Releases: mar-file-system/GUFI
0.5.1-rc0
Renamed
contrib/benchmark.sh
→contrib/canned_queries.sh
scripts/query_builder.py
→scripts/gufi_common.py
Internal Interface Changes
gufi_query
no longer has the options-P
and-p
- Updated scripts and tests to not pass these flags
- Removed unnecessary arguments from
dbutils
sqlite3 wrapper functions - OutputBuffers now has functions to flush to
FILE *
andFILE **
- QueuePerThreadPool
- No longer has thread pinning as part of the interface
- Fixed bug with terminating condition
QPTPool_start
can now be called before any items are placed in the work queue
- The function passed to
QPTPool_start
is now the default function to run- A function can also be provided to
QPTPool_enqueue
to change the behavior for that one work item
- A function can also be provided to
Updated travis scripts
Added --path-index
to gufi_find
to allow for selection of a single path from the list of paths provided in the config file.
Added debug prints to gufi_trace2index
Debug prints in gufi_query
and gufi_trace2index
go through OutputBuffers
, so they have very little overhead.
0.5.0-rc0
gufi_find
, gufi_ls
, and gufi_stats
now use /etc/GUFI/config
to set runtime settings.
gufi_config
can be used to generate configuration files.
0.4.1-rc0
Optimizations
- Results are now buffered instead of immediately printed
- the
-B
flag has been added togufi_query
to allow for the buffer size to be modified
- the
- Now using
jemalloc
as a replacement malloc to speed up allocations SNPRINTF
with only%s
format specifiers have been replaced withmemcpy
s- Calls to
access(2)
from the worker threads have been removed- Duplicate filesystem access since
opendir
also checks for accessibility - Potential security hole, since permissions might change between call to
access
andopendir
- Duplicate filesystem access since
Fixes:
- Removed all memory leaks from
gufi_query
- Fixed number of queries performed calculation
0.4.0-rc1
buffered output and fine grained debug macros
v0.4.0-rc0
Renamed bfq.c
to gufi_query.c
Split bfwi.c
into gufi_dir2index.c
, gufi_dir2trace.c
, and gufi_trace2index.c
gufi_trace2index.c
wasbfwi_ingest.cpp
Replaced thread pool library in gufi_*.c
with a queue-per-thread thread pool
- Added tests
Added make benchmark
target to CMake
- generates a tree using CMake options
- indexes the tree
- queries the tree (get everything)
Removed C++ requirement for core functionality
Removed make_testdirs.c
Cleaned up memory leaks
Unified opendb
s into opendb2
, separate from the one in dbuitls.h
gufi_find
updates
-size
can be used twice to create a range to search--size%
can be used to create a range to search, given a single-size
- default selected columns is now the columns instead of
*
--aggregate
is now a flag, and does not require any arguments- multiple
--group_by
s can be provided
Documentation updates
v0.3.0-rc0
All processed directories now have databases when being generated. Missing databases should not cause errors during queries.
Summary tables are now populated.
All paths provided to bfq are pushed onto the queue before queue processing begins rather than fully processing one path before pushing the next one onto the queue.
v0.2.0-rc2: Build release version of GUFI when tagged
Miscellaneous changes - capitalization, indentation, ordering, etc
v0.2.0-rc1
use CentOS 7 to generate RPMs
v0.2.0-rc0: Fixed RPM dependencies
Require rpm-build