tok): continue - # A catch all for tricky sizeof cases, including 'sizeof expression', - # 'sizeof(*type)', 'sizeof(const type)', 'sizeof(struct StructName)' - # requires skipping the next token because we split on ' ' and '*'. - if tok.startswith('sizeof'): - skip_next = True - continue - is_const = False - break - if not is_const: - error(filename, linenum, 'runtime/arrays', 1, - 'Do not use variable-length arrays. Use an appropriately named ' - "('k' followed by CamelCase) compile-time constant for the size.") - - # Check for use of unnamed namespaces in header files. Registration - # macros are typically OK, so we allow use of "namespace {" on lines - # that end with backslashes. - if (IsHeaderExtension(file_extension) - and Search(r'\bnamespace\s*{', line) - and line[-1] != '\\'): - error(filename, linenum, 'build/namespaces', 4, - 'Do not use unnamed namespaces in header files. See ' - 'https://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Namespaces' - ' for more information.') + for tok in tokens: + if skip_next: + skip_next = False + continue + + if Search(r"sizeof\(.+\)", tok): + continue + if Search(r"arraysize\(\w+\)", tok): + continue + + tok = tok.lstrip("(") + tok = tok.rstrip(")") + if not tok: + continue + if Match(r"\d+", tok): + continue + if Match(r"0[xX][0-9a-fA-F]+", tok): + continue + if Match(r"k[A-Z0-9]\w*", tok): + continue + if Match(r"(.+::)?k[A-Z0-9]\w*", tok): + continue + if Match(r"(.+::)?[A-Z][A-Z0-9_]*", tok): + continue + # A catch all for tricky sizeof cases, including 'sizeof expression', + # 'sizeof(*type)', 'sizeof(const type)', 'sizeof(struct StructName)' + # requires skipping the next token because we split on ' ' and '*'. + if tok.startswith("sizeof"): + skip_next = True + continue + is_const = False + break + if not is_const: + error( + filename, + linenum, + "runtime/arrays", + 1, + "Do not use variable-length arrays. Use an appropriately named " + "('k' followed by CamelCase) compile-time constant for the size.", + ) + + # Check for use of unnamed namespaces in header files. Registration + # macros are typically OK, so we allow use of "namespace {" on lines + # that end with backslashes. + if ( + IsHeaderExtension(file_extension) + and Search(r"\bnamespace\s*{", line) + and line[-1] != "\\" + ): + error( + filename, + linenum, + "build/namespaces", + 4, + "Do not use unnamed namespaces in header files. See " + "https://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Namespaces" + " for more information.", + ) def CheckGlobalStatic(filename, clean_lines, linenum, error): - """Check for unsafe global or static objects. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - # Match two lines at a time to support multiline declarations - if linenum + 1 < clean_lines.NumLines() and not Search(r'[;({]', line): - line += clean_lines.elided[linenum + 1].strip() - - # Check for people declaring static/global STL strings at the top level. - # This is dangerous because the C++ language does not guarantee that - # globals with constructors are initialized before the first access, and - # also because globals can be destroyed when some threads are still running. - # TODO(unknown): Generalize this to also find static unique_ptr instances. - # TODO(unknown): File bugs for clang-tidy to find these. - match = Match( - r'((?:|static +)(?:|const +))(?::*std::)?string( +const)? +' - r'([a-zA-Z0-9_:]+)\b(.*)', - line) - - # Remove false positives: - # - String pointers (as opposed to values). - # string *pointer - # const string *pointer - # string const *pointer - # string *const pointer - # - # - Functions and template specializations. - # string Function(... - # string Class::Method(... - # - # - Operators. These are matched separately because operator names - # cross non-word boundaries, and trying to match both operators - # and functions at the same time would decrease accuracy of - # matching identifiers. - # string Class::operator*() - if (match and - not Search(r'\bstring\b(\s+const)?\s*[\*\&]\s*(const\s+)?\w', line) and - not Search(r'\boperator\W', line) and - not Match(r'\s*(<.*>)?(::[a-zA-Z0-9_]+)*\s*\(([^"]|$)', match.group(4))): - if Search(r'\bconst\b', line): - error(filename, linenum, 'runtime/string', 4, - 'For a static/global string constant, use a C style string ' - 'instead: "%schar%s %s[]".' % - (match.group(1), match.group(2) or '', match.group(3))) - else: - error(filename, linenum, 'runtime/string', 4, - 'Static/global string variables are not permitted.') + """Check for unsafe global or static objects. - if (Search(r'\b([A-Za-z0-9_]*_)\(\1\)', line) or - Search(r'\b([A-Za-z0-9_]*_)\(CHECK_NOTNULL\(\1\)\)', line)): - error(filename, linenum, 'runtime/init', 4, - 'You seem to be initializing a member variable with itself.') + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Match two lines at a time to support multiline declarations + if linenum + 1 < clean_lines.NumLines() and not Search(r"[;({]", line): + line += clean_lines.elided[linenum + 1].strip() + + # Check for people declaring static/global STL strings at the top level. + # This is dangerous because the C++ language does not guarantee that + # globals with constructors are initialized before the first access, and + # also because globals can be destroyed when some threads are still running. + # TODO(unknown): Generalize this to also find static unique_ptr instances. + # TODO(unknown): File bugs for clang-tidy to find these. + match = Match( + r"((?:|static +)(?:|const +))(?::*std::)?string( +const)? +" + r"([a-zA-Z0-9_:]+)\b(.*)", + line, + ) + + # Remove false positives: + # - String pointers (as opposed to values). + # string *pointer + # const string *pointer + # string const *pointer + # string *const pointer + # + # - Functions and template specializations. + # string Function(... + # string Class::Method(... + # + # - Operators. These are matched separately because operator names + # cross non-word boundaries, and trying to match both operators + # and functions at the same time would decrease accuracy of + # matching identifiers. + # string Class::operator*() + if ( + match + and not Search(r"\bstring\b(\s+const)?\s*[\*\&]\s*(const\s+)?\w", line) + and not Search(r"\boperator\W", line) + and not Match(r'\s*(<.*>)?(::[a-zA-Z0-9_]+)*\s*\(([^"]|$)', match.group(4)) + ): + if Search(r"\bconst\b", line): + error( + filename, + linenum, + "runtime/string", + 4, + "For a static/global string constant, use a C style string " + 'instead: "%schar%s %s[]".' + % (match.group(1), match.group(2) or "", match.group(3)), + ) + else: + error( + filename, + linenum, + "runtime/string", + 4, + "Static/global string variables are not permitted.", + ) + + if Search(r"\b([A-Za-z0-9_]*_)\(\1\)", line) or Search( + r"\b([A-Za-z0-9_]*_)\(CHECK_NOTNULL\(\1\)\)", line + ): + error( + filename, + linenum, + "runtime/init", + 4, + "You seem to be initializing a member variable with itself.", + ) def CheckPrintf(filename, clean_lines, linenum, error): - """Check for printf related issues. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - # When snprintf is used, the second argument shouldn't be a literal. - match = Search(r'snprintf\s*\(([^,]*),\s*([0-9]*)\s*,', line) - if match and match.group(2) != '0': - # If 2nd arg is zero, snprintf is used to calculate size. - error(filename, linenum, 'runtime/printf', 3, - 'If you can, use sizeof(%s) instead of %s as the 2nd arg ' - 'to snprintf.' % (match.group(1), match.group(2))) - - # Check if some verboten C functions are being used. - if Search(r'\bsprintf\s*\(', line): - error(filename, linenum, 'runtime/printf', 5, - 'Never use sprintf. Use snprintf instead.') - match = Search(r'\b(strcpy|strcat)\s*\(', line) - if match: - error(filename, linenum, 'runtime/printf', 4, - 'Almost always, snprintf is better than %s' % match.group(1)) + """Check for printf related issues. + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] -def IsDerivedFunction(clean_lines, linenum): - """Check if current line contains an inherited function. - - Args: - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - Returns: - True if current line contains a function with "override" - virt-specifier. - """ - # Scan back a few lines for start of current function - for i in xrange(linenum, max(-1, linenum - 10), -1): - match = Match(r'^([^()]*\w+)\(', clean_lines.elided[i]) + # When snprintf is used, the second argument shouldn't be a literal. + match = Search(r"snprintf\s*\(([^,]*),\s*([0-9]*)\s*,", line) + if match and match.group(2) != "0": + # If 2nd arg is zero, snprintf is used to calculate size. + error( + filename, + linenum, + "runtime/printf", + 3, + "If you can, use sizeof(%s) instead of %s as the 2nd arg " + "to snprintf." % (match.group(1), match.group(2)), + ) + + # Check if some verboten C functions are being used. + if Search(r"\bsprintf\s*\(", line): + error( + filename, + linenum, + "runtime/printf", + 5, + "Never use sprintf. Use snprintf instead.", + ) + match = Search(r"\b(strcpy|strcat)\s*\(", line) if match: - # Look for "override" after the matching closing parenthesis - line, _, closing_paren = CloseExpression( - clean_lines, i, len(match.group(1))) - return (closing_paren >= 0 and - Search(r'\boverride\b', line[closing_paren:])) - return False + error( + filename, + linenum, + "runtime/printf", + 4, + "Almost always, snprintf is better than %s" % match.group(1), + ) + + +def IsDerivedFunction(clean_lines, linenum): + """Check if current line contains an inherited function. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + Returns: + True if current line contains a function with "override" + virt-specifier. + """ + # Scan back a few lines for start of current function + for i in xrange(linenum, max(-1, linenum - 10), -1): + match = Match(r"^([^()]*\w+)\(", clean_lines.elided[i]) + if match: + # Look for "override" after the matching closing parenthesis + line, _, closing_paren = CloseExpression( + clean_lines, i, len(match.group(1)) + ) + return closing_paren >= 0 and Search(r"\boverride\b", line[closing_paren:]) + return False def IsOutOfLineMethodDefinition(clean_lines, linenum): - """Check if current line contains an out-of-line method definition. + """Check if current line contains an out-of-line method definition. - Args: - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - Returns: - True if current line contains an out-of-line method definition. - """ - # Scan back a few lines for start of current function - for i in xrange(linenum, max(-1, linenum - 10), -1): - if Match(r'^([^()]*\w+)\(', clean_lines.elided[i]): - return Match(r'^[^()]*\w+::\w+\(', clean_lines.elided[i]) is not None - return False + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + Returns: + True if current line contains an out-of-line method definition. + """ + # Scan back a few lines for start of current function + for i in xrange(linenum, max(-1, linenum - 10), -1): + if Match(r"^([^()]*\w+)\(", clean_lines.elided[i]): + return Match(r"^[^()]*\w+::\w+\(", clean_lines.elided[i]) is not None + return False def IsInitializerList(clean_lines, linenum): - """Check if current line is inside constructor initializer list. - - Args: - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - Returns: - True if current line appears to be inside constructor initializer - list, False otherwise. - """ - for i in xrange(linenum, 1, -1): - line = clean_lines.elided[i] - if i == linenum: - remove_function_body = Match(r'^(.*)\{\s*$', line) - if remove_function_body: - line = remove_function_body.group(1) - - if Search(r'\s:\s*\w+[({]', line): - # A lone colon tend to indicate the start of a constructor - # initializer list. It could also be a ternary operator, which - # also tend to appear in constructor initializer lists as - # opposed to parameter lists. - return True - if Search(r'\}\s*,\s*$', line): - # A closing brace followed by a comma is probably the end of a - # brace-initialized member in constructor initializer list. - return True - if Search(r'[{};]\s*$', line): - # Found one of the following: - # - A closing brace or semicolon, probably the end of the previous - # function. - # - An opening brace, probably the start of current class or namespace. - # - # Current line is probably not inside an initializer list since - # we saw one of those things without seeing the starting colon. - return False - - # Got to the beginning of the file without seeing the start of - # constructor initializer list. - return False - - -def CheckForNonConstReference(filename, clean_lines, linenum, - nesting_state, error): - """Check for non-const references. - - Separate from CheckLanguage since it scans backwards from current - line, instead of scanning forward. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - nesting_state: A NestingState instance which maintains information about - the current stack of nested blocks being parsed. - error: The function to call with any errors found. - """ - # Do nothing if there is no '&' on current line. - line = clean_lines.elided[linenum] - if '&' not in line: - return - - # If a function is inherited, current function doesn't have much of - # a choice, so any non-const references should not be blamed on - # derived function. - if IsDerivedFunction(clean_lines, linenum): - return - - # Don't warn on out-of-line method definitions, as we would warn on the - # in-line declaration, if it isn't marked with 'override'. - if IsOutOfLineMethodDefinition(clean_lines, linenum): - return - - # Long type names may be broken across multiple lines, usually in one - # of these forms: - # LongType - # ::LongTypeContinued &identifier - # LongType:: - # LongTypeContinued &identifier - # LongType< - # ...>::LongTypeContinued &identifier - # - # If we detected a type split across two lines, join the previous - # line to current line so that we can match const references - # accordingly. - # - # Note that this only scans back one line, since scanning back - # arbitrary number of lines would be expensive. If you have a type - # that spans more than 2 lines, please use a typedef. - if linenum > 1: - previous = None - if Match(r'\s*::(?:[\w<>]|::)+\s*&\s*\S', line): - # previous_line\n + ::current_line - previous = Search(r'\b((?:const\s*)?(?:[\w<>]|::)+[\w<>])\s*$', - clean_lines.elided[linenum - 1]) - elif Match(r'\s*[a-zA-Z_]([\w<>]|::)+\s*&\s*\S', line): - # previous_line::\n + current_line - previous = Search(r'\b((?:const\s*)?(?:[\w<>]|::)+::)\s*$', - clean_lines.elided[linenum - 1]) - if previous: - line = previous.group(1) + line.lstrip() - else: - # Check for templated parameter that is split across multiple lines - endpos = line.rfind('>') - if endpos > -1: - (_, startline, startpos) = ReverseCloseExpression( - clean_lines, linenum, endpos) - if startpos > -1 and startline < linenum: - # Found the matching < on an earlier line, collect all - # pieces up to current line. - line = '' - for i in xrange(startline, linenum + 1): - line += clean_lines.elided[i].strip() - - # Check for non-const references in function parameters. A single '&' may - # found in the following places: - # inside expression: binary & for bitwise AND - # inside expression: unary & for taking the address of something - # inside declarators: reference parameter - # We will exclude the first two cases by checking that we are not inside a - # function body, including one that was just introduced by a trailing '{'. - # TODO(unknown): Doesn't account for 'catch(Exception& e)' [rare]. - if (nesting_state.previous_stack_top and - not (isinstance(nesting_state.previous_stack_top, _ClassInfo) or - isinstance(nesting_state.previous_stack_top, _NamespaceInfo))): - # Not at toplevel, not within a class, and not within a namespace - return - - # Avoid initializer lists. We only need to scan back from the - # current line for something that starts with ':'. - # - # We don't need to check the current line, since the '&' would - # appear inside the second set of parentheses on the current line as - # opposed to the first set. - if linenum > 0: - for i in xrange(linenum - 1, max(0, linenum - 10), -1): - previous_line = clean_lines.elided[i] - if not Search(r'[),]\s*$', previous_line): - break - if Match(r'^\s*:\s+\S', previous_line): + """Check if current line is inside constructor initializer list. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + Returns: + True if current line appears to be inside constructor initializer + list, False otherwise. + """ + for i in xrange(linenum, 1, -1): + line = clean_lines.elided[i] + if i == linenum: + remove_function_body = Match(r"^(.*)\{\s*$", line) + if remove_function_body: + line = remove_function_body.group(1) + + if Search(r"\s:\s*\w+[({]", line): + # A lone colon tend to indicate the start of a constructor + # initializer list. It could also be a ternary operator, which + # also tend to appear in constructor initializer lists as + # opposed to parameter lists. + return True + if Search(r"\}\s*,\s*$", line): + # A closing brace followed by a comma is probably the end of a + # brace-initialized member in constructor initializer list. + return True + if Search(r"[{};]\s*$", line): + # Found one of the following: + # - A closing brace or semicolon, probably the end of the previous + # function. + # - An opening brace, probably the start of current class or namespace. + # + # Current line is probably not inside an initializer list since + # we saw one of those things without seeing the starting colon. + return False + + # Got to the beginning of the file without seeing the start of + # constructor initializer list. + return False + + +def CheckForNonConstReference(filename, clean_lines, linenum, nesting_state, error): + """Check for non-const references. + + Separate from CheckLanguage since it scans backwards from current + line, instead of scanning forward. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: The function to call with any errors found. + """ + # Do nothing if there is no '&' on current line. + line = clean_lines.elided[linenum] + if "&" not in line: return - # Avoid preprocessors - if Search(r'\\\s*$', line): - return - - # Avoid constructor initializer lists - if IsInitializerList(clean_lines, linenum): - return - - # We allow non-const references in a few standard places, like functions - # called "swap()" or iostream operators like "<<" or ">>". Do not check - # those function parameters. - # - # We also accept & in static_assert, which looks like a function but - # it's actually a declaration expression. - allowed_functions = (r'(?:[sS]wap(?:<\w:+>)?|' - r'operator\s*[<>][<>]|' - r'static_assert|COMPILE_ASSERT' - r')\s*\(') - if Search(allowed_functions, line): - return - elif not Search(r'\S+\([^)]*$', line): - # Don't see an allowed function entry on this line. Actually we - # didn't see any function name on this line, so this is likely a - # multi-line parameter list. Try a bit harder to catch this case. - for i in xrange(2): - if (linenum > i and - Search(allowed_functions, clean_lines.elided[linenum - i - 1])): + # If a function is inherited, current function doesn't have much of + # a choice, so any non-const references should not be blamed on + # derived function. + if IsDerivedFunction(clean_lines, linenum): return - decls = ReplaceAll(r'{[^}]*}', ' ', line) # exclude function body - for parameter in re.findall(_RE_PATTERN_REF_PARAM, decls): - if (not Match(_RE_PATTERN_CONST_REF_PARAM, parameter) and - not Match(_RE_PATTERN_REF_STREAM_PARAM, parameter)): - error(filename, linenum, 'runtime/references', 2, - 'Is this a non-const reference? ' - 'If so, make const or use a pointer: ' + - ReplaceAll(' *<', '<', parameter)) + # Don't warn on out-of-line method definitions, as we would warn on the + # in-line declaration, if it isn't marked with 'override'. + if IsOutOfLineMethodDefinition(clean_lines, linenum): + return + + # Long type names may be broken across multiple lines, usually in one + # of these forms: + # LongType + # ::LongTypeContinued &identifier + # LongType:: + # LongTypeContinued &identifier + # LongType< + # ...>::LongTypeContinued &identifier + # + # If we detected a type split across two lines, join the previous + # line to current line so that we can match const references + # accordingly. + # + # Note that this only scans back one line, since scanning back + # arbitrary number of lines would be expensive. If you have a type + # that spans more than 2 lines, please use a typedef. + if linenum > 1: + previous = None + if Match(r"\s*::(?:[\w<>]|::)+\s*&\s*\S", line): + # previous_line\n + ::current_line + previous = Search( + r"\b((?:const\s*)?(?:[\w<>]|::)+[\w<>])\s*$", + clean_lines.elided[linenum - 1], + ) + elif Match(r"\s*[a-zA-Z_]([\w<>]|::)+\s*&\s*\S", line): + # previous_line::\n + current_line + previous = Search( + r"\b((?:const\s*)?(?:[\w<>]|::)+::)\s*$", + clean_lines.elided[linenum - 1], + ) + if previous: + line = previous.group(1) + line.lstrip() + else: + # Check for templated parameter that is split across multiple lines + endpos = line.rfind(">") + if endpos > -1: + (_, startline, startpos) = ReverseCloseExpression( + clean_lines, linenum, endpos + ) + if startpos > -1 and startline < linenum: + # Found the matching < on an earlier line, collect all + # pieces up to current line. + line = "" + for i in xrange(startline, linenum + 1): + line += clean_lines.elided[i].strip() + + # Check for non-const references in function parameters. A single '&' may + # found in the following places: + # inside expression: binary & for bitwise AND + # inside expression: unary & for taking the address of something + # inside declarators: reference parameter + # We will exclude the first two cases by checking that we are not inside a + # function body, including one that was just introduced by a trailing '{'. + # TODO(unknown): Doesn't account for 'catch(Exception& e)' [rare]. + if nesting_state.previous_stack_top and not ( + isinstance(nesting_state.previous_stack_top, _ClassInfo) + or isinstance(nesting_state.previous_stack_top, _NamespaceInfo) + ): + # Not at toplevel, not within a class, and not within a namespace + return + + # Avoid initializer lists. We only need to scan back from the + # current line for something that starts with ':'. + # + # We don't need to check the current line, since the '&' would + # appear inside the second set of parentheses on the current line as + # opposed to the first set. + if linenum > 0: + for i in xrange(linenum - 1, max(0, linenum - 10), -1): + previous_line = clean_lines.elided[i] + if not Search(r"[),]\s*$", previous_line): + break + if Match(r"^\s*:\s+\S", previous_line): + return + + # Avoid preprocessors + if Search(r"\\\s*$", line): + return + + # Avoid constructor initializer lists + if IsInitializerList(clean_lines, linenum): + return + + # We allow non-const references in a few standard places, like functions + # called "swap()" or iostream operators like "<<" or ">>". Do not check + # those function parameters. + # + # We also accept & in static_assert, which looks like a function but + # it's actually a declaration expression. + allowed_functions = ( + r"(?:[sS]wap(?:<\w:+>)?|" + r"operator\s*[<>][<>]|" + r"static_assert|COMPILE_ASSERT" + r")\s*\(" + ) + if Search(allowed_functions, line): + return + elif not Search(r"\S+\([^)]*$", line): + # Don't see an allowed function entry on this line. Actually we + # didn't see any function name on this line, so this is likely a + # multi-line parameter list. Try a bit harder to catch this case. + for i in xrange(2): + if linenum > i and Search( + allowed_functions, clean_lines.elided[linenum - i - 1] + ): + return + + decls = ReplaceAll(r"{[^}]*}", " ", line) # exclude function body + for parameter in re.findall(_RE_PATTERN_REF_PARAM, decls): + if not Match(_RE_PATTERN_CONST_REF_PARAM, parameter) and not Match( + _RE_PATTERN_REF_STREAM_PARAM, parameter + ): + error( + filename, + linenum, + "runtime/references", + 2, + "Is this a non-const reference? " + "If so, make const or use a pointer: " + + ReplaceAll(" *<", "<", parameter), + ) def CheckCasts(filename, clean_lines, linenum, error): - """Various cast related checks. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - # Check to see if they're using an conversion function cast. - # I just try to capture the most common basic types, though there are more. - # Parameterless conversion functions, such as bool(), are allowed as they are - # probably a member operator declaration or default constructor. - match = Search( - r'(\bnew\s+(?:const\s+)?|\S<\s*(?:const\s+)?)?\b' - r'(int|float|double|bool|char|int32|uint32|int64|uint64)' - r'(\([^)].*)', line) - expecting_function = ExpectingFunctionArgs(clean_lines, linenum) - if match and not expecting_function: - matched_type = match.group(2) - - # matched_new_or_template is used to silence two false positives: - # - New operators - # - Template arguments with function types + """Various cast related checks. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Check to see if they're using an conversion function cast. + # I just try to capture the most common basic types, though there are more. + # Parameterless conversion functions, such as bool(), are allowed as they are + # probably a member operator declaration or default constructor. + match = Search( + r"(\bnew\s+(?:const\s+)?|\S<\s*(?:const\s+)?)?\b" + r"(int|float|double|bool|char|int32|uint32|int64|uint64)" + r"(\([^)].*)", + line, + ) + expecting_function = ExpectingFunctionArgs(clean_lines, linenum) + if match and not expecting_function: + matched_type = match.group(2) + + # matched_new_or_template is used to silence two false positives: + # - New operators + # - Template arguments with function types + # + # For template arguments, we match on types immediately following + # an opening bracket without any spaces. This is a fast way to + # silence the common case where the function type is the first + # template argument. False negative with less-than comparison is + # avoided because those operators are usually followed by a space. + # + # function // bracket + no space = false positive + # value < double(42) // bracket + space = true positive + matched_new_or_template = match.group(1) + + # Avoid arrays by looking for brackets that come after the closing + # parenthesis. + if Match(r"\([^()]+\)\s*\[", match.group(3)): + return + + # Other things to ignore: + # - Function pointers + # - Casts to pointer types + # - Placement new + # - Alias declarations + matched_funcptr = match.group(3) + if ( + matched_new_or_template is None + and not ( + matched_funcptr + and ( + Match(r"\((?:[^() ]+::\s*\*\s*)?[^() ]+\)\s*\(", matched_funcptr) + or matched_funcptr.startswith("(*)") + ) + ) + and not Match(r"\s*using\s+\S+\s*=\s*" + matched_type, line) + and not Search(r"new\(\S+\)\s*" + matched_type, line) + ): + error( + filename, + linenum, + "readability/casting", + 4, + "Using deprecated casting style. " + "Use static_cast<%s>(...) instead" % matched_type, + ) + + if not expecting_function: + CheckCStyleCast( + filename, + clean_lines, + linenum, + "static_cast", + r"\((int|float|double|bool|char|u?int(16|32|64))\)", + error, + ) + + # This doesn't catch all cases. Consider (const char * const)"hello". # - # For template arguments, we match on types immediately following - # an opening bracket without any spaces. This is a fast way to - # silence the common case where the function type is the first - # template argument. False negative with less-than comparison is - # avoided because those operators are usually followed by a space. + # (char *) "foo" should always be a const_cast (reinterpret_cast won't + # compile). + if CheckCStyleCast( + filename, clean_lines, linenum, "const_cast", r'\((char\s?\*+\s?)\)\s*"', error + ): + pass + else: + # Check pointer casts for other than string constants + CheckCStyleCast( + filename, + clean_lines, + linenum, + "reinterpret_cast", + r"\((\w+\s?\*+\s?)\)", + error, + ) + + # In addition, we look for people taking the address of a cast. This + # is dangerous -- casts can assign to temporaries, so the pointer doesn't + # point where you think. # - # function // bracket + no space = false positive - # value < double(42) // bracket + space = true positive - matched_new_or_template = match.group(1) - - # Avoid arrays by looking for brackets that come after the closing - # parenthesis. - if Match(r'\([^()]+\)\s*\[', match.group(3)): - return - - # Other things to ignore: - # - Function pointers - # - Casts to pointer types - # - Placement new - # - Alias declarations - matched_funcptr = match.group(3) - if (matched_new_or_template is None and - not (matched_funcptr and - (Match(r'\((?:[^() ]+::\s*\*\s*)?[^() ]+\)\s*\(', - matched_funcptr) or - matched_funcptr.startswith('(*)'))) and - not Match(r'\s*using\s+\S+\s*=\s*' + matched_type, line) and - not Search(r'new\(\S+\)\s*' + matched_type, line)): - error(filename, linenum, 'readability/casting', 4, - 'Using deprecated casting style. ' - 'Use static_cast<%s>(...) instead' % - matched_type) - - if not expecting_function: - CheckCStyleCast(filename, clean_lines, linenum, 'static_cast', - r'\((int|float|double|bool|char|u?int(16|32|64))\)', error) - - # This doesn't catch all cases. Consider (const char * const)"hello". - # - # (char *) "foo" should always be a const_cast (reinterpret_cast won't - # compile). - if CheckCStyleCast(filename, clean_lines, linenum, 'const_cast', - r'\((char\s?\*+\s?)\)\s*"', error): - pass - else: - # Check pointer casts for other than string constants - CheckCStyleCast(filename, clean_lines, linenum, 'reinterpret_cast', - r'\((\w+\s?\*+\s?)\)', error) - - # In addition, we look for people taking the address of a cast. This - # is dangerous -- casts can assign to temporaries, so the pointer doesn't - # point where you think. - # - # Some non-identifier character is required before the '&' for the - # expression to be recognized as a cast. These are casts: - # expression = &static_cast(temporary()); - # function(&(int*)(temporary())); - # - # This is not a cast: - # reference_type&(int* function_param); - match = Search( - r'(?:[^\w]&\(([^)*][^)]*)\)[\w(])|' - r'(?:[^\w]&(static|dynamic|down|reinterpret)_cast\b)', line) - if match: - # Try a better error message when the & is bound to something - # dereferenced by the casted pointer, as opposed to the casted - # pointer itself. - parenthesis_error = False - match = Match(r'^(.*&(?:static|dynamic|down|reinterpret)_cast\b)<', line) + # Some non-identifier character is required before the '&' for the + # expression to be recognized as a cast. These are casts: + # expression = &static_cast(temporary()); + # function(&(int*)(temporary())); + # + # This is not a cast: + # reference_type&(int* function_param); + match = Search( + r"(?:[^\w]&\(([^)*][^)]*)\)[\w(])|" + r"(?:[^\w]&(static|dynamic|down|reinterpret)_cast\b)", + line, + ) if match: - _, y1, x1 = CloseExpression(clean_lines, linenum, len(match.group(1))) - if x1 >= 0 and clean_lines.elided[y1][x1] == '(': - _, y2, x2 = CloseExpression(clean_lines, y1, x1) - if x2 >= 0: - extended_line = clean_lines.elided[y2][x2:] - if y2 < clean_lines.NumLines() - 1: - extended_line += clean_lines.elided[y2 + 1] - if Match(r'\s*(?:->|\[)', extended_line): - parenthesis_error = True - - if parenthesis_error: - error(filename, linenum, 'readability/casting', 4, - ('Are you taking an address of something dereferenced ' - 'from a cast? Wrapping the dereferenced expression in ' - 'parentheses will make the binding more obvious')) - else: - error(filename, linenum, 'runtime/casting', 4, - ('Are you taking an address of a cast? ' - 'This is dangerous: could be a temp var. ' - 'Take the address before doing the cast, rather than after')) + # Try a better error message when the & is bound to something + # dereferenced by the casted pointer, as opposed to the casted + # pointer itself. + parenthesis_error = False + match = Match(r"^(.*&(?:static|dynamic|down|reinterpret)_cast\b)<", line) + if match: + _, y1, x1 = CloseExpression(clean_lines, linenum, len(match.group(1))) + if x1 >= 0 and clean_lines.elided[y1][x1] == "(": + _, y2, x2 = CloseExpression(clean_lines, y1, x1) + if x2 >= 0: + extended_line = clean_lines.elided[y2][x2:] + if y2 < clean_lines.NumLines() - 1: + extended_line += clean_lines.elided[y2 + 1] + if Match(r"\s*(?:->|\[)", extended_line): + parenthesis_error = True + + if parenthesis_error: + error( + filename, + linenum, + "readability/casting", + 4, + ( + "Are you taking an address of something dereferenced " + "from a cast? Wrapping the dereferenced expression in " + "parentheses will make the binding more obvious" + ), + ) + else: + error( + filename, + linenum, + "runtime/casting", + 4, + ( + "Are you taking an address of a cast? " + "This is dangerous: could be a temp var. " + "Take the address before doing the cast, rather than after" + ), + ) def CheckCStyleCast(filename, clean_lines, linenum, cast_type, pattern, error): - """Checks for a C-style cast by looking for the pattern. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - cast_type: The string for the C++ cast to recommend. This is either - reinterpret_cast, static_cast, or const_cast, depending. - pattern: The regular expression used to find C-style casts. - error: The function to call with any errors found. - - Returns: - True if an error was emitted. - False otherwise. - """ - line = clean_lines.elided[linenum] - match = Search(pattern, line) - if not match: - return False + """Checks for a C-style cast by looking for the pattern. - # Exclude lines with keywords that tend to look like casts - context = line[0:match.start(1) - 1] - if Match(r'.*\b(?:sizeof|alignof|alignas|[_A-Z][_A-Z0-9]*)\s*$', context): - return False + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + cast_type: The string for the C++ cast to recommend. This is either + reinterpret_cast, static_cast, or const_cast, depending. + pattern: The regular expression used to find C-style casts. + error: The function to call with any errors found. - # Try expanding current context to see if we one level of - # parentheses inside a macro. - if linenum > 0: - for i in xrange(linenum - 1, max(0, linenum - 5), -1): - context = clean_lines.elided[i] + context - if Match(r'.*\b[_A-Z][_A-Z0-9]*\s*\((?:\([^()]*\)|[^()])*$', context): - return False + Returns: + True if an error was emitted. + False otherwise. + """ + line = clean_lines.elided[linenum] + match = Search(pattern, line) + if not match: + return False - # operator++(int) and operator--(int) - if context.endswith(' operator++') or context.endswith(' operator--'): - return False + # Exclude lines with keywords that tend to look like casts + context = line[0 : match.start(1) - 1] + if Match(r".*\b(?:sizeof|alignof|alignas|[_A-Z][_A-Z0-9]*)\s*$", context): + return False - # A single unnamed argument for a function tends to look like old style cast. - # If we see those, don't issue warnings for deprecated casts. - remainder = line[match.end(0):] - if Match(r'^\s*(?:;|const\b|throw\b|final\b|override\b|[=>{),]|->)', - remainder): - return False + # Try expanding current context to see if we one level of + # parentheses inside a macro. + if linenum > 0: + for i in xrange(linenum - 1, max(0, linenum - 5), -1): + context = clean_lines.elided[i] + context + if Match(r".*\b[_A-Z][_A-Z0-9]*\s*\((?:\([^()]*\)|[^()])*$", context): + return False + + # operator++(int) and operator--(int) + if context.endswith(" operator++") or context.endswith(" operator--"): + return False - # At this point, all that should be left is actual casts. - error(filename, linenum, 'readability/casting', 4, - 'Using C-style cast. Use %s<%s>(...) instead' % - (cast_type, match.group(1))) + # A single unnamed argument for a function tends to look like old style cast. + # If we see those, don't issue warnings for deprecated casts. + remainder = line[match.end(0) :] + if Match(r"^\s*(?:;|const\b|throw\b|final\b|override\b|[=>{),]|->)", remainder): + return False + + # At this point, all that should be left is actual casts. + error( + filename, + linenum, + "readability/casting", + 4, + "Using C-style cast. Use %s<%s>(...) instead" % (cast_type, match.group(1)), + ) - return True + return True def ExpectingFunctionArgs(clean_lines, linenum): - """Checks whether where function type arguments are expected. - - Args: - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - - Returns: - True if the line at 'linenum' is inside something that expects arguments - of function types. - """ - line = clean_lines.elided[linenum] - return (Match(r'^\s*MOCK_(CONST_)?METHOD\d+(_T)?\(', line) or - (linenum >= 2 and - (Match(r'^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\((?:\S+,)?\s*$', - clean_lines.elided[linenum - 1]) or - Match(r'^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\(\s*$', - clean_lines.elided[linenum - 2]) or - Search(r'\bstd::m?function\s*\<\s*$', - clean_lines.elided[linenum - 1])))) + """Checks whether where function type arguments are expected. + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. -_HEADERS_CONTAINING_TEMPLATES = ( - ('', ('deque',)), - ('', ('unary_function', 'binary_function', - 'plus', 'minus', 'multiplies', 'divides', 'modulus', - 'negate', - 'equal_to', 'not_equal_to', 'greater', 'less', - 'greater_equal', 'less_equal', - 'logical_and', 'logical_or', 'logical_not', - 'unary_negate', 'not1', 'binary_negate', 'not2', - 'bind1st', 'bind2nd', - 'pointer_to_unary_function', - 'pointer_to_binary_function', - 'ptr_fun', - 'mem_fun_t', 'mem_fun', 'mem_fun1_t', 'mem_fun1_ref_t', - 'mem_fun_ref_t', - 'const_mem_fun_t', 'const_mem_fun1_t', - 'const_mem_fun_ref_t', 'const_mem_fun1_ref_t', - 'mem_fun_ref', - )), - ('', ('numeric_limits',)), - ('', ('list',)), - ('', ('map', 'multimap',)), - ('', ('allocator', 'make_shared', 'make_unique', 'shared_ptr', - 'unique_ptr', 'weak_ptr')), - ('', ('queue', 'priority_queue',)), - ('', ('set', 'multiset',)), - ('', ('stack',)), - ('', ('char_traits', 'basic_string',)), - ('', ('tuple',)), - ('', ('unordered_map', 'unordered_multimap')), - ('', ('unordered_set', 'unordered_multiset')), - ('', ('pair',)), - ('', ('vector',)), + Returns: + True if the line at 'linenum' is inside something that expects arguments + of function types. + """ + line = clean_lines.elided[linenum] + return Match(r"^\s*MOCK_(CONST_)?METHOD\d+(_T)?\(", line) or ( + linenum >= 2 + and ( + Match( + r"^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\((?:\S+,)?\s*$", + clean_lines.elided[linenum - 1], + ) + or Match( + r"^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\(\s*$", + clean_lines.elided[linenum - 2], + ) + or Search(r"\bstd::m?function\s*\<\s*$", clean_lines.elided[linenum - 1]) + ) + ) + +_HEADERS_CONTAINING_TEMPLATES = ( + ("", ("deque",)), + ( + "", + ( + "unary_function", + "binary_function", + "plus", + "minus", + "multiplies", + "divides", + "modulus", + "negate", + "equal_to", + "not_equal_to", + "greater", + "less", + "greater_equal", + "less_equal", + "logical_and", + "logical_or", + "logical_not", + "unary_negate", + "not1", + "binary_negate", + "not2", + "bind1st", + "bind2nd", + "pointer_to_unary_function", + "pointer_to_binary_function", + "ptr_fun", + "mem_fun_t", + "mem_fun", + "mem_fun1_t", + "mem_fun1_ref_t", + "mem_fun_ref_t", + "const_mem_fun_t", + "const_mem_fun1_t", + "const_mem_fun_ref_t", + "const_mem_fun1_ref_t", + "mem_fun_ref", + ), + ), + ("", ("numeric_limits",)), + ("", ("list",)), + ( + "", + ( + "map", + "multimap", + ), + ), + ( + "", + ( + "allocator", + "make_shared", + "make_unique", + "shared_ptr", + "unique_ptr", + "weak_ptr", + ), + ), + ( + "", + ( + "queue", + "priority_queue", + ), + ), + ( + "", + ( + "set", + "multiset", + ), + ), + ("", ("stack",)), + ( + "", + ( + "char_traits", + "basic_string", + ), + ), + ("", ("tuple",)), + ("", ("unordered_map", "unordered_multimap")), + ("", ("unordered_set", "unordered_multiset")), + ("", ("pair",)), + ("", ("vector",)), # gcc extensions. # Note: std::hash is their hash, ::hash is our hash - ('', ('hash_map', 'hash_multimap',)), - ('', ('hash_set', 'hash_multiset',)), - ('', ('slist',)), - ) + ( + "", + ( + "hash_map", + "hash_multimap", + ), + ), + ( + "", + ( + "hash_set", + "hash_multiset", + ), + ), + ("", ("slist",)), +) _HEADERS_MAYBE_TEMPLATES = ( - ('', ('copy', 'max', 'min', 'min_element', 'sort', - 'transform', - )), - ('', ('forward', 'make_pair', 'move', 'swap')), - ) - -_RE_PATTERN_STRING = re.compile(r'\bstring\b') + ( + "", + ( + "copy", + "max", + "min", + "min_element", + "sort", + "transform", + ), + ), + ("", ("forward", "make_pair", "move", "swap")), +) + +_RE_PATTERN_STRING = re.compile(r"\bstring\b") _re_pattern_headers_maybe_templates = [] for _header, _templates in _HEADERS_MAYBE_TEMPLATES: - for _template in _templates: - # Match max(..., ...), max(..., ...), but not foo->max, foo.max or - # type::max(). - _re_pattern_headers_maybe_templates.append( - (re.compile(r'[^>.]\b' + _template + r'(<.*?>)?\([^\)]'), - _template, - _header)) + for _template in _templates: + # Match max(..., ...), max(..., ...), but not foo->max, foo.max or + # type::max(). + _re_pattern_headers_maybe_templates.append( + ( + re.compile(r"[^>.]\b" + _template + r"(<.*?>)?\([^\)]"), + _template, + _header, + ) + ) # Other scripts may reach in and modify this pattern. _re_pattern_templates = [] for _header, _templates in _HEADERS_CONTAINING_TEMPLATES: - for _template in _templates: - _re_pattern_templates.append( - (re.compile(r'(\<|\b)' + _template + r'\s*\<'), - _template + '<>', - _header)) + for _template in _templates: + _re_pattern_templates.append( + (re.compile(r"(\<|\b)" + _template + r"\s*\<"), _template + "<>", _header) + ) def FilesBelongToSameModule(filename_cc, filename_h): - """Check if these two filenames belong to the same module. - - The concept of a 'module' here is a as follows: - foo.h, foo-inl.h, foo.cc, foo_test.cc and foo_unittest.cc belong to the - same 'module' if they are in the same directory. - some/path/public/xyzzy and some/path/internal/xyzzy are also considered - to belong to the same module here. - - If the filename_cc contains a longer path than the filename_h, for example, - '/absolute/path/to/base/sysinfo.cc', and this file would include - 'base/sysinfo.h', this function also produces the prefix needed to open the - header. This is used by the caller of this function to more robustly open the - header file. We don't have access to the real include paths in this context, - so we need this guesswork here. - - Known bugs: tools/base/bar.cc and base/bar.h belong to the same module - according to this implementation. Because of this, this function gives - some false positives. This should be sufficiently rare in practice. - - Args: - filename_cc: is the path for the .cc file - filename_h: is the path for the header path - - Returns: - Tuple with a bool and a string: - bool: True if filename_cc and filename_h belong to the same module. - string: the additional prefix needed to open the header file. - """ - - fileinfo = FileInfo(filename_cc) - if not fileinfo.IsSource(): - return (False, '') - filename_cc = filename_cc[:-len(fileinfo.Extension())] - matched_test_suffix = Search(_TEST_FILE_SUFFIX, fileinfo.BaseName()) - if matched_test_suffix: - filename_cc = filename_cc[:-len(matched_test_suffix.group(1))] - filename_cc = filename_cc.replace('/public/', '/') - filename_cc = filename_cc.replace('/internal/', '/') - - if not filename_h.endswith('.h'): - return (False, '') - filename_h = filename_h[:-len('.h')] - if filename_h.endswith('-inl'): - filename_h = filename_h[:-len('-inl')] - filename_h = filename_h.replace('/public/', '/') - filename_h = filename_h.replace('/internal/', '/') - - files_belong_to_same_module = filename_cc.endswith(filename_h) - common_path = '' - if files_belong_to_same_module: - common_path = filename_cc[:-len(filename_h)] - return files_belong_to_same_module, common_path + """Check if these two filenames belong to the same module. + + The concept of a 'module' here is a as follows: + foo.h, foo-inl.h, foo.cc, foo_test.cc and foo_unittest.cc belong to the + same 'module' if they are in the same directory. + some/path/public/xyzzy and some/path/internal/xyzzy are also considered + to belong to the same module here. + + If the filename_cc contains a longer path than the filename_h, for example, + '/absolute/path/to/base/sysinfo.cc', and this file would include + 'base/sysinfo.h', this function also produces the prefix needed to open the + header. This is used by the caller of this function to more robustly open the + header file. We don't have access to the real include paths in this context, + so we need this guesswork here. + + Known bugs: tools/base/bar.cc and base/bar.h belong to the same module + according to this implementation. Because of this, this function gives + some false positives. This should be sufficiently rare in practice. + + Args: + filename_cc: is the path for the .cc file + filename_h: is the path for the header path + + Returns: + Tuple with a bool and a string: + bool: True if filename_cc and filename_h belong to the same module. + string: the additional prefix needed to open the header file. + """ + + fileinfo = FileInfo(filename_cc) + if not fileinfo.IsSource(): + return (False, "") + filename_cc = filename_cc[: -len(fileinfo.Extension())] + matched_test_suffix = Search(_TEST_FILE_SUFFIX, fileinfo.BaseName()) + if matched_test_suffix: + filename_cc = filename_cc[: -len(matched_test_suffix.group(1))] + filename_cc = filename_cc.replace("/public/", "/") + filename_cc = filename_cc.replace("/internal/", "/") + + if not filename_h.endswith(".h"): + return (False, "") + filename_h = filename_h[: -len(".h")] + if filename_h.endswith("-inl"): + filename_h = filename_h[: -len("-inl")] + filename_h = filename_h.replace("/public/", "/") + filename_h = filename_h.replace("/internal/", "/") + + files_belong_to_same_module = filename_cc.endswith(filename_h) + common_path = "" + if files_belong_to_same_module: + common_path = filename_cc[: -len(filename_h)] + return files_belong_to_same_module, common_path def UpdateIncludeState(filename, include_dict, io=codecs): - """Fill up the include_dict with new includes found from the file. - - Args: - filename: the name of the header to read. - include_dict: a dictionary in which the headers are inserted. - io: The io factory to use to read the file. Provided for testability. - - Returns: - True if a header was successfully added. False otherwise. - """ - headerfile = None - try: - headerfile = io.open(filename, 'r', 'utf8', 'replace') - except IOError: - return False - linenum = 0 - for line in headerfile: - linenum += 1 - clean_line = CleanseComments(line) - match = _RE_PATTERN_INCLUDE.search(clean_line) - if match: - include = match.group(2) - include_dict.setdefault(include, linenum) - return True - - -def CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error, - io=codecs): - """Reports for missing stl includes. - - This function will output warnings to make sure you are including the headers - necessary for the stl containers and functions that you use. We only give one - reason to include a header. For example, if you use both equal_to<> and - less<> in a .h file, only one (the latter in the file) of these will be - reported as a reason to include the . - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - include_state: An _IncludeState instance. - error: The function to call with any errors found. - io: The IO factory to use to read the header file. Provided for unittest - injection. - """ - required = {} # A map of header name to linenumber and the template entity. - # Example of required: { '': (1219, 'less<>') } - - for linenum in xrange(clean_lines.NumLines()): - line = clean_lines.elided[linenum] - if not line or line[0] == '#': - continue + """Fill up the include_dict with new includes found from the file. - # String is special -- it is a non-templatized type in STL. - matched = _RE_PATTERN_STRING.search(line) - if matched: - # Don't warn about strings in non-STL namespaces: - # (We check only the first match per line; good enough.) - prefix = line[:matched.start()] - if prefix.endswith('std::') or not prefix.endswith('::'): - required[''] = (linenum, 'string') - - for pattern, template, header in _re_pattern_headers_maybe_templates: - if pattern.search(line): - required[header] = (linenum, template) - - # The following function is just a speed up, no semantics are changed. - if not '<' in line: # Reduces the cpu time usage by skipping lines. - continue - - for pattern, template, header in _re_pattern_templates: - matched = pattern.search(line) - if matched: - # Don't warn about IWYU in non-STL namespaces: - # (We check only the first match per line; good enough.) - prefix = line[:matched.start()] - if prefix.endswith('std::') or not prefix.endswith('::'): - required[header] = (linenum, template) - - # The policy is that if you #include something in foo.h you don't need to - # include it again in foo.cc. Here, we will look at possible includes. - # Let's flatten the include_state include_list and copy it into a dictionary. - include_dict = dict([item for sublist in include_state.include_list - for item in sublist]) - - # Did we find the header for this file (if any) and successfully load it? - header_found = False - - # Use the absolute path so that matching works properly. - abs_filename = FileInfo(filename).FullName() - - # For Emacs's flymake. - # If cpplint is invoked from Emacs's flymake, a temporary file is generated - # by flymake and that file name might end with '_flymake.cc'. In that case, - # restore original file name here so that the corresponding header file can be - # found. - # e.g. If the file name is 'foo_flymake.cc', we should search for 'foo.h' - # instead of 'foo_flymake.h' - abs_filename = re.sub(r'_flymake\.cc$', '.cc', abs_filename) - - # include_dict is modified during iteration, so we iterate over a copy of - # the keys. - header_keys = include_dict.keys() - for header in header_keys: - (same_module, common_path) = FilesBelongToSameModule(abs_filename, header) - fullpath = common_path + header - if same_module and UpdateIncludeState(fullpath, include_dict, io): - header_found = True - - # If we can't find the header file for a .cc, assume it's because we don't - # know where to look. In that case we'll give up as we're not sure they - # didn't include it in the .h file. - # TODO(unknown): Do a better job of finding .h files so we are confident that - # not having the .h file means there isn't one. - if filename.endswith('.cc') and not header_found: - return - - # All the lines have been processed, report the errors found. - for required_header_unstripped in required: - template = required[required_header_unstripped][1] - if required_header_unstripped.strip('<>"') not in include_dict: - error(filename, required[required_header_unstripped][0], - 'build/include_what_you_use', 4, - 'Add #include ' + required_header_unstripped + ' for ' + template) - - -_RE_PATTERN_EXPLICIT_MAKEPAIR = re.compile(r'\bmake_pair\s*<') + Args: + filename: the name of the header to read. + include_dict: a dictionary in which the headers are inserted. + io: The io factory to use to read the file. Provided for testability. + + Returns: + True if a header was successfully added. False otherwise. + """ + headerfile = None + try: + headerfile = io.open(filename, "r", "utf8", "replace") + except IOError: + return False + linenum = 0 + for line in headerfile: + linenum += 1 + clean_line = CleanseComments(line) + match = _RE_PATTERN_INCLUDE.search(clean_line) + if match: + include = match.group(2) + include_dict.setdefault(include, linenum) + return True + + +def CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error, io=codecs): + """Reports for missing stl includes. + + This function will output warnings to make sure you are including the headers + necessary for the stl containers and functions that you use. We only give one + reason to include a header. For example, if you use both equal_to<> and + less<> in a .h file, only one (the latter in the file) of these will be + reported as a reason to include the . + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + include_state: An _IncludeState instance. + error: The function to call with any errors found. + io: The IO factory to use to read the header file. Provided for unittest + injection. + """ + required = {} # A map of header name to linenumber and the template entity. + # Example of required: { '': (1219, 'less<>') } + + for linenum in xrange(clean_lines.NumLines()): + line = clean_lines.elided[linenum] + if not line or line[0] == "#": + continue + + # String is special -- it is a non-templatized type in STL. + matched = _RE_PATTERN_STRING.search(line) + if matched: + # Don't warn about strings in non-STL namespaces: + # (We check only the first match per line; good enough.) + prefix = line[: matched.start()] + if prefix.endswith("std::") or not prefix.endswith("::"): + required[""] = (linenum, "string") + + for pattern, template, header in _re_pattern_headers_maybe_templates: + if pattern.search(line): + required[header] = (linenum, template) + + # The following function is just a speed up, no semantics are changed. + if not "<" in line: # Reduces the cpu time usage by skipping lines. + continue + + for pattern, template, header in _re_pattern_templates: + matched = pattern.search(line) + if matched: + # Don't warn about IWYU in non-STL namespaces: + # (We check only the first match per line; good enough.) + prefix = line[: matched.start()] + if prefix.endswith("std::") or not prefix.endswith("::"): + required[header] = (linenum, template) + + # The policy is that if you #include something in foo.h you don't need to + # include it again in foo.cc. Here, we will look at possible includes. + # Let's flatten the include_state include_list and copy it into a dictionary. + include_dict = dict( + [item for sublist in include_state.include_list for item in sublist] + ) + + # Did we find the header for this file (if any) and successfully load it? + header_found = False + + # Use the absolute path so that matching works properly. + abs_filename = FileInfo(filename).FullName() + + # For Emacs's flymake. + # If cpplint is invoked from Emacs's flymake, a temporary file is generated + # by flymake and that file name might end with '_flymake.cc'. In that case, + # restore original file name here so that the corresponding header file can be + # found. + # e.g. If the file name is 'foo_flymake.cc', we should search for 'foo.h' + # instead of 'foo_flymake.h' + abs_filename = re.sub(r"_flymake\.cc$", ".cc", abs_filename) + + # include_dict is modified during iteration, so we iterate over a copy of + # the keys. + header_keys = include_dict.keys() + for header in header_keys: + (same_module, common_path) = FilesBelongToSameModule(abs_filename, header) + fullpath = common_path + header + if same_module and UpdateIncludeState(fullpath, include_dict, io): + header_found = True + + # If we can't find the header file for a .cc, assume it's because we don't + # know where to look. In that case we'll give up as we're not sure they + # didn't include it in the .h file. + # TODO(unknown): Do a better job of finding .h files so we are confident that + # not having the .h file means there isn't one. + if filename.endswith(".cc") and not header_found: + return + + # All the lines have been processed, report the errors found. + for required_header_unstripped in required: + template = required[required_header_unstripped][1] + if required_header_unstripped.strip('<>"') not in include_dict: + error( + filename, + required[required_header_unstripped][0], + "build/include_what_you_use", + 4, + "Add #include " + required_header_unstripped + " for " + template, + ) + + +_RE_PATTERN_EXPLICIT_MAKEPAIR = re.compile(r"\bmake_pair\s*<") def CheckMakePairUsesDeduction(filename, clean_lines, linenum, error): - """Check that make_pair's template arguments are deduced. - - G++ 4.6 in C++11 mode fails badly if make_pair's template arguments are - specified explicitly, and such use isn't intended in any case. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - match = _RE_PATTERN_EXPLICIT_MAKEPAIR.search(line) - if match: - error(filename, linenum, 'build/explicit_make_pair', - 4, # 4 = high confidence - 'For C++11-compatibility, omit template arguments from make_pair' - ' OR use pair directly OR if appropriate, construct a pair directly') + """Check that make_pair's template arguments are deduced. + G++ 4.6 in C++11 mode fails badly if make_pair's template arguments are + specified explicitly, and such use isn't intended in any case. -def CheckRedundantVirtual(filename, clean_lines, linenum, error): - """Check if line contains a redundant "virtual" function-specifier. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - # Look for "virtual" on current line. - line = clean_lines.elided[linenum] - virtual = Match(r'^(.*)(\bvirtual\b)(.*)$', line) - if not virtual: return - - # Ignore "virtual" keywords that are near access-specifiers. These - # are only used in class base-specifier and do not apply to member - # functions. - if (Search(r'\b(public|protected|private)\s+$', virtual.group(1)) or - Match(r'^\s+(public|protected|private)\b', virtual.group(3))): - return - - # Ignore the "virtual" keyword from virtual base classes. Usually - # there is a column on the same line in these cases (virtual base - # classes are rare in google3 because multiple inheritance is rare). - if Match(r'^.*[^:]:[^:].*$', line): return - - # Look for the next opening parenthesis. This is the start of the - # parameter list (possibly on the next line shortly after virtual). - # TODO(unknown): doesn't work if there are virtual functions with - # decltype() or other things that use parentheses, but csearch suggests - # that this is rare. - end_col = -1 - end_line = -1 - start_col = len(virtual.group(2)) - for start_line in xrange(linenum, min(linenum + 3, clean_lines.NumLines())): - line = clean_lines.elided[start_line][start_col:] - parameter_list = Match(r'^([^(]*)\(', line) - if parameter_list: - # Match parentheses to find the end of the parameter list - (_, end_line, end_col) = CloseExpression( - clean_lines, start_line, start_col + len(parameter_list.group(1))) - break - start_col = 0 - - if end_col < 0: - return # Couldn't find end of parameter list, give up - - # Look for "override" or "final" after the parameter list - # (possibly on the next few lines). - for i in xrange(end_line, min(end_line + 3, clean_lines.NumLines())): - line = clean_lines.elided[i][end_col:] - match = Search(r'\b(override|final)\b', line) + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + match = _RE_PATTERN_EXPLICIT_MAKEPAIR.search(line) if match: - error(filename, linenum, 'readability/inheritance', 4, - ('"virtual" is redundant since function is ' - 'already declared as "%s"' % match.group(1))) + error( + filename, + linenum, + "build/explicit_make_pair", + 4, # 4 = high confidence + "For C++11-compatibility, omit template arguments from make_pair" + " OR use pair directly OR if appropriate, construct a pair directly", + ) - # Set end_col to check whole lines after we are done with the - # first line. - end_col = 0 - if Search(r'[^\w]\s*$', line): - break +def CheckRedundantVirtual(filename, clean_lines, linenum, error): + """Check if line contains a redundant "virtual" function-specifier. -def CheckRedundantOverrideOrFinal(filename, clean_lines, linenum, error): - """Check if line contains a redundant "override" or "final" virt-specifier. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - # Look for closing parenthesis nearby. We need one to confirm where - # the declarator ends and where the virt-specifier starts to avoid - # false positives. - line = clean_lines.elided[linenum] - declarator_end = line.rfind(')') - if declarator_end >= 0: - fragment = line[declarator_end:] - else: - if linenum > 1 and clean_lines.elided[linenum - 1].rfind(')') >= 0: - fragment = line - else: - return + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + # Look for "virtual" on current line. + line = clean_lines.elided[linenum] + virtual = Match(r"^(.*)(\bvirtual\b)(.*)$", line) + if not virtual: + return - # Check that at most one of "override" or "final" is present, not both - if Search(r'\boverride\b', fragment) and Search(r'\bfinal\b', fragment): - error(filename, linenum, 'readability/inheritance', 4, - ('"override" is redundant since function is ' - 'already declared as "final"')) + # Ignore "virtual" keywords that are near access-specifiers. These + # are only used in class base-specifier and do not apply to member + # functions. + if Search(r"\b(public|protected|private)\s+$", virtual.group(1)) or Match( + r"^\s+(public|protected|private)\b", virtual.group(3) + ): + return + # Ignore the "virtual" keyword from virtual base classes. Usually + # there is a column on the same line in these cases (virtual base + # classes are rare in google3 because multiple inheritance is rare). + if Match(r"^.*[^:]:[^:].*$", line): + return + # Look for the next opening parenthesis. This is the start of the + # parameter list (possibly on the next line shortly after virtual). + # TODO(unknown): doesn't work if there are virtual functions with + # decltype() or other things that use parentheses, but csearch suggests + # that this is rare. + end_col = -1 + end_line = -1 + start_col = len(virtual.group(2)) + for start_line in xrange(linenum, min(linenum + 3, clean_lines.NumLines())): + line = clean_lines.elided[start_line][start_col:] + parameter_list = Match(r"^([^(]*)\(", line) + if parameter_list: + # Match parentheses to find the end of the parameter list + (_, end_line, end_col) = CloseExpression( + clean_lines, start_line, start_col + len(parameter_list.group(1)) + ) + break + start_col = 0 + + if end_col < 0: + return # Couldn't find end of parameter list, give up + + # Look for "override" or "final" after the parameter list + # (possibly on the next few lines). + for i in xrange(end_line, min(end_line + 3, clean_lines.NumLines())): + line = clean_lines.elided[i][end_col:] + match = Search(r"\b(override|final)\b", line) + if match: + error( + filename, + linenum, + "readability/inheritance", + 4, + ( + '"virtual" is redundant since function is ' + 'already declared as "%s"' % match.group(1) + ), + ) + + # Set end_col to check whole lines after we are done with the + # first line. + end_col = 0 + if Search(r"[^\w]\s*$", line): + break + + +def CheckRedundantOverrideOrFinal(filename, clean_lines, linenum, error): + """Check if line contains a redundant "override" or "final" virt-specifier. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + # Look for closing parenthesis nearby. We need one to confirm where + # the declarator ends and where the virt-specifier starts to avoid + # false positives. + line = clean_lines.elided[linenum] + declarator_end = line.rfind(")") + if declarator_end >= 0: + fragment = line[declarator_end:] + else: + if linenum > 1 and clean_lines.elided[linenum - 1].rfind(")") >= 0: + fragment = line + else: + return + + # Check that at most one of "override" or "final" is present, not both + if Search(r"\boverride\b", fragment) and Search(r"\bfinal\b", fragment): + error( + filename, + linenum, + "readability/inheritance", + 4, + ( + '"override" is redundant since function is ' + 'already declared as "final"' + ), + ) # Returns true if we are at a new block, and it is directly # inside of a namespace. def IsBlockInNameSpace(nesting_state, is_forward_declaration): - """Checks that the new block is directly in a namespace. - - Args: - nesting_state: The _NestingState object that contains info about our state. - is_forward_declaration: If the class is a forward declared class. - Returns: - Whether or not the new block is directly in a namespace. - """ - if is_forward_declaration: - if len(nesting_state.stack) >= 1 and ( - isinstance(nesting_state.stack[-1], _NamespaceInfo)): - return True - else: - return False + """Checks that the new block is directly in a namespace. - return (len(nesting_state.stack) > 1 and - nesting_state.stack[-1].check_namespace_indentation and - isinstance(nesting_state.stack[-2], _NamespaceInfo)) + Args: + nesting_state: The _NestingState object that contains info about our state. + is_forward_declaration: If the class is a forward declared class. + Returns: + Whether or not the new block is directly in a namespace. + """ + if is_forward_declaration: + if len(nesting_state.stack) >= 1 and ( + isinstance(nesting_state.stack[-1], _NamespaceInfo) + ): + return True + else: + return False + return ( + len(nesting_state.stack) > 1 + and nesting_state.stack[-1].check_namespace_indentation + and isinstance(nesting_state.stack[-2], _NamespaceInfo) + ) -def ShouldCheckNamespaceIndentation(nesting_state, is_namespace_indent_item, - raw_lines_no_comments, linenum): - """This method determines if we should apply our namespace indentation check. - Args: - nesting_state: The current nesting state. - is_namespace_indent_item: If we just put a new class on the stack, True. - If the top of the stack is not a class, or we did not recently - add the class, False. - raw_lines_no_comments: The lines without the comments. - linenum: The current line number we are processing. +def ShouldCheckNamespaceIndentation( + nesting_state, is_namespace_indent_item, raw_lines_no_comments, linenum +): + """This method determines if we should apply our namespace indentation check. - Returns: - True if we should apply our namespace indentation check. Currently, it - only works for classes and namespaces inside of a namespace. - """ + Args: + nesting_state: The current nesting state. + is_namespace_indent_item: If we just put a new class on the stack, True. + If the top of the stack is not a class, or we did not recently + add the class, False. + raw_lines_no_comments: The lines without the comments. + linenum: The current line number we are processing. - is_forward_declaration = IsForwardClassDeclaration(raw_lines_no_comments, - linenum) + Returns: + True if we should apply our namespace indentation check. Currently, it + only works for classes and namespaces inside of a namespace. + """ - if not (is_namespace_indent_item or is_forward_declaration): - return False + is_forward_declaration = IsForwardClassDeclaration(raw_lines_no_comments, linenum) - # If we are in a macro, we do not want to check the namespace indentation. - if IsMacroDefinition(raw_lines_no_comments, linenum): - return False + if not (is_namespace_indent_item or is_forward_declaration): + return False + + # If we are in a macro, we do not want to check the namespace indentation. + if IsMacroDefinition(raw_lines_no_comments, linenum): + return False - return IsBlockInNameSpace(nesting_state, is_forward_declaration) + return IsBlockInNameSpace(nesting_state, is_forward_declaration) # Call this method if the line is directly inside of a namespace. # If the line above is blank (excluding comments) or the start of # an inner namespace, it cannot be indented. -def CheckItemIndentationInNamespace(filename, raw_lines_no_comments, linenum, - error): - line = raw_lines_no_comments[linenum] - if Match(r'^\s+', line): - error(filename, linenum, 'runtime/indentation_namespace', 4, - 'Do not indent within a namespace') - - -def ProcessLine(filename, file_extension, clean_lines, line, - include_state, function_state, nesting_state, error, - extra_check_functions=[]): - """Processes a single line in the file. - - Args: - filename: Filename of the file that is being processed. - file_extension: The extension (dot not included) of the file. - clean_lines: An array of strings, each representing a line of the file, - with comments stripped. - line: Number of line being processed. - include_state: An _IncludeState instance in which the headers are inserted. - function_state: A _FunctionState instance which counts function lines, etc. - nesting_state: A NestingState instance which maintains information about - the current stack of nested blocks being parsed. - error: A callable to which errors are reported, which takes 4 arguments: - filename, line number, error level, and message - extra_check_functions: An array of additional check functions that will be - run on each source line. Each function takes 4 - arguments: filename, clean_lines, line, error - """ - raw_lines = clean_lines.raw_lines - ParseNolintSuppressions(filename, raw_lines[line], line, error) - nesting_state.Update(filename, clean_lines, line, error) - CheckForNamespaceIndentation(filename, nesting_state, clean_lines, line, - error) - if nesting_state.InAsmBlock(): return - CheckForFunctionLengths(filename, clean_lines, line, function_state, error) - CheckForMultilineCommentsAndStrings(filename, clean_lines, line, error) - CheckStyle(filename, clean_lines, line, file_extension, nesting_state, error) - CheckLanguage(filename, clean_lines, line, file_extension, include_state, - nesting_state, error) - CheckForNonConstReference(filename, clean_lines, line, nesting_state, error) - CheckForNonStandardConstructs(filename, clean_lines, line, - nesting_state, error) - CheckVlogArguments(filename, clean_lines, line, error) - CheckPosixThreading(filename, clean_lines, line, error) - CheckInvalidIncrement(filename, clean_lines, line, error) - CheckMakePairUsesDeduction(filename, clean_lines, line, error) - CheckRedundantVirtual(filename, clean_lines, line, error) - CheckRedundantOverrideOrFinal(filename, clean_lines, line, error) - for check_fn in extra_check_functions: - check_fn(filename, clean_lines, line, error) +def CheckItemIndentationInNamespace(filename, raw_lines_no_comments, linenum, error): + line = raw_lines_no_comments[linenum] + if Match(r"^\s+", line): + error( + filename, + linenum, + "runtime/indentation_namespace", + 4, + "Do not indent within a namespace", + ) + + +def ProcessLine( + filename, + file_extension, + clean_lines, + line, + include_state, + function_state, + nesting_state, + error, + extra_check_functions=[], +): + """Processes a single line in the file. + + Args: + filename: Filename of the file that is being processed. + file_extension: The extension (dot not included) of the file. + clean_lines: An array of strings, each representing a line of the file, + with comments stripped. + line: Number of line being processed. + include_state: An _IncludeState instance in which the headers are inserted. + function_state: A _FunctionState instance which counts function lines, etc. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: A callable to which errors are reported, which takes 4 arguments: + filename, line number, error level, and message + extra_check_functions: An array of additional check functions that will be + run on each source line. Each function takes 4 + arguments: filename, clean_lines, line, error + """ + raw_lines = clean_lines.raw_lines + ParseNolintSuppressions(filename, raw_lines[line], line, error) + nesting_state.Update(filename, clean_lines, line, error) + CheckForNamespaceIndentation(filename, nesting_state, clean_lines, line, error) + if nesting_state.InAsmBlock(): + return + CheckForFunctionLengths(filename, clean_lines, line, function_state, error) + CheckForMultilineCommentsAndStrings(filename, clean_lines, line, error) + CheckStyle(filename, clean_lines, line, file_extension, nesting_state, error) + CheckLanguage( + filename, clean_lines, line, file_extension, include_state, nesting_state, error + ) + CheckForNonConstReference(filename, clean_lines, line, nesting_state, error) + CheckForNonStandardConstructs(filename, clean_lines, line, nesting_state, error) + CheckVlogArguments(filename, clean_lines, line, error) + CheckPosixThreading(filename, clean_lines, line, error) + CheckInvalidIncrement(filename, clean_lines, line, error) + CheckMakePairUsesDeduction(filename, clean_lines, line, error) + CheckRedundantVirtual(filename, clean_lines, line, error) + CheckRedundantOverrideOrFinal(filename, clean_lines, line, error) + for check_fn in extra_check_functions: + check_fn(filename, clean_lines, line, error) + def FlagCxx11Features(filename, clean_lines, linenum, error): - """Flag those c++11 features that we only allow in certain places. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - include = Match(r'\s*#\s*include\s+[<"]([^<"]+)[">]', line) - - # Flag unapproved C++ TR1 headers. - if include and include.group(1).startswith('tr1/'): - error(filename, linenum, 'build/c++tr1', 5, - ('C++ TR1 headers such as <%s> are unapproved.') % include.group(1)) - - # Flag unapproved C++11 headers. - if include and include.group(1) in ('cfenv', - 'condition_variable', - 'fenv.h', - 'future', - 'mutex', - 'thread', - 'chrono', - 'ratio', - 'regex', - 'system_error', - ): - error(filename, linenum, 'build/c++11', 5, - ('<%s> is an unapproved C++11 header.') % include.group(1)) - - # The only place where we need to worry about C++11 keywords and library - # features in preprocessor directives is in macro definitions. - if Match(r'\s*#', line) and not Match(r'\s*#\s*define\b', line): return - - # These are classes and free functions. The classes are always - # mentioned as std::*, but we only catch the free functions if - # they're not found by ADL. They're alphabetical by header. - for top_name in ( - # type_traits - 'alignment_of', - 'aligned_union', - ): - if Search(r'\bstd::%s\b' % top_name, line): - error(filename, linenum, 'build/c++11', 5, - ('std::%s is an unapproved C++11 class or function. Send c-style ' - 'an example of where it would make your code more readable, and ' - 'they may let you use it.') % top_name) + """Flag those c++11 features that we only allow in certain places. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + include = Match(r'\s*#\s*include\s+[<"]([^<"]+)[">]', line) + + # Flag unapproved C++ TR1 headers. + if include and include.group(1).startswith("tr1/"): + error( + filename, + linenum, + "build/c++tr1", + 5, + ("C++ TR1 headers such as <%s> are unapproved.") % include.group(1), + ) + + # Flag unapproved C++11 headers. + if include and include.group(1) in ( + "cfenv", + "condition_variable", + "fenv.h", + "future", + "mutex", + "thread", + "chrono", + "ratio", + "regex", + "system_error", + ): + error( + filename, + linenum, + "build/c++11", + 5, + ("<%s> is an unapproved C++11 header.") % include.group(1), + ) + + # The only place where we need to worry about C++11 keywords and library + # features in preprocessor directives is in macro definitions. + if Match(r"\s*#", line) and not Match(r"\s*#\s*define\b", line): + return + + # These are classes and free functions. The classes are always + # mentioned as std::*, but we only catch the free functions if + # they're not found by ADL. They're alphabetical by header. + for top_name in ( + # type_traits + "alignment_of", + "aligned_union", + ): + if Search(r"\bstd::%s\b" % top_name, line): + error( + filename, + linenum, + "build/c++11", + 5, + ( + "std::%s is an unapproved C++11 class or function. Send c-style " + "an example of where it would make your code more readable, and " + "they may let you use it." + ) + % top_name, + ) def FlagCxx14Features(filename, clean_lines, linenum, error): - """Flag those C++14 features that we restrict. - - Args: - filename: The name of the current file. - clean_lines: A CleansedLines instance containing the file. - linenum: The number of the line to check. - error: The function to call with any errors found. - """ - line = clean_lines.elided[linenum] - - include = Match(r'\s*#\s*include\s+[<"]([^<"]+)[">]', line) - - # Flag unapproved C++14 headers. - if include and include.group(1) in ('scoped_allocator', 'shared_mutex'): - error(filename, linenum, 'build/c++14', 5, - ('<%s> is an unapproved C++14 header.') % include.group(1)) - - -def ProcessFileData(filename, file_extension, lines, error, - extra_check_functions=[]): - """Performs lint checks and reports any errors to the given error function. - - Args: - filename: Filename of the file that is being processed. - file_extension: The extension (dot not included) of the file. - lines: An array of strings, each representing a line of the file, with the - last element being empty if the file is terminated with a newline. - error: A callable to which errors are reported, which takes 4 arguments: - filename, line number, error level, and message - extra_check_functions: An array of additional check functions that will be - run on each source line. Each function takes 4 - arguments: filename, clean_lines, line, error - """ - lines = (['// marker so line numbers and indices both start at 1'] + lines + - ['// marker so line numbers end in a known way']) - - include_state = _IncludeState() - function_state = _FunctionState() - nesting_state = NestingState() - - ResetNolintSuppressions() - - CheckForCopyright(filename, lines, error) - ProcessGlobalSuppresions(lines) - RemoveMultiLineComments(filename, lines, error) - clean_lines = CleansedLines(lines) - - if IsHeaderExtension(file_extension): - CheckForHeaderGuard(filename, clean_lines, error) - - for line in xrange(clean_lines.NumLines()): - ProcessLine(filename, file_extension, clean_lines, line, - include_state, function_state, nesting_state, error, - extra_check_functions) - FlagCxx11Features(filename, clean_lines, line, error) - nesting_state.CheckCompletedBlocks(filename, error) - - CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error) - - # Check that the .cc file has included its header if it exists. - if _IsSourceExtension(file_extension): - CheckHeaderFileIncluded(filename, include_state, error) - - # We check here rather than inside ProcessLine so that we see raw - # lines rather than "cleaned" lines. - CheckForBadCharacters(filename, lines, error) - - CheckForNewlineAtEOF(filename, lines, error) + """Flag those C++14 features that we restrict. -def ProcessConfigOverrides(filename): - """ Loads the configuration files and processes the config overrides. + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] - Args: - filename: The name of the file being processed by the linter. + include = Match(r'\s*#\s*include\s+[<"]([^<"]+)[">]', line) - Returns: - False if the current |filename| should not be processed further. - """ + # Flag unapproved C++14 headers. + if include and include.group(1) in ("scoped_allocator", "shared_mutex"): + error( + filename, + linenum, + "build/c++14", + 5, + ("<%s> is an unapproved C++14 header.") % include.group(1), + ) - abs_filename = os.path.abspath(filename) - cfg_filters = [] - keep_looking = True - while keep_looking: - abs_path, base_name = os.path.split(abs_filename) - if not base_name: - break # Reached the root directory. - cfg_file = os.path.join(abs_path, "CPPLINT.cfg") - abs_filename = abs_path - if not os.path.isfile(cfg_file): - continue +def ProcessFileData(filename, file_extension, lines, error, extra_check_functions=[]): + """Performs lint checks and reports any errors to the given error function. - try: - with open(cfg_file) as file_handle: - for line in file_handle: - line, _, _ = line.partition('#') # Remove comments. - if not line.strip(): + Args: + filename: Filename of the file that is being processed. + file_extension: The extension (dot not included) of the file. + lines: An array of strings, each representing a line of the file, with the + last element being empty if the file is terminated with a newline. + error: A callable to which errors are reported, which takes 4 arguments: + filename, line number, error level, and message + extra_check_functions: An array of additional check functions that will be + run on each source line. Each function takes 4 + arguments: filename, clean_lines, line, error + """ + lines = ( + ["// marker so line numbers and indices both start at 1"] + + lines + + ["// marker so line numbers end in a known way"] + ) + + include_state = _IncludeState() + function_state = _FunctionState() + nesting_state = NestingState() + + ResetNolintSuppressions() + + CheckForCopyright(filename, lines, error) + ProcessGlobalSuppresions(lines) + RemoveMultiLineComments(filename, lines, error) + clean_lines = CleansedLines(lines) + + if IsHeaderExtension(file_extension): + CheckForHeaderGuard(filename, clean_lines, error) + + for line in xrange(clean_lines.NumLines()): + ProcessLine( + filename, + file_extension, + clean_lines, + line, + include_state, + function_state, + nesting_state, + error, + extra_check_functions, + ) + FlagCxx11Features(filename, clean_lines, line, error) + nesting_state.CheckCompletedBlocks(filename, error) + + CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error) + + # Check that the .cc file has included its header if it exists. + if _IsSourceExtension(file_extension): + CheckHeaderFileIncluded(filename, include_state, error) + + # We check here rather than inside ProcessLine so that we see raw + # lines rather than "cleaned" lines. + CheckForBadCharacters(filename, lines, error) + + CheckForNewlineAtEOF(filename, lines, error) + + +def ProcessConfigOverrides(filename): + """Loads the configuration files and processes the config overrides. + + Args: + filename: The name of the file being processed by the linter. + + Returns: + False if the current |filename| should not be processed further. + """ + + abs_filename = os.path.abspath(filename) + cfg_filters = [] + keep_looking = True + while keep_looking: + abs_path, base_name = os.path.split(abs_filename) + if not base_name: + break # Reached the root directory. + + cfg_file = os.path.join(abs_path, "CPPLINT.cfg") + abs_filename = abs_path + if not os.path.isfile(cfg_file): continue - name, _, val = line.partition('=') - name = name.strip() - val = val.strip() - if name == 'set noparent': - keep_looking = False - elif name == 'filter': - cfg_filters.append(val) - elif name == 'exclude_files': - # When matching exclude_files pattern, use the base_name of - # the current file name or the directory name we are processing. - # For example, if we are checking for lint errors in /foo/bar/baz.cc - # and we found the .cfg file at /foo/CPPLINT.cfg, then the config - # file's "exclude_files" filter is meant to be checked against "bar" - # and not "baz" nor "bar/baz.cc". - if base_name: - pattern = re.compile(val) - if pattern.match(base_name): - if _cpplint_state.quiet: - # Suppress "Ignoring file" warning when using --quiet. - return False - sys.stderr.write('Ignoring "%s": file excluded by "%s". ' - 'File path component "%s" matches ' - 'pattern "%s"\n' % - (filename, cfg_file, base_name, val)) - return False - elif name == 'linelength': - global _line_length - try: - _line_length = int(val) - except ValueError: - sys.stderr.write('Line length must be numeric.') - elif name == 'root': - global _root - # root directories are specified relative to CPPLINT.cfg dir. - _root = os.path.join(os.path.dirname(cfg_file), val) - elif name == 'headers': - ProcessHppHeadersOption(val) - else: + try: + with open(cfg_file) as file_handle: + for line in file_handle: + line, _, _ = line.partition("#") # Remove comments. + if not line.strip(): + continue + + name, _, val = line.partition("=") + name = name.strip() + val = val.strip() + if name == "set noparent": + keep_looking = False + elif name == "filter": + cfg_filters.append(val) + elif name == "exclude_files": + # When matching exclude_files pattern, use the base_name of + # the current file name or the directory name we are processing. + # For example, if we are checking for lint errors in /foo/bar/baz.cc + # and we found the .cfg file at /foo/CPPLINT.cfg, then the config + # file's "exclude_files" filter is meant to be checked against "bar" + # and not "baz" nor "bar/baz.cc". + if base_name: + pattern = re.compile(val) + if pattern.match(base_name): + if _cpplint_state.quiet: + # Suppress "Ignoring file" warning when using --quiet. + return False + sys.stderr.write( + 'Ignoring "%s": file excluded by "%s". ' + 'File path component "%s" matches ' + 'pattern "%s"\n' + % (filename, cfg_file, base_name, val) + ) + return False + elif name == "linelength": + global _line_length + try: + _line_length = int(val) + except ValueError: + sys.stderr.write("Line length must be numeric.") + elif name == "root": + global _root + # root directories are specified relative to CPPLINT.cfg dir. + _root = os.path.join(os.path.dirname(cfg_file), val) + elif name == "headers": + ProcessHppHeadersOption(val) + else: + sys.stderr.write( + "Invalid configuration option (%s) in file %s\n" + % (name, cfg_file) + ) + + except IOError: sys.stderr.write( - 'Invalid configuration option (%s) in file %s\n' % - (name, cfg_file)) - - except IOError: - sys.stderr.write( - "Skipping config file '%s': Can't open for reading\n" % cfg_file) - keep_looking = False + "Skipping config file '%s': Can't open for reading\n" % cfg_file + ) + keep_looking = False - # Apply all the accumulated filters in reverse order (top-level directory - # config options having the least priority). - for filter in reversed(cfg_filters): - _AddFilters(filter) + # Apply all the accumulated filters in reverse order (top-level directory + # config options having the least priority). + for filter in reversed(cfg_filters): + _AddFilters(filter) - return True + return True def ProcessFile(filename, vlevel, extra_check_functions=[]): - """Does google-lint on a single file. + """Does google-lint on a single file. - Args: - filename: The name of the file to parse. + Args: + filename: The name of the file to parse. - vlevel: The level of errors to report. Every error of confidence - >= verbose_level will be reported. 0 is a good default. + vlevel: The level of errors to report. Every error of confidence + >= verbose_level will be reported. 0 is a good default. - extra_check_functions: An array of additional check functions that will be - run on each source line. Each function takes 4 - arguments: filename, clean_lines, line, error - """ + extra_check_functions: An array of additional check functions that will be + run on each source line. Each function takes 4 + arguments: filename, clean_lines, line, error + """ - _SetVerboseLevel(vlevel) - _BackupFilters() - old_errors = _cpplint_state.error_count + _SetVerboseLevel(vlevel) + _BackupFilters() + old_errors = _cpplint_state.error_count - if not ProcessConfigOverrides(filename): - _RestoreFilters() - return - - lf_lines = [] - crlf_lines = [] - try: - # Support the UNIX convention of using "-" for stdin. Note that - # we are not opening the file with universal newline support - # (which codecs doesn't support anyway), so the resulting lines do - # contain trailing '\r' characters if we are reading a file that - # has CRLF endings. - # If after the split a trailing '\r' is present, it is removed - # below. - if filename == '-': - lines = codecs.StreamReaderWriter(sys.stdin, - codecs.getreader('utf8'), - codecs.getwriter('utf8'), - 'replace').read().split('\n') + if not ProcessConfigOverrides(filename): + _RestoreFilters() + return + + lf_lines = [] + crlf_lines = [] + try: + # Support the UNIX convention of using "-" for stdin. Note that + # we are not opening the file with universal newline support + # (which codecs doesn't support anyway), so the resulting lines do + # contain trailing '\r' characters if we are reading a file that + # has CRLF endings. + # If after the split a trailing '\r' is present, it is removed + # below. + if filename == "-": + lines = ( + codecs.StreamReaderWriter( + sys.stdin, + codecs.getreader("utf8"), + codecs.getwriter("utf8"), + "replace", + ) + .read() + .split("\n") + ) + else: + lines = codecs.open(filename, "r", "utf8", "replace").read().split("\n") + + # Remove trailing '\r'. + # The -1 accounts for the extra trailing blank line we get from split() + for linenum in range(len(lines) - 1): + if lines[linenum].endswith("\r"): + lines[linenum] = lines[linenum].rstrip("\r") + crlf_lines.append(linenum + 1) + else: + lf_lines.append(linenum + 1) + + except IOError: + sys.stderr.write("Skipping input '%s': Can't open for reading\n" % filename) + _RestoreFilters() + return + + # Note, if no dot is found, this will give the entire filename as the ext. + file_extension = filename[filename.rfind(".") + 1 :] + + # When reading from stdin, the extension is unknown, so no cpplint tests + # should rely on the extension. + if filename != "-" and file_extension not in _valid_extensions: + sys.stderr.write( + "Ignoring %s; not a valid file name " + "(%s)\n" % (filename, ", ".join(_valid_extensions)) + ) else: - lines = codecs.open(filename, 'r', 'utf8', 'replace').read().split('\n') - - # Remove trailing '\r'. - # The -1 accounts for the extra trailing blank line we get from split() - for linenum in range(len(lines) - 1): - if lines[linenum].endswith('\r'): - lines[linenum] = lines[linenum].rstrip('\r') - crlf_lines.append(linenum + 1) - else: - lf_lines.append(linenum + 1) - - except IOError: - sys.stderr.write( - "Skipping input '%s': Can't open for reading\n" % filename) + ProcessFileData(filename, file_extension, lines, Error, extra_check_functions) + + # If end-of-line sequences are a mix of LF and CR-LF, issue + # warnings on the lines with CR. + # + # Don't issue any warnings if all lines are uniformly LF or CR-LF, + # since critique can handle these just fine, and the style guide + # doesn't dictate a particular end of line sequence. + # + # We can't depend on os.linesep to determine what the desired + # end-of-line sequence should be, since that will return the + # server-side end-of-line sequence. + if lf_lines and crlf_lines: + # Warn on every line with CR. An alternative approach might be to + # check whether the file is mostly CRLF or just LF, and warn on the + # minority, we bias toward LF here since most tools prefer LF. + for linenum in crlf_lines: + Error( + filename, + linenum, + "whitespace/newline", + 1, + "Unexpected \\r (^M) found; better to use only \\n", + ) + + # Suppress printing anything if --quiet was passed unless the error + # count has increased after processing this file. + if not _cpplint_state.quiet or old_errors != _cpplint_state.error_count: + sys.stdout.write("Done processing %s\n" % filename) _RestoreFilters() - return - - # Note, if no dot is found, this will give the entire filename as the ext. - file_extension = filename[filename.rfind('.') + 1:] - - # When reading from stdin, the extension is unknown, so no cpplint tests - # should rely on the extension. - if filename != '-' and file_extension not in _valid_extensions: - sys.stderr.write('Ignoring %s; not a valid file name ' - '(%s)\n' % (filename, ', '.join(_valid_extensions))) - else: - ProcessFileData(filename, file_extension, lines, Error, - extra_check_functions) - - # If end-of-line sequences are a mix of LF and CR-LF, issue - # warnings on the lines with CR. - # - # Don't issue any warnings if all lines are uniformly LF or CR-LF, - # since critique can handle these just fine, and the style guide - # doesn't dictate a particular end of line sequence. - # - # We can't depend on os.linesep to determine what the desired - # end-of-line sequence should be, since that will return the - # server-side end-of-line sequence. - if lf_lines and crlf_lines: - # Warn on every line with CR. An alternative approach might be to - # check whether the file is mostly CRLF or just LF, and warn on the - # minority, we bias toward LF here since most tools prefer LF. - for linenum in crlf_lines: - Error(filename, linenum, 'whitespace/newline', 1, - 'Unexpected \\r (^M) found; better to use only \\n') - - # Suppress printing anything if --quiet was passed unless the error - # count has increased after processing this file. - if not _cpplint_state.quiet or old_errors != _cpplint_state.error_count: - sys.stdout.write('Done processing %s\n' % filename) - _RestoreFilters() def PrintUsage(message): - """Prints a brief usage string and exits, optionally with an error message. + """Prints a brief usage string and exits, optionally with an error message. - Args: - message: The optional error message. - """ - sys.stderr.write(_USAGE) - if message: - sys.exit('\nFATAL ERROR: ' + message) - else: - sys.exit(1) + Args: + message: The optional error message. + """ + sys.stderr.write(_USAGE) + if message: + sys.exit("\nFATAL ERROR: " + message) + else: + sys.exit(1) def PrintCategories(): - """Prints a list of all the error-categories used by error messages. + """Prints a list of all the error-categories used by error messages. - These are the categories used to filter messages via --filter. - """ - sys.stderr.write(''.join(' %s\n' % cat for cat in _ERROR_CATEGORIES)) - sys.exit(0) + These are the categories used to filter messages via --filter. + """ + sys.stderr.write("".join(" %s\n" % cat for cat in _ERROR_CATEGORIES)) + sys.exit(0) def ParseArguments(args): - """Parses the command line arguments. - - This may set the output format and verbosity level as side-effects. - - Args: - args: The command line arguments: - - Returns: - The list of filenames to lint. - """ - try: - (opts, filenames) = getopt.getopt(args, '', ['help', 'output=', 'verbose=', - 'counting=', - 'filter=', - 'root=', - 'linelength=', - 'extensions=', - 'headers=', - 'quiet']) - except getopt.GetoptError: - PrintUsage('Invalid arguments.') - - verbosity = _VerboseLevel() - output_format = _OutputFormat() - filters = '' - quiet = _Quiet() - counting_style = '' - - for (opt, val) in opts: - if opt == '--help': - PrintUsage(None) - elif opt == '--output': - if val not in ('emacs', 'vs7', 'eclipse'): - PrintUsage('The only allowed output formats are emacs, vs7 and eclipse.') - output_format = val - elif opt == '--quiet': - quiet = True - elif opt == '--verbose': - verbosity = int(val) - elif opt == '--filter': - filters = val - if not filters: - PrintCategories() - elif opt == '--counting': - if val not in ('total', 'toplevel', 'detailed'): - PrintUsage('Valid counting options are total, toplevel, and detailed') - counting_style = val - elif opt == '--root': - global _root - _root = val - elif opt == '--linelength': - global _line_length - try: - _line_length = int(val) - except ValueError: - PrintUsage('Line length must be digits.') - elif opt == '--extensions': - global _valid_extensions - try: - _valid_extensions = set(val.split(',')) - except ValueError: - PrintUsage('Extensions must be comma separated list.') - elif opt == '--headers': - ProcessHppHeadersOption(val) - - if not filenames: - PrintUsage('No files were specified.') - - _SetOutputFormat(output_format) - _SetQuiet(quiet) - _SetVerboseLevel(verbosity) - _SetFilters(filters) - _SetCountingStyle(counting_style) - - return filenames + """Parses the command line arguments. + + This may set the output format and verbosity level as side-effects. + + Args: + args: The command line arguments: + + Returns: + The list of filenames to lint. + """ + try: + (opts, filenames) = getopt.getopt( + args, + "", + [ + "help", + "output=", + "verbose=", + "counting=", + "filter=", + "root=", + "linelength=", + "extensions=", + "headers=", + "quiet", + ], + ) + except getopt.GetoptError: + PrintUsage("Invalid arguments.") + + verbosity = _VerboseLevel() + output_format = _OutputFormat() + filters = "" + quiet = _Quiet() + counting_style = "" + + for opt, val in opts: + if opt == "--help": + PrintUsage(None) + elif opt == "--output": + if val not in ("emacs", "vs7", "eclipse"): + PrintUsage( + "The only allowed output formats are emacs, vs7 and eclipse." + ) + output_format = val + elif opt == "--quiet": + quiet = True + elif opt == "--verbose": + verbosity = int(val) + elif opt == "--filter": + filters = val + if not filters: + PrintCategories() + elif opt == "--counting": + if val not in ("total", "toplevel", "detailed"): + PrintUsage("Valid counting options are total, toplevel, and detailed") + counting_style = val + elif opt == "--root": + global _root + _root = val + elif opt == "--linelength": + global _line_length + try: + _line_length = int(val) + except ValueError: + PrintUsage("Line length must be digits.") + elif opt == "--extensions": + global _valid_extensions + try: + _valid_extensions = set(val.split(",")) + except ValueError: + PrintUsage("Extensions must be comma separated list.") + elif opt == "--headers": + ProcessHppHeadersOption(val) + + if not filenames: + PrintUsage("No files were specified.") + + _SetOutputFormat(output_format) + _SetQuiet(quiet) + _SetVerboseLevel(verbosity) + _SetFilters(filters) + _SetCountingStyle(counting_style) + + return filenames def main(): - filenames = ParseArguments(sys.argv[1:]) + filenames = ParseArguments(sys.argv[1:]) - # Change stderr to write with replacement characters so we don't die - # if we try to print something containing non-ASCII characters. - sys.stderr = codecs.StreamReaderWriter(sys.stderr, - codecs.getreader('utf8'), - codecs.getwriter('utf8'), - 'replace') + # Change stderr to write with replacement characters so we don't die + # if we try to print something containing non-ASCII characters. + sys.stderr = codecs.StreamReaderWriter( + sys.stderr, codecs.getreader("utf8"), codecs.getwriter("utf8"), "replace" + ) - _cpplint_state.ResetErrorCounts() - for filename in filenames: - ProcessFile(filename, _cpplint_state.verbose_level) - # If --quiet is passed, suppress printing error count unless there are errors. - if not _cpplint_state.quiet or _cpplint_state.error_count > 0: - _cpplint_state.PrintErrorCounts() + _cpplint_state.ResetErrorCounts() + for filename in filenames: + ProcessFile(filename, _cpplint_state.verbose_level) + # If --quiet is passed, suppress printing error count unless there are errors. + if not _cpplint_state.quiet or _cpplint_state.error_count > 0: + _cpplint_state.PrintErrorCounts() - sys.exit(_cpplint_state.error_count > 0) + sys.exit(_cpplint_state.error_count > 0) -if __name__ == '__main__': - main() +if __name__ == "__main__": + main() diff --git a/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/gen_dynamic_list.py b/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/gen_dynamic_list.py index 6585a42d..8c0a1ea8 100755 --- a/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/gen_dynamic_list.py +++ b/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/gen_dynamic_list.py @@ -1,18 +1,18 @@ #!/usr/bin/env python -#===- lib/sanitizer_common/scripts/gen_dynamic_list.py ---------------------===# +# ===- lib/sanitizer_common/scripts/gen_dynamic_list.py ---------------------===# # # Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. # See https://llvm.org/LICENSE.txt for license information. # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception # -#===------------------------------------------------------------------------===# +# ===------------------------------------------------------------------------===# # # Generates the list of functions that should be exported from sanitizer # runtimes. The output format is recognized by --dynamic-list linker option. # Usage: # gen_dynamic_list.py libclang_rt.*san*.a [ files ... ] # -#===------------------------------------------------------------------------===# +# ===------------------------------------------------------------------------===# from __future__ import print_function import argparse import os @@ -21,115 +21,143 @@ import sys import platform -new_delete = set([ - '_Znam', '_ZnamRKSt9nothrow_t', # operator new[](unsigned long) - '_Znwm', '_ZnwmRKSt9nothrow_t', # operator new(unsigned long) - '_Znaj', '_ZnajRKSt9nothrow_t', # operator new[](unsigned int) - '_Znwj', '_ZnwjRKSt9nothrow_t', # operator new(unsigned int) - # operator new(unsigned long, std::align_val_t) - '_ZnwmSt11align_val_t', '_ZnwmSt11align_val_tRKSt9nothrow_t', - # operator new(unsigned int, std::align_val_t) - '_ZnwjSt11align_val_t', '_ZnwjSt11align_val_tRKSt9nothrow_t', - # operator new[](unsigned long, std::align_val_t) - '_ZnamSt11align_val_t', '_ZnamSt11align_val_tRKSt9nothrow_t', - # operator new[](unsigned int, std::align_val_t) - '_ZnajSt11align_val_t', '_ZnajSt11align_val_tRKSt9nothrow_t', - '_ZdaPv', '_ZdaPvRKSt9nothrow_t', # operator delete[](void *) - '_ZdlPv', '_ZdlPvRKSt9nothrow_t', # operator delete(void *) - '_ZdaPvm', # operator delete[](void*, unsigned long) - '_ZdlPvm', # operator delete(void*, unsigned long) - '_ZdaPvj', # operator delete[](void*, unsigned int) - '_ZdlPvj', # operator delete(void*, unsigned int) - # operator delete(void*, std::align_val_t) - '_ZdlPvSt11align_val_t', '_ZdlPvSt11align_val_tRKSt9nothrow_t', - # operator delete[](void*, std::align_val_t) - '_ZdaPvSt11align_val_t', '_ZdaPvSt11align_val_tRKSt9nothrow_t', - # operator delete(void*, unsigned long, std::align_val_t) - '_ZdlPvmSt11align_val_t', - # operator delete[](void*, unsigned long, std::align_val_t) - '_ZdaPvmSt11align_val_t', - # operator delete(void*, unsigned int, std::align_val_t) - '_ZdlPvjSt11align_val_t', - # operator delete[](void*, unsigned int, std::align_val_t) - '_ZdaPvjSt11align_val_t', - ]) +new_delete = set( + [ + "_Znam", + "_ZnamRKSt9nothrow_t", # operator new[](unsigned long) + "_Znwm", + "_ZnwmRKSt9nothrow_t", # operator new(unsigned long) + "_Znaj", + "_ZnajRKSt9nothrow_t", # operator new[](unsigned int) + "_Znwj", + "_ZnwjRKSt9nothrow_t", # operator new(unsigned int) + # operator new(unsigned long, std::align_val_t) + "_ZnwmSt11align_val_t", + "_ZnwmSt11align_val_tRKSt9nothrow_t", + # operator new(unsigned int, std::align_val_t) + "_ZnwjSt11align_val_t", + "_ZnwjSt11align_val_tRKSt9nothrow_t", + # operator new[](unsigned long, std::align_val_t) + "_ZnamSt11align_val_t", + "_ZnamSt11align_val_tRKSt9nothrow_t", + # operator new[](unsigned int, std::align_val_t) + "_ZnajSt11align_val_t", + "_ZnajSt11align_val_tRKSt9nothrow_t", + "_ZdaPv", + "_ZdaPvRKSt9nothrow_t", # operator delete[](void *) + "_ZdlPv", + "_ZdlPvRKSt9nothrow_t", # operator delete(void *) + "_ZdaPvm", # operator delete[](void*, unsigned long) + "_ZdlPvm", # operator delete(void*, unsigned long) + "_ZdaPvj", # operator delete[](void*, unsigned int) + "_ZdlPvj", # operator delete(void*, unsigned int) + # operator delete(void*, std::align_val_t) + "_ZdlPvSt11align_val_t", + "_ZdlPvSt11align_val_tRKSt9nothrow_t", + # operator delete[](void*, std::align_val_t) + "_ZdaPvSt11align_val_t", + "_ZdaPvSt11align_val_tRKSt9nothrow_t", + # operator delete(void*, unsigned long, std::align_val_t) + "_ZdlPvmSt11align_val_t", + # operator delete[](void*, unsigned long, std::align_val_t) + "_ZdaPvmSt11align_val_t", + # operator delete(void*, unsigned int, std::align_val_t) + "_ZdlPvjSt11align_val_t", + # operator delete[](void*, unsigned int, std::align_val_t) + "_ZdaPvjSt11align_val_t", + ] +) + +versioned_functions = set( + [ + "memcpy", + "pthread_attr_getaffinity_np", + "pthread_cond_broadcast", + "pthread_cond_destroy", + "pthread_cond_init", + "pthread_cond_signal", + "pthread_cond_timedwait", + "pthread_cond_wait", + "realpath", + "sched_getaffinity", + ] +) -versioned_functions = set(['memcpy', 'pthread_attr_getaffinity_np', - 'pthread_cond_broadcast', - 'pthread_cond_destroy', 'pthread_cond_init', - 'pthread_cond_signal', 'pthread_cond_timedwait', - 'pthread_cond_wait', 'realpath', - 'sched_getaffinity']) def get_global_functions(nm_executable, library): - functions = [] - nm = os.environ.get('NM', nm_executable) - nm_proc = subprocess.Popen([nm, library], stdout=subprocess.PIPE, - stderr=subprocess.PIPE) - nm_out = nm_proc.communicate()[0].decode().split('\n') - if nm_proc.returncode != 0: - raise subprocess.CalledProcessError(nm_proc.returncode, nm) - func_symbols = ['T', 'W'] - # On PowerPC, nm prints function descriptors from .data section. - if platform.uname()[4] in ["powerpc", "ppc64"]: - func_symbols += ['D'] - for line in nm_out: - cols = line.split(' ') - if len(cols) == 3 and cols[1] in func_symbols : - functions.append(cols[2]) - return functions + functions = [] + nm = os.environ.get("NM", nm_executable) + nm_proc = subprocess.Popen( + [nm, library], stdout=subprocess.PIPE, stderr=subprocess.PIPE + ) + nm_out = nm_proc.communicate()[0].decode().split("\n") + if nm_proc.returncode != 0: + raise subprocess.CalledProcessError(nm_proc.returncode, nm) + func_symbols = ["T", "W"] + # On PowerPC, nm prints function descriptors from .data section. + if platform.uname()[4] in ["powerpc", "ppc64"]: + func_symbols += ["D"] + for line in nm_out: + cols = line.split(" ") + if len(cols) == 3 and cols[1] in func_symbols: + functions.append(cols[2]) + return functions + def main(argv): - parser = argparse.ArgumentParser() - parser.add_argument('--version-list', action='store_true') - parser.add_argument('--extra', default=[], action='append') - parser.add_argument('libraries', default=[], nargs='+') - parser.add_argument('--nm-executable', required=True) - parser.add_argument('-o', '--output', required=True) - args = parser.parse_args() + parser = argparse.ArgumentParser() + parser.add_argument("--version-list", action="store_true") + parser.add_argument("--extra", default=[], action="append") + parser.add_argument("libraries", default=[], nargs="+") + parser.add_argument("--nm-executable", required=True) + parser.add_argument("-o", "--output", required=True) + args = parser.parse_args() + + result = [] - result = [] + all_functions = [] + for library in args.libraries: + all_functions.extend(get_global_functions(args.nm_executable, library)) + function_set = set(all_functions) + for func in all_functions: + # Export new/delete operators. + if func in new_delete: + result.append(func) + continue + # Export interceptors. + match = re.match("__interceptor_(.*)", func) + if match: + result.append(func) + # We have to avoid exporting the interceptors for versioned library + # functions due to gold internal error. + orig_name = match.group(1) + if orig_name in function_set and ( + args.version_list or orig_name not in versioned_functions + ): + result.append(orig_name) + continue + # Export sanitizer interface functions. + if re.match("__sanitizer_(.*)", func): + result.append(func) - all_functions = [] - for library in args.libraries: - all_functions.extend(get_global_functions(args.nm_executable, library)) - function_set = set(all_functions) - for func in all_functions: - # Export new/delete operators. - if func in new_delete: - result.append(func) - continue - # Export interceptors. - match = re.match('__interceptor_(.*)', func) - if match: - result.append(func) - # We have to avoid exporting the interceptors for versioned library - # functions due to gold internal error. - orig_name = match.group(1) - if orig_name in function_set and (args.version_list or orig_name not in versioned_functions): - result.append(orig_name) - continue - # Export sanitizer interface functions. - if re.match('__sanitizer_(.*)', func): - result.append(func) + # Additional exported functions from files. + for fname in args.extra: + f = open(fname, "r") + for line in f: + result.append(line.rstrip()) + # Print the resulting list in the format recognized by ld. + with open(args.output, "w") as f: + print("{", file=f) + if args.version_list: + print("global:", file=f) + result.sort() + for sym in result: + print(" %s;" % sym, file=f) + if args.version_list: + print("local:", file=f) + print(" *;", file=f) + print("};", file=f) - # Additional exported functions from files. - for fname in args.extra: - f = open(fname, 'r') - for line in f: - result.append(line.rstrip()) - # Print the resulting list in the format recognized by ld. - with open(args.output, 'w') as f: - print('{', file=f) - if args.version_list: - print('global:', file=f) - result.sort() - for sym in result: - print(u' %s;' % sym, file=f) - if args.version_list: - print('local:', file=f) - print(' *;', file=f) - print('};', file=f) -if __name__ == '__main__': - main(sys.argv) +if __name__ == "__main__": + main(sys.argv) diff --git a/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/litlint.py b/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/litlint.py index da6ef4ba..8e6db83d 100755 --- a/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/litlint.py +++ b/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/litlint.py @@ -13,61 +13,62 @@ from io import open # Compile regex once for all files -runRegex = re.compile(r'(? 0: - sys.exit(1) + # Parse args + parser = optparse.OptionParser() + parser.add_option("--filter") # ignored + (options, filenames) = parser.parse_args() + + # Lint each file + errs = 0 + for p in filenames: + errs += LintFile(p) + + # If errors, return nonzero + if errs > 0: + sys.exit(1) diff --git a/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/litlint_test.py b/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/litlint_test.py index 30c9f16e..2862fc30 100755 --- a/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/litlint_test.py +++ b/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/litlint_test.py @@ -9,15 +9,17 @@ import litlint import unittest + class TestLintLine(unittest.TestCase): - def test_missing_run(self): - f = litlint.LintLine - self.assertEqual(f(' %t '), ('missing %run before %t', 2)) - self.assertEqual(f(' %t\n'), ('missing %run before %t', 2)) - self.assertEqual(f(' %t.so '), (None, None)) - self.assertEqual(f(' %t.o '), (None, None)) - self.assertEqual(f('%run %t '), (None, None)) - self.assertEqual(f('-o %t '), (None, None)) + def test_missing_run(self): + f = litlint.LintLine + self.assertEqual(f(" %t "), ("missing %run before %t", 2)) + self.assertEqual(f(" %t\n"), ("missing %run before %t", 2)) + self.assertEqual(f(" %t.so "), (None, None)) + self.assertEqual(f(" %t.o "), (None, None)) + self.assertEqual(f("%run %t "), (None, None)) + self.assertEqual(f("-o %t "), (None, None)) + -if __name__ == '__main__': - unittest.main() +if __name__ == "__main__": + unittest.main() diff --git a/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/sancov.py b/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/sancov.py index 759eb0cb..5b920a7f 100755 --- a/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/sancov.py +++ b/polytracker/src/compiler-rt/lib/sanitizer_common/scripts/sancov.py @@ -13,237 +13,276 @@ prog_name = "" + def Usage(): - sys.stderr.write( - "Usage: \n" + \ - " " + prog_name + " merge FILE [FILE...] > OUTPUT\n" \ - " " + prog_name + " print FILE [FILE...]\n" \ - " " + prog_name + " unpack FILE [FILE...]\n" \ - " " + prog_name + " rawunpack FILE [FILE ...]\n" \ - " " + prog_name + " missing BINARY < LIST_OF_PCS\n" \ - "\n") - exit(1) + sys.stderr.write( + "Usage: \n" + " " + prog_name + " merge FILE [FILE...] > OUTPUT\n" + " " + prog_name + " print FILE [FILE...]\n" + " " + prog_name + " unpack FILE [FILE...]\n" + " " + prog_name + " rawunpack FILE [FILE ...]\n" + " " + prog_name + " missing BINARY < LIST_OF_PCS\n" + "\n" + ) + exit(1) + def CheckBits(bits): - if bits != 32 and bits != 64: - raise Exception("Wrong bitness: %d" % bits) + if bits != 32 and bits != 64: + raise Exception("Wrong bitness: %d" % bits) + def TypeCodeForBits(bits): - CheckBits(bits) - return 'L' if bits == 64 else 'I' + CheckBits(bits) + return "L" if bits == 64 else "I" + def TypeCodeForStruct(bits): - CheckBits(bits) - return 'Q' if bits == 64 else 'I' + CheckBits(bits) + return "Q" if bits == 64 else "I" + + +kMagic32SecondHalf = 0xFFFFFF32 +kMagic64SecondHalf = 0xFFFFFF64 +kMagicFirstHalf = 0xC0BFFFFF -kMagic32SecondHalf = 0xFFFFFF32; -kMagic64SecondHalf = 0xFFFFFF64; -kMagicFirstHalf = 0xC0BFFFFF; def MagicForBits(bits): - CheckBits(bits) - if sys.byteorder == 'little': - return [kMagic64SecondHalf if bits == 64 else kMagic32SecondHalf, kMagicFirstHalf] - else: - return [kMagicFirstHalf, kMagic64SecondHalf if bits == 64 else kMagic32SecondHalf] + CheckBits(bits) + if sys.byteorder == "little": + return [ + kMagic64SecondHalf if bits == 64 else kMagic32SecondHalf, + kMagicFirstHalf, + ] + else: + return [ + kMagicFirstHalf, + kMagic64SecondHalf if bits == 64 else kMagic32SecondHalf, + ] + def ReadMagicAndReturnBitness(f, path): - magic_bytes = f.read(8) - magic_words = struct.unpack('II', magic_bytes); - bits = 0 - idx = 1 if sys.byteorder == 'little' else 0 - if magic_words[idx] == kMagicFirstHalf: - if magic_words[1-idx] == kMagic64SecondHalf: - bits = 64 - elif magic_words[1-idx] == kMagic32SecondHalf: - bits = 32 - if bits == 0: - raise Exception('Bad magic word in %s' % path) - return bits + magic_bytes = f.read(8) + magic_words = struct.unpack("II", magic_bytes) + bits = 0 + idx = 1 if sys.byteorder == "little" else 0 + if magic_words[idx] == kMagicFirstHalf: + if magic_words[1 - idx] == kMagic64SecondHalf: + bits = 64 + elif magic_words[1 - idx] == kMagic32SecondHalf: + bits = 32 + if bits == 0: + raise Exception("Bad magic word in %s" % path) + return bits + def ReadOneFile(path): - with open(path, mode="rb") as f: - f.seek(0, 2) - size = f.tell() - f.seek(0, 0) - if size < 8: - raise Exception('File %s is short (< 8 bytes)' % path) - bits = ReadMagicAndReturnBitness(f, path) - size -= 8 - w = size * 8 // bits - s = struct.unpack_from(TypeCodeForStruct(bits) * (w), f.read(size)) - sys.stderr.write( - "%s: read %d %d-bit PCs from %s\n" % (prog_name, w, bits, path)) - return s + with open(path, mode="rb") as f: + f.seek(0, 2) + size = f.tell() + f.seek(0, 0) + if size < 8: + raise Exception("File %s is short (< 8 bytes)" % path) + bits = ReadMagicAndReturnBitness(f, path) + size -= 8 + w = size * 8 // bits + s = struct.unpack_from(TypeCodeForStruct(bits) * (w), f.read(size)) + sys.stderr.write("%s: read %d %d-bit PCs from %s\n" % (prog_name, w, bits, path)) + return s + def Merge(files): - s = set() - for f in files: - s = s.union(set(ReadOneFile(f))) - sys.stderr.write( - "%s: %d files merged; %d PCs total\n" % (prog_name, len(files), len(s)) - ) - return sorted(s) + s = set() + for f in files: + s = s.union(set(ReadOneFile(f))) + sys.stderr.write( + "%s: %d files merged; %d PCs total\n" % (prog_name, len(files), len(s)) + ) + return sorted(s) + def PrintFiles(files): - if len(files) > 1: - s = Merge(files) - else: # If there is just on file, print the PCs in order. - s = ReadOneFile(files[0]) - sys.stderr.write("%s: 1 file merged; %d PCs total\n" % (prog_name, len(s))) - for i in s: - print("0x%x" % i) + if len(files) > 1: + s = Merge(files) + else: # If there is just on file, print the PCs in order. + s = ReadOneFile(files[0]) + sys.stderr.write("%s: 1 file merged; %d PCs total\n" % (prog_name, len(s))) + for i in s: + print("0x%x" % i) + def MergeAndPrint(files): - if sys.stdout.isatty(): - Usage() - s = Merge(files) - bits = 32 - if max(s) > 0xFFFFFFFF: - bits = 64 - stdout_buf = getattr(sys.stdout, 'buffer', sys.stdout) - array.array('I', MagicForBits(bits)).tofile(stdout_buf) - a = struct.pack(TypeCodeForStruct(bits) * len(s), *s) - stdout_buf.write(a) + if sys.stdout.isatty(): + Usage() + s = Merge(files) + bits = 32 + if max(s) > 0xFFFFFFFF: + bits = 64 + stdout_buf = getattr(sys.stdout, "buffer", sys.stdout) + array.array("I", MagicForBits(bits)).tofile(stdout_buf) + a = struct.pack(TypeCodeForStruct(bits) * len(s), *s) + stdout_buf.write(a) def UnpackOneFile(path): - with open(path, mode="rb") as f: - sys.stderr.write("%s: unpacking %s\n" % (prog_name, path)) - while True: - header = f.read(12) - if not header: return - if len(header) < 12: - break - pid, module_length, blob_size = struct.unpack('iII', header) - module = f.read(module_length).decode('utf-8') - blob = f.read(blob_size) - assert(len(module) == module_length) - assert(len(blob) == blob_size) - extracted_file = "%s.%d.sancov" % (module, pid) - sys.stderr.write("%s: extracting %s\n" % (prog_name, extracted_file)) - # The packed file may contain multiple blobs for the same pid/module - # pair. Append to the end of the file instead of overwriting. - with open(extracted_file, 'ab') as f2: - f2.write(blob) - # fail - raise Exception('Error reading file %s' % path) + with open(path, mode="rb") as f: + sys.stderr.write("%s: unpacking %s\n" % (prog_name, path)) + while True: + header = f.read(12) + if not header: + return + if len(header) < 12: + break + pid, module_length, blob_size = struct.unpack("iII", header) + module = f.read(module_length).decode("utf-8") + blob = f.read(blob_size) + assert len(module) == module_length + assert len(blob) == blob_size + extracted_file = "%s.%d.sancov" % (module, pid) + sys.stderr.write("%s: extracting %s\n" % (prog_name, extracted_file)) + # The packed file may contain multiple blobs for the same pid/module + # pair. Append to the end of the file instead of overwriting. + with open(extracted_file, "ab") as f2: + f2.write(blob) + # fail + raise Exception("Error reading file %s" % path) def Unpack(files): - for f in files: - UnpackOneFile(f) + for f in files: + UnpackOneFile(f) + def UnpackOneRawFile(path, map_path): - mem_map = [] - with open(map_path, mode="rt") as f_map: - sys.stderr.write("%s: reading map %s\n" % (prog_name, map_path)) - bits = int(f_map.readline()) - if bits != 32 and bits != 64: - raise Exception('Wrong bits size in the map') - for line in f_map: - parts = line.rstrip().split() - mem_map.append((int(parts[0], 16), - int(parts[1], 16), - int(parts[2], 16), - ' '.join(parts[3:]))) - mem_map.sort(key=lambda m : m[0]) - mem_map_keys = [m[0] for m in mem_map] - - with open(path, mode="rb") as f: - sys.stderr.write("%s: unpacking %s\n" % (prog_name, path)) - - f.seek(0, 2) - size = f.tell() - f.seek(0, 0) - pcs = struct.unpack_from(TypeCodeForStruct(bits) * (size * 8 // bits), f.read(size)) - mem_map_pcs = [[] for i in range(0, len(mem_map))] - - for pc in pcs: - if pc == 0: continue - map_idx = bisect.bisect(mem_map_keys, pc) - 1 - (start, end, base, module_path) = mem_map[map_idx] - assert pc >= start - if pc >= end: - sys.stderr.write("warning: %s: pc %x outside of any known mapping\n" % (prog_name, pc)) - continue - mem_map_pcs[map_idx].append(pc - base) - - for ((start, end, base, module_path), pc_list) in zip(mem_map, mem_map_pcs): - if len(pc_list) == 0: continue - assert path.endswith('.sancov.raw') - dst_path = module_path + '.' + os.path.basename(path)[:-4] - sys.stderr.write("%s: writing %d PCs to %s\n" % (prog_name, len(pc_list), dst_path)) - sorted_pc_list = sorted(pc_list) - pc_buffer = struct.pack(TypeCodeForStruct(bits) * len(pc_list), *sorted_pc_list) - with open(dst_path, 'ab+') as f2: - array.array('I', MagicForBits(bits)).tofile(f2) - f2.seek(0, 2) - f2.write(pc_buffer) + mem_map = [] + with open(map_path, mode="rt") as f_map: + sys.stderr.write("%s: reading map %s\n" % (prog_name, map_path)) + bits = int(f_map.readline()) + if bits != 32 and bits != 64: + raise Exception("Wrong bits size in the map") + for line in f_map: + parts = line.rstrip().split() + mem_map.append( + ( + int(parts[0], 16), + int(parts[1], 16), + int(parts[2], 16), + " ".join(parts[3:]), + ) + ) + mem_map.sort(key=lambda m: m[0]) + mem_map_keys = [m[0] for m in mem_map] + + with open(path, mode="rb") as f: + sys.stderr.write("%s: unpacking %s\n" % (prog_name, path)) + + f.seek(0, 2) + size = f.tell() + f.seek(0, 0) + pcs = struct.unpack_from( + TypeCodeForStruct(bits) * (size * 8 // bits), f.read(size) + ) + mem_map_pcs = [[] for i in range(0, len(mem_map))] + + for pc in pcs: + if pc == 0: + continue + map_idx = bisect.bisect(mem_map_keys, pc) - 1 + (start, end, base, module_path) = mem_map[map_idx] + assert pc >= start + if pc >= end: + sys.stderr.write( + "warning: %s: pc %x outside of any known mapping\n" + % (prog_name, pc) + ) + continue + mem_map_pcs[map_idx].append(pc - base) + + for (start, end, base, module_path), pc_list in zip(mem_map, mem_map_pcs): + if len(pc_list) == 0: + continue + assert path.endswith(".sancov.raw") + dst_path = module_path + "." + os.path.basename(path)[:-4] + sys.stderr.write( + "%s: writing %d PCs to %s\n" % (prog_name, len(pc_list), dst_path) + ) + sorted_pc_list = sorted(pc_list) + pc_buffer = struct.pack( + TypeCodeForStruct(bits) * len(pc_list), *sorted_pc_list + ) + with open(dst_path, "ab+") as f2: + array.array("I", MagicForBits(bits)).tofile(f2) + f2.seek(0, 2) + f2.write(pc_buffer) + def RawUnpack(files): - for f in files: - if not f.endswith('.sancov.raw'): - raise Exception('Unexpected raw file name %s' % f) - f_map = f[:-3] + 'map' - UnpackOneRawFile(f, f_map) + for f in files: + if not f.endswith(".sancov.raw"): + raise Exception("Unexpected raw file name %s" % f) + f_map = f[:-3] + "map" + UnpackOneRawFile(f, f_map) + def GetInstrumentedPCs(binary): - # This looks scary, but all it does is extract all offsets where we call: - # - __sanitizer_cov() or __sanitizer_cov_with_check(), - # - with call or callq, - # - directly or via PLT. - cmd = r"objdump --no-show-raw-insn -d %s | " \ - r"grep '^\s\+[0-9a-f]\+:\s\+call\(q\|\)\s\+\(0x\|\)[0-9a-f]\+ <__sanitizer_cov\(_with_check\|\|_trace_pc_guard\)\(@plt\|\)>' | " \ + # This looks scary, but all it does is extract all offsets where we call: + # - __sanitizer_cov() or __sanitizer_cov_with_check(), + # - with call or callq, + # - directly or via PLT. + cmd = ( + r"objdump --no-show-raw-insn -d %s | " + r"grep '^\s\+[0-9a-f]\+:\s\+call\(q\|\)\s\+\(0x\|\)[0-9a-f]\+ <__sanitizer_cov\(_with_check\|\|_trace_pc_guard\)\(@plt\|\)>' | " r"grep -o '^\s\+[0-9a-f]\+'" % binary - lines = subprocess.check_output(cmd, stdin=subprocess.PIPE, shell=True).splitlines() - # The PCs we get from objdump are off by 4 bytes, as they point to the - # beginning of the callq instruction. Empirically this is true on x86 and - # x86_64. - return set(int(line.strip(), 16) + 4 for line in lines) + ) + lines = subprocess.check_output(cmd, stdin=subprocess.PIPE, shell=True).splitlines() + # The PCs we get from objdump are off by 4 bytes, as they point to the + # beginning of the callq instruction. Empirically this is true on x86 and + # x86_64. + return set(int(line.strip(), 16) + 4 for line in lines) + def PrintMissing(binary): - if not os.path.isfile(binary): - raise Exception('File not found: %s' % binary) - instrumented = GetInstrumentedPCs(binary) - sys.stderr.write("%s: found %d instrumented PCs in %s\n" % (prog_name, - len(instrumented), - binary)) - covered = set(int(line, 16) for line in sys.stdin) - sys.stderr.write("%s: read %d PCs from stdin\n" % (prog_name, len(covered))) - missing = instrumented - covered - sys.stderr.write("%s: %d PCs missing from coverage\n" % (prog_name, len(missing))) - if (len(missing) > len(instrumented) - len(covered)): + if not os.path.isfile(binary): + raise Exception("File not found: %s" % binary) + instrumented = GetInstrumentedPCs(binary) sys.stderr.write( - "%s: WARNING: stdin contains PCs not found in binary\n" % prog_name + "%s: found %d instrumented PCs in %s\n" % (prog_name, len(instrumented), binary) ) - for pc in sorted(missing): - print("0x%x" % pc) - -if __name__ == '__main__': - prog_name = sys.argv[0] - if len(sys.argv) <= 2: - Usage(); - - if sys.argv[1] == "missing": - if len(sys.argv) != 3: - Usage() - PrintMissing(sys.argv[2]) - exit(0) - - file_list = [] - for f in sys.argv[2:]: - file_list += glob.glob(f) - if not file_list: - Usage() - - if sys.argv[1] == "print": - PrintFiles(file_list) - elif sys.argv[1] == "merge": - MergeAndPrint(file_list) - elif sys.argv[1] == "unpack": - Unpack(file_list) - elif sys.argv[1] == "rawunpack": - RawUnpack(file_list) - else: - Usage() + covered = set(int(line, 16) for line in sys.stdin) + sys.stderr.write("%s: read %d PCs from stdin\n" % (prog_name, len(covered))) + missing = instrumented - covered + sys.stderr.write("%s: %d PCs missing from coverage\n" % (prog_name, len(missing))) + if len(missing) > len(instrumented) - len(covered): + sys.stderr.write( + "%s: WARNING: stdin contains PCs not found in binary\n" % prog_name + ) + for pc in sorted(missing): + print("0x%x" % pc) + + +if __name__ == "__main__": + prog_name = sys.argv[0] + if len(sys.argv) <= 2: + Usage() + + if sys.argv[1] == "missing": + if len(sys.argv) != 3: + Usage() + PrintMissing(sys.argv[2]) + exit(0) + + file_list = [] + for f in sys.argv[2:]: + file_list += glob.glob(f) + if not file_list: + Usage() + + if sys.argv[1] == "print": + PrintFiles(file_list) + elif sys.argv[1] == "merge": + MergeAndPrint(file_list) + elif sys.argv[1] == "unpack": + Unpack(file_list) + elif sys.argv[1] == "rawunpack": + RawUnpack(file_list) + else: + Usage() diff --git a/polytracker/src/passes/CMakeLists.txt b/polytracker/src/passes/CMakeLists.txt index 87080e67..af6aaa9d 100644 --- a/polytracker/src/passes/CMakeLists.txt +++ b/polytracker/src/passes/CMakeLists.txt @@ -6,7 +6,7 @@ endif(APPLE) add_library( PolytrackerPass SHARED - taint_tracking.cpp remove_fn_attr.cpp function_tracing.cpp + taint_tracking.cpp remove_fn_attr.cpp function_tracing.cpp tainted_control_flow.cpp DataFlowSanitizer.cpp utils.cpp pass_plugin.cpp) target_link_libraries( diff --git a/polytracker/src/passes/pass_plugin.cpp b/polytracker/src/passes/pass_plugin.cpp index c026d5ba..e8ad4a1e 100644 --- a/polytracker/src/passes/pass_plugin.cpp +++ b/polytracker/src/passes/pass_plugin.cpp @@ -13,6 +13,7 @@ #include "polytracker/passes/function_tracing.h" #include "polytracker/passes/remove_fn_attr.h" #include "polytracker/passes/taint_tracking.h" +#include "polytracker/passes/tainted_control_flow.h" llvm::PassPluginLibraryInfo getPolyTrackerPluginInfo() { return {LLVM_PLUGIN_API_VERSION, "PolyTracker", "", @@ -36,6 +37,10 @@ llvm::PassPluginLibraryInfo getPolyTrackerPluginInfo() { mpm.addPass(polytracker::FunctionTracingPass()); return true; } + if (name == "pt-tcf") { + mpm.addPass(polytracker::TaintedControlFlowPass()); + return true; + } return false; }); }}; diff --git a/polytracker/src/passes/tainted_control_flow.cpp b/polytracker/src/passes/tainted_control_flow.cpp new file mode 100644 index 00000000..d8142794 --- /dev/null +++ b/polytracker/src/passes/tainted_control_flow.cpp @@ -0,0 +1,198 @@ +/* + * Copyright (c) 2022-present, Trail of Bits, Inc. + * All rights reserved. + * + * This source code is licensed in accordance with the terms specified in + * the LICENSE file found in the root directory of this source tree. + */ + +#include "polytracker/passes/tainted_control_flow.h" + +#include +#include +#include +#include + +#include + +#include "polytracker/dfsan_types.h" +#include "polytracker/passes/utils.h" + +#include + +namespace polytracker { + +namespace detail { +// Helper type to produce the json file of function names by functionid +class FunctionMappingJSONWriter { +public: + FunctionMappingJSONWriter(std::string_view filename) + : file(filename.data(), std::ios::binary) { + file << "["; + } + + ~FunctionMappingJSONWriter() { + // Back up and erase the last ",\n" + file.seekp(-2, std::ios::cur); + file << "\n]\n"; + } + + void append(std::string_view name) { + // Will cause an additional ',' but don't care about that right now... + // The destructor will back up two steps and replace the ',' with a newline + // and array termination. + file << "\"" << name << "\",\n"; + } + +private: + std::ofstream file; +}; +} // namespace detail + +namespace { +uint32_t +get_or_add_mapping(uintptr_t key, std::unordered_map &m, + uint32_t &counter, std::string_view name, + polytracker::detail::FunctionMappingJSONWriter &js) { + if (auto it = m.find(key); it != m.end()) { + return it->second; + } else { + js.append(name); + return m[key] = counter++; + } +} + +} // namespace +void TaintedControlFlowPass::insertCondBrLogCall(llvm::Instruction &inst, + llvm::Value *val) { + llvm::IRBuilder<> ir(&inst); + auto dummy_val{val}; + if (inst.getType()->isVectorTy()) { + dummy_val = ir.CreateExtractElement(val, uint64_t(0)); + } + ir.CreateCall(cond_br_log_fn, {ir.CreateSExtOrTrunc(dummy_val, label_ty)}); +} + +llvm::ConstantInt * +TaintedControlFlowPass::get_function_id_const(llvm::Function &func) { + auto func_address = reinterpret_cast(&func); + std::string_view name = func.getName(); + auto fid = get_or_add_mapping(func_address, function_ids_, function_counter_, + name, *function_mapping_writer_); + return llvm::ConstantInt::get(func.getContext(), llvm::APInt(32, fid, false)); +} + +llvm::ConstantInt * +TaintedControlFlowPass::get_function_id_const(llvm::Instruction &i) { + return get_function_id_const(*(i.getParent()->getParent())); +} + +void TaintedControlFlowPass::visitGetElementPtrInst( + llvm::GetElementPtrInst &gep) { + llvm::IRBuilder<> ir(&gep); + for (auto &idx : gep.indices()) { + if (llvm::isa(idx)) { + continue; + } + + auto callret = ir.CreateCall(cond_br_log_fn, + {ir.CreateSExtOrTrunc(idx, ir.getInt64Ty()), + get_function_id_const(gep)}); + + idx = ir.CreateSExtOrTrunc(callret, idx->getType()); + } +} + +void TaintedControlFlowPass::visitBranchInst(llvm::BranchInst &bi) { + if (bi.isUnconditional()) { + return; + } + + llvm::IRBuilder<> ir(&bi); + auto cond = bi.getCondition(); + + auto callret = ir.CreateCall( + cond_br_log_fn, + {ir.CreateSExtOrTrunc(cond, ir.getInt64Ty()), get_function_id_const(bi)}); + + bi.setCondition(ir.CreateSExtOrTrunc(callret, cond->getType())); +} + +void TaintedControlFlowPass::visitSwitchInst(llvm::SwitchInst &si) { + llvm::IRBuilder<> ir(&si); + auto cond = si.getCondition(); + + auto callret = ir.CreateCall( + cond_br_log_fn, + {ir.CreateSExtOrTrunc(cond, ir.getInt64Ty()), get_function_id_const(si)}); + + si.setCondition(ir.CreateSExtOrTrunc(callret, cond->getType())); +} + +void TaintedControlFlowPass::visitSelectInst(llvm::SelectInst &si) { + // TODO(hbrodin): Can't handle atm. + if (si.getType()->isVectorTy()) { + return; + } + llvm::IRBuilder<> ir(&si); + auto cond = si.getCondition(); + + auto callret = ir.CreateCall( + cond_br_log_fn, + {ir.CreateSExtOrTrunc(cond, ir.getInt64Ty()), get_function_id_const(si)}); + + si.setCondition(ir.CreateSExtOrTrunc(callret, cond->getType())); +} + +void TaintedControlFlowPass::declareLoggingFunctions(llvm::Module &mod) { + llvm::IRBuilder<> ir(mod.getContext()); + cond_br_log_fn = mod.getOrInsertFunction( + "__polytracker_log_tainted_control_flow", + llvm::AttributeList::get( + mod.getContext(), + {{llvm::AttributeList::FunctionIndex, + llvm::Attribute::get(mod.getContext(), + llvm::Attribute::ReadNone)}}), + ir.getInt64Ty(), ir.getInt64Ty(), ir.getInt32Ty()); + + fn_enter_log_fn = mod.getOrInsertFunction("__polytracker_enter_function", + ir.getVoidTy(), ir.getInt32Ty()); + + fn_leave_log_fn = mod.getOrInsertFunction("__polytracker_leave_function", + ir.getVoidTy(), ir.getInt32Ty()); +} + +void TaintedControlFlowPass::instrumentFunctionEnter(llvm::Function &func) { + if (func.isDeclaration()) { + return; + } + llvm::IRBuilder<> ir(&*func.getEntryBlock().begin()); + ir.CreateCall(fn_enter_log_fn, get_function_id_const(func)); +} + +void TaintedControlFlowPass::visitReturnInst(llvm::ReturnInst &ri) { + llvm::IRBuilder<> ir(&ri); + ir.CreateCall(fn_leave_log_fn, get_function_id_const(ri)); +} + +llvm::PreservedAnalyses +TaintedControlFlowPass::run(llvm::Module &mod, + llvm::ModuleAnalysisManager &mam) { + label_ty = llvm::IntegerType::get(mod.getContext(), DFSAN_LABEL_BITS); + declareLoggingFunctions(mod); + for (auto &fn : mod) { + instrumentFunctionEnter(fn); + visit(fn); + } + return llvm::PreservedAnalyses::none(); +} + +TaintedControlFlowPass::TaintedControlFlowPass() + : function_mapping_writer_( + std::make_unique( + "functionid.json")) {} + +TaintedControlFlowPass::~TaintedControlFlowPass() = default; +TaintedControlFlowPass::TaintedControlFlowPass(TaintedControlFlowPass &&) = + default; +} // namespace polytracker \ No newline at end of file diff --git a/polytracker/src/polytracker/polytracker.cpp b/polytracker/src/polytracker/polytracker.cpp index 477c7e33..1f30d4f4 100644 --- a/polytracker/src/polytracker/polytracker.cpp +++ b/polytracker/src/polytracker/polytracker.cpp @@ -41,4 +41,23 @@ extern "C" void __taint_start() { taint_start(); } extern "C" void __polytracker_taint_argv(int argc, char *argv[]) { polytracker::taint_argv(argc, argv); +} + +extern "C" uint64_t __dfsw___polytracker_log_tainted_control_flow( + uint64_t conditional, uint32_t functionid, dfsan_label conditional_label, + dfsan_label function_label, dfsan_label *ret_label) { + if (conditional_label > 0) { + get_polytracker_tdag().log_tainted_control_flow(conditional_label, + functionid); + } + *ret_label = conditional_label; + return conditional; +} + +extern "C" void __polytracker_enter_function(uint32_t function_id) { + get_polytracker_tdag().enter_function(function_id); +} + +extern "C" void __polytracker_leave_function(uint32_t function_id) { + get_polytracker_tdag().leave_function(function_id); } \ No newline at end of file diff --git a/polytracker/src/taintdag/polytracker.cpp b/polytracker/src/taintdag/polytracker.cpp index 89f41bad..65b683a8 100644 --- a/polytracker/src/taintdag/polytracker.cpp +++ b/polytracker/src/taintdag/polytracker.cpp @@ -175,6 +175,18 @@ void PolyTracker::affects_control_flow(label_t lbl) { output_file_.section().affects_control_flow(lbl); } +void PolyTracker::log_tainted_control_flow(label_t lbl, uint32_t function_id) { + output_file_.section().tainted_control_flow(lbl, function_id); +} + +void PolyTracker::enter_function(uint32_t function_id) { + output_file_.section().enter_function(function_id); +} + +void PolyTracker::leave_function(uint32_t function_id) { + output_file_.section().leave_function(function_id); +} + Functions::index_t PolyTracker::function_entry(std::string_view name) { auto &functions{output_file_.section()}; auto maybe_index{functions.add_mapping(name)}; diff --git a/polytracker/taint_dag.py b/polytracker/taint_dag.py index 46e7b5e8..744e5761 100644 --- a/polytracker/taint_dag.py +++ b/polytracker/taint_dag.py @@ -8,6 +8,7 @@ Tuple, List, Set, + Type, cast, ) @@ -123,12 +124,144 @@ def __init__(self, mem, hdr): self.section = mem[hdr.offset : hdr.offset + hdr.size] def read_raw(self, label): - return c_uint64.from_buffer_copy(self.section[label * sizeof(c_uint64) :]).value + return c_uint64.from_buffer_copy(self.section, label * sizeof(c_uint64)).value def count(self): return len(self.section) // sizeof(c_uint64) +class TDEnterFunctionEvent: + """Emitted whenever execution enters a function. + The callstack member is the callstack right before entering the function, + having the function just entered as the last member of the callstack. + """ + + def __init__(self, callstack): + """Callstack after entering function""" + self.callstack = callstack + + def __repr__(self) -> str: + return f"Enter: {self.callstack}" + + def __eq__(self, __o: object) -> bool: + if isinstance(__o, TDEnterFunctionEvent): + return self.callstack == __o.callstack + return False + + +class TDLeaveFunctionEvent: + """Emitted whenever execution leaves a function. + The callstack member is the callstack right before leaving the function, + having the function about to leave as the last member of the callstack. + """ + + def __init__(self, callstack): + """Callstack before leaving function""" + self.callstack = callstack + + def __repr__(self) -> str: + return f"Leave: {self.callstack}" + + def __eq__(self, __o: object) -> bool: + if isinstance(__o, TDLeaveFunctionEvent): + return self.callstack == __o.callstack + return False + + +class TDTaintedControlFlowEvent: + """Emitted whenever a control flow change is influenced by tainted data. + The label that influenced the control flow is available in the `label` member. + Current callstack (including the function the control flow happened in) is available + in the `callstack` member.""" + + def __init__(self, callstack, label): + self.callstack = callstack + self.label = label + + def __repr__(self) -> str: + return f"TaintedControlFlow label {self.label} callstack {self.callstack}" + + def __eq__(self, __o: object) -> bool: + if isinstance(__o, TDTaintedControlFlowEvent): + return self.label == __o.label and self.callstack == __o.callstack + return False + + +class TDControlFlowLogSection: + """TDAG Control flow log section + + Interprets the control flow log section in a TDAG file. + Enables enumeration/random access of items + """ + + # NOTE: MUST correspond to the members in the `ControlFlowLog::EventType`` in `control_flog_log.h`. + ENTER_FUNCTION = 0 + LEAVE_FUNCTION = 1 + TAINTED_CONTROL_FLOW = 2 + + @staticmethod + def _decode_varint(buffer): + shift = 0 + val = 0 + while buffer: + curr = c_uint8.from_buffer_copy(buffer, 0).value + val |= (curr & 0x7F) << shift + shift += 7 + buffer = buffer[1:] + if curr & 0x80 == 0: + break + + return val, buffer + + @staticmethod + def _align_callstack(target_function_id, callstack): + while callstack and callstack[-1] != target_function_id: + yield TDLeaveFunctionEvent(callstack[:]) + callstack.pop() + + def __init__(self, mem, hdr): + self.section = mem[hdr.offset : hdr.offset + hdr.size] + self.funcmapping = None + + def __iter__(self): + buffer = self.section + callstack = [] + while buffer: + event = c_uint8.from_buffer_copy(buffer, 0).value + buffer = buffer[1:] + function_id, buffer = TDControlFlowLogSection._decode_varint(buffer) + if self.funcmapping != None: + function_id = self.funcmapping[function_id] + + if event == TDControlFlowLogSection.ENTER_FUNCTION: + callstack.append(function_id) + yield TDEnterFunctionEvent(callstack[:]) + elif event == TDControlFlowLogSection.LEAVE_FUNCTION: + # Align call stack, if needed + yield from TDControlFlowLogSection._align_callstack( + function_id, callstack + ) + + # TODO(hbrodin): If the callstack doesn't contain function_id at all, this will break. + yield TDLeaveFunctionEvent(callstack[:]) + callstack.pop() + else: + # Align call stack, if needed + yield from TDControlFlowLogSection._align_callstack( + function_id, callstack + ) + + label, buffer = TDControlFlowLogSection._decode_varint(buffer) + yield TDTaintedControlFlowEvent(callstack[:], label) + + # Drain callstack with artifical TDLeaveFunction events (using a dummy function id that doesn't exist) + yield from TDControlFlowLogSection._align_callstack(-1, callstack) + + def function_id_mapping(self, id_to_name_array): + """This method stores an array used to translate from function id to symbolic names""" + self.funcmapping = id_to_name_array + + class TDSinkSection: """TDAG Sinks section @@ -163,7 +296,7 @@ def enumerate_set_bits(self): """ index = 0 for offset in range(0, len(self.section), sizeof(c_uint64)): - bucket = c_uint64.from_buffer_copy(self.section[offset:]).value + bucket = c_uint64.from_buffer_copy(self.section, offset).value if bucket == 0: index += 64 # No bits set, just advance the bit index else: @@ -291,6 +424,18 @@ def __repr__(self) -> str: return f"kind: {self.Kind(self.kind).name} fnidx: {self.fnidx}" +TDSection = Union[ + TDLabelSection, + TDSourceSection, + TDStringSection, + TDSinkSection, + TDSourceIndexSection, + TDFunctionsSection, + TDEventsSection, + TDControlFlowLogSection, +] + + class TDFile: def __init__(self, file: BinaryIO) -> None: # This needs to be kept in sync with implementation in encoding.cpp @@ -307,35 +452,36 @@ def __init__(self, file: BinaryIO) -> None: self.filemeta = TDFileMeta.from_buffer_copy(self.buffer) section_offset = sizeof(TDFileMeta) - self.sections: List[ - Union[ - TDLabelSection, - TDSourceSection, - TDStringSection, - TDSinkSection, - TDSourceIndexSection, - TDFunctionsSection, - TDEventsSection, - ] - ] = [] + self.sections: List[TDSection] = [] + self.sections_by_type: Dict[Type[TDSection], TDSection] = {} for i in range(0, self.filemeta.section_count): hdr = TDSectionMeta.from_buffer_copy(self.buffer, section_offset) if hdr.tag == 1: self.sections.append(TDSourceSection(self.buffer, hdr)) + self.sections_by_type[TDSourceSection] = self.sections[-1] elif hdr.tag == 2: self.sections.append(TDLabelSection(self.buffer, hdr)) + self.sections_by_type[TDLabelSection] = self.sections[-1] elif hdr.tag == 3: self.sections.append(TDStringSection(self.buffer, hdr)) + self.sections_by_type[TDStringSection] = self.sections[-1] elif hdr.tag == 4: self.sections.append(TDSinkSection(self.buffer, hdr)) + self.sections_by_type[TDSinkSection] = self.sections[-1] elif hdr.tag == 5: self.sections.append(TDSourceIndexSection(self.buffer, hdr)) + self.sections_by_type[TDSourceIndexSection] = self.sections[-1] elif hdr.tag == 6: self.sections.append(TDFunctionsSection(self.buffer, hdr)) + self.sections_by_type[TDFunctionsSection] = self.sections[-1] elif hdr.tag == 7: self.sections.append(TDEventsSection(self.buffer, hdr)) + self.sections_by_type[TDEventsSection] = self.sections[-1] + elif hdr.tag == 8: + self.sections.append(TDControlFlowLogSection(self.buffer, hdr)) + self.sections_by_type[TDControlFlowLogSection] = self.sections[-1] else: - raise Exception("Unsupported section tag") + raise NotImplementedError("Unsupported section tag") section_offset += sizeof(TDSectionMeta) @@ -345,38 +491,47 @@ def __init__(self, file: BinaryIO) -> None: self.fd_headers: List[Tuple[Path, TDFDHeader]] = list(self.read_fd_headers()) self.fn_headers: List[Tuple[str, TDFnHeader]] = list(self.read_fn_headers()) - def _get_section(self, wanted_type): - return next(filter(lambda x: isinstance(x, wanted_type), self.sections)) + def _get_section(self, wanted_type: Type[TDSection]) -> TDSection: + return self.sections_by_type[wanted_type] def read_fd_headers(self) -> Iterator[Tuple[Path, TDFDHeader]]: - sources = self._get_section(TDSourceSection) - strings = self._get_section(TDStringSection) + sources = self.sections_by_type[TDSourceSection] + strings = self.sections_by_type[TDStringSection] + assert isinstance(sources, TDSourceSection) + assert isinstance(strings, TDStringSection) - yield from map( - lambda x: (Path(strings.read_string(x.name_offset)), x), sources.enumerate() + yield from ( + (Path(strings.read_string(x.name_offset)), x) for x in sources.enumerate() ) def read_fn_headers(self) -> Iterator[Tuple[str, TDFnHeader]]: - functions = self._get_section(TDFunctionsSection) - strings = self._get_section(TDStringSection) + functions = self.sections_by_type[TDFunctionsSection] + strings = self.sections_by_type[TDStringSection] + assert isinstance(functions, TDFunctionsSection) + assert isinstance(strings, TDStringSection) for header in functions: name = strings.read_string(header.name_offset) - yield (name, header) + yield name, header def input_labels(self) -> Iterator[int]: """Enumerates all taint labels that are input labels (source taint)""" - return self._get_section(TDSourceIndexSection).enumerate_set_bits() + source_index_section = self.sections_by_type[TDSourceIndexSection] + assert isinstance(source_index_section, TDSourceIndexSection) + return source_index_section.enumerate_set_bits() @property def label_count(self): - return self._get_section(TDLabelSection).count() + label_section = self.sections_by_type[TDLabelSection] + assert isinstance(label_section, TDLabelSection) + return label_section.count() def read_node(self, label: int) -> int: if label in self.raw_nodes: return self.raw_nodes[label] - - result = self._get_section(TDLabelSection).read_raw(label) + label_section = self.sections_by_type[TDLabelSection] + assert isinstance(label_section, TDLabelSection) + result = label_section.read_raw(label) self.raw_nodes[label] = result return result @@ -410,14 +565,18 @@ def nodes(self) -> Iterator[TDNode]: @property def sinks(self) -> Iterator[TDSink]: - yield from self._get_section(TDSinkSection).enumerate() + sink_section = self.sections_by_type[TDSinkSection] + assert isinstance(sink_section, TDSinkSection) + yield from sink_section.enumerate() def read_event(self, offset: int) -> TDEvent: return TDEvent.from_buffer_copy(self.buffer, offset) @property def events(self) -> Iterator[TDEvent]: - yield from self._get_section(TDEventsSection) + events_section = self.sections_by_type[TDEventsSection] + assert isinstance(events_section, TDEventsSection) + yield from events_section class TDTaintOutput(TaintOutput): @@ -693,6 +852,13 @@ def __init_arguments__(self, parser): help="print function trace events", ) + parser.add_argument( + "--print-control-flow-log", + "-c", + action="store_true", + help="print function trace events", + ) + def run(self, args): with open(args.POLYTRACKER_TF, "rb") as f: tdfile = TDFile(f) @@ -719,3 +885,9 @@ def run(self, args): if args.print_function_trace: for e in tdfile.events: print(f"{e}") + + if args.print_control_flow_log: + cflog = tdfile._get_section(TDControlFlowLogSection) + assert isinstance(cflog, TDControlFlowLogSection) + for obj in cflog: + print(f"{obj}") diff --git a/tests/conftest.py b/tests/conftest.py index 6f5f9df8..8114bafc 100644 --- a/tests/conftest.py +++ b/tests/conftest.py @@ -35,7 +35,7 @@ def build(target: Path, binary: Path) -> None: def instrument(target: str) -> None: - cmd = ["instrument-targets", "--taint", "--ftrace", target] + cmd = ["instrument-targets", "--taint", "--ftrace", "--cflog", target] run_polytracker(cmd) diff --git a/tests/test_cf_log.cpp b/tests/test_cf_log.cpp new file mode 100644 index 00000000..5b992a9e --- /dev/null +++ b/tests/test_cf_log.cpp @@ -0,0 +1,71 @@ +#include +#include +#include +#include + +int f2(uint8_t val) { + if (val & 1) { + return val + 4; + } else { + return val + 3; + } +} + +int f3(uint8_t val) { + if (val > 0) { + return val - 1; + } else { + return val; + } +} + +int f1(uint8_t val) { + if (val > 5) { + return f2(val); + } else { + return 2 + f3(val + 2); + } +} + +// Some dummy control flow to test that control flow logging works as expected +int main(int argc, char *argv[]) { + uint8_t buffer[sizeof(uint64_t)]; + if (sizeof(buffer) != read(0, buffer, sizeof(buffer))) { + exit(EXIT_FAILURE); + } + + bool good = true; + + // Control flow 1, affects label 1 + if (buffer[0] == 'a') { + // Control flow 2, affects label 2 + if (buffer[1] != 'c') { + // Control flow labels 3 trough 8 + for (size_t i = 2; i < sizeof(buffer); i++) { + if (buffer[i] == '\0') { + good = false; + } + } + } + } + + if (good) { + // Union/range + uint64_t val = 0; + for (auto v : buffer) { + val = (val << 8) | v; + } + // Control flow label 15. The range node covering the full input buffer + if (val == 1) { + printf("Wow, that was unexpected\n"); + } + + // Control flow label 3 (again) + if (buffer[2] < 16) { + printf("OK, buffer[2] < 16\n"); + } + + auto v = f1(buffer[6]); + } + exit(EXIT_SUCCESS); +} \ No newline at end of file diff --git a/tests/test_cf_log.py b/tests/test_cf_log.py new file mode 100644 index 00000000..5316fac7 --- /dev/null +++ b/tests/test_cf_log.py @@ -0,0 +1,69 @@ +import cxxfilt +import json +import pytest +import subprocess + +import polytracker +from pathlib import Path + +from polytracker.taint_dag import ( + TDEnterFunctionEvent, + TDLeaveFunctionEvent, + TDTaintedControlFlowEvent, +) + + +@pytest.mark.program_trace("test_cf_log.cpp") +def test_cf_log(instrumented_binary: Path, trace_file: Path): + # Data to write to stdin, one byte at a time + stdin_data = "abcdefgh" + + subprocess.run( + [str(instrumented_binary)], + input=stdin_data.encode("utf-8"), + env={ + "POLYDB": str(trace_file), + "POLYTRACKER_STDIN_SOURCE": "1", + "POLYTRACKER_LOG_CONTROL_FLOW": "1", + }, + ) + + program_trace = polytracker.PolyTrackerTrace.load(trace_file) + + cflog = program_trace.tdfile._get_section( + polytracker.taint_dag.TDControlFlowLogSection + ) + + # The functionid mapping is available next to the built binary + with open(instrumented_binary.parent / "functionid.json", "rb") as f: + functionid_mapping = list(map(cxxfilt.demangle, json.load(f))) + + # Apply the id to function mappign + cflog.function_id_mapping(functionid_mapping) + + expected_seq = [ + TDEnterFunctionEvent(["main"]), + TDTaintedControlFlowEvent(["main"], 1), + TDTaintedControlFlowEvent(["main"], 2), + TDTaintedControlFlowEvent(["main"], 3), + TDTaintedControlFlowEvent(["main"], 4), + TDTaintedControlFlowEvent(["main"], 5), + TDTaintedControlFlowEvent(["main"], 6), + TDTaintedControlFlowEvent(["main"], 7), + TDTaintedControlFlowEvent(["main"], 8), + TDTaintedControlFlowEvent(["main"], 15), + TDTaintedControlFlowEvent(["main"], 3), + TDEnterFunctionEvent(["main", "f1(unsigned char)"]), + TDTaintedControlFlowEvent(["main", "f1(unsigned char)"], 7), + TDEnterFunctionEvent(["main", "f1(unsigned char)", "f2(unsigned char)"]), + TDTaintedControlFlowEvent( + ["main", "f1(unsigned char)", "f2(unsigned char)"], 7 + ), + TDLeaveFunctionEvent(["main", "f1(unsigned char)", "f2(unsigned char)"]), + TDLeaveFunctionEvent(["main", "f1(unsigned char)"]), + TDLeaveFunctionEvent(["main"]), # This is artifical as there is a call to exit + ] + + # NOTE(hbrodin): Could have done assert list(cflog) == expected_seq, but this provides the failed element + for got, expected in zip(cflog, expected_seq): + assert got == expected diff --git a/tests/test_polytracker.py b/tests/test_polytracker.py index a228f81d..13bf6e29 100644 --- a/tests/test_polytracker.py +++ b/tests/test_polytracker.py @@ -3,8 +3,6 @@ from subprocess import CalledProcessError from typing import Dict, Union -from tqdm import tqdm - from polytracker import ( BasicBlockEntry, FunctionEntry, diff --git a/tests/test_socket_read_write.py b/tests/test_socket_read_write.py index bf88346a..b6cc9da3 100644 --- a/tests/test_socket_read_write.py +++ b/tests/test_socket_read_write.py @@ -16,6 +16,7 @@ def expected_source_name(binary_ip, binary_port, python_ip, python_port): return f"socket:{binary_ip}:{binary_port}-{python_ip}:{python_port}" +@pytest.mark.skip(reason="test may time out and never complete") @pytest.mark.program_trace("test_socket_read_write.cpp") @pytest.mark.parametrize("mode", ["client", "server"]) def test_socket_read_write(instrumented_binary: Path, trace_file: Path, mode: str): diff --git a/unittests/src/taintdag/CMakeLists.txt b/unittests/src/taintdag/CMakeLists.txt index fee4b776..b620b84a 100644 --- a/unittests/src/taintdag/CMakeLists.txt +++ b/unittests/src/taintdag/CMakeLists.txt @@ -11,7 +11,8 @@ add_executable( fntrace.cpp union.cpp labeldeq.cpp - stream_offset.cpp) + stream_offset.cpp + control_flow_log.cpp) target_include_directories(${TAINTDAG_UNITTEST} PRIVATE ${CMAKE_SOURCE_DIR}/polytracker/include) diff --git a/unittests/src/taintdag/control_flow_log.cpp b/unittests/src/taintdag/control_flow_log.cpp new file mode 100644 index 00000000..fcafe61b --- /dev/null +++ b/unittests/src/taintdag/control_flow_log.cpp @@ -0,0 +1,57 @@ + +/* + * Copyright (c) 2022-present, Trail of Bits, Inc. + * All rights reserved. + * + * This source code is licensed in accordance with the terms specified in + * the LICENSE file found in the root directory of this source tree. + */ + +#include "taintdag/control_flow_log.h" +#include "taintdag/section.h" +#include + +TEST_CASE("Simple varint encoding") { + using namespace taintdag::detail; + uint8_t buffer[5]; + + SECTION("Encode 0") { + auto n = varint_encode(0, buffer); + REQUIRE(n == 1); + REQUIRE(buffer[0] == 0); + } + + SECTION("Encode 1") { + auto n = varint_encode(1, buffer); + REQUIRE(n == 1); + REQUIRE(buffer[0] == 1); + } + + SECTION("Encode 0x7f") { + auto n = varint_encode(0x7f, buffer); + REQUIRE(n == 1); + REQUIRE(buffer[0] == 0x7f); + } + + SECTION("Encode 0x80") { + auto n = varint_encode(0x80, buffer); + REQUIRE(n == 2); + REQUIRE(buffer[0] == 0x80); + REQUIRE(buffer[1] == 0x01); + } + SECTION("Encode 0x3ffe") { + auto n = varint_encode(0x3ffe, buffer); + REQUIRE(n == 2); + REQUIRE(buffer[0] == 0xfe); + REQUIRE(buffer[1] == 0x7f); + } + SECTION("Encode 0xffffffff") { + auto n = varint_encode(0xffffffff, buffer); + REQUIRE(n == 5); + REQUIRE(buffer[0] == 0xff); + REQUIRE(buffer[1] == 0xff); + REQUIRE(buffer[2] == 0xff); + REQUIRE(buffer[3] == 0xff); + REQUIRE(buffer[4] == 0x0f); + } +}