Skip to content

Commit 5776aad

Browse files
deeprobinSingleAccretionradicaljakobbotschGrabYourPitchforks
authored
Sfh fix tests (#8)
* Use the signature types when building ABI info (#70635) * Use signature types when building the ABI info This will allow us to let "mismatched" struct types as call arguments, which in turn is expected to simplify and de-pessimize much of the code working around the need to keep precise handles on struct nodes. Credit to @jakobbotsch for being able to make this change. * Work around the "InferOpSizeAlign" problem Handling it generally requires solving a larger problem of "eeGetArgSizeAlignment" not working well on ARM. (See also the use of "eeGetArgSizeAlignment" in "lvaInitUserArgs") * [wasm] Build WasmAppHost with UseAppHost=false (#70606) * [wasm] Build WasmAppHost with UseAppHost=false - WasmAppHost is included in the WebAssembly.Sdk pack, and doesn't have a platform specific package. - Since, this is being built for x64, the build defaults to using the app host, which means that we get a single binary file that can be run directly. - But this also means that the binary included in the WebAssembly.Sdk platform-agnostic package will target the platform where it was built. - Instead, build with UseAppHost=false, and update the wasm host to use the managed assembly instead. And update the RunCommand to use `dotnet exec WasmAppHost.dll`. * fix * JIT: Optimize expansion of indir cell addresses for CFG (#70491) On ARM64 it is expensive to materialize the address of an indirection cell since that requires multiple instructions. This is particular a problem when CFG is enabled where validation + call uses the indir cell twice. This changes the CFG logic to create a local in this case. We also were forcing the indir cell argument node into the register. This is not ideal for this particular case, so remove this logic and let LSRA deal with it. ARM still needs this logic. Fix #65076 * Handle mis-sized structs in XARCH `PUTARG_STK` codegen (#68611) * Handle "mis-sized" structs in XARCH PUTARG_STK codegen Previously, for structs which were not a multiple of TARGET_POINTER_SIZE in size: 1) On x86, we copied them to a local in morph. This was a small CQ issue. 2) On Unix x64, we didn't handle them at all and could read out of bounds in "genStructPutArgUnroll". This change fixes both issues by properly detecting cases where we need to be careful and updating codegen to use correct load sizes. * Add a test * Improve distribution of CoseHeaderLabel hash codes (#70695) * Implement digest one shots for browser implementation * [wasm] Don't use a fixed listening port for the proxy (#70681) Fixes https://github.com/dotnet/runtime/issues/70670 . * Fix SOS tests on 7.0 (#70677) The wcscpy_s in ClrDataAccess::GetRegisterName was failing with an invalid parameter exception because the prefixLen and regLen didn't include the terminating null wchar. * [wasm] Enable some previously failing WBT/AOT tests on windows (#70595) * [wasm] Enable some previously failing WBT/AOT tests on windows Fixes https://github.com/dotnet/runtime/issues/61725 . * [wasm] WBT: Fix a failing satellite assembly test The test fails because it is trying to find the string `got: こんにちは` in the app output, but it sees `?????`. This is due to xharness not setting the console output to UTF8, and then the tests not doing the same. Instead of that, we can move the check to be in the app itself, thus removing the need for any of this. Fixes https://github.com/dotnet/runtime/issues/58709 . * List other architectures and DOTNET_ROOT environment variables in dotnet --info (#70403) * [mono] Remove dead code. (#70671) A variable initialized to false and never modified means we can remove the variable and any code that depends on that variable not being false. This is a consequence of 15ab9f985ed45feaa619df70d288fbd0acd5c45f, where code that assigned the variable was removed. * Avoid throwing Win32Exception from HTTP authentication (#70474) * Avoid throwing Win32Exception from HTTP authentication When server sends malformed NTLM challenge the NT authentication processing would throw an unexpected Win32Exception from HttpClientHandler.Send[Async] calls. This aligns the behavior to WinHTTP handler where the Unauthorized reply with challenge token is returned back to the client. Similarly, failure to validate the last MIC token in Negotiate scheme could result in Win32Exception. Handle it by throwing HttpRequestException instead. * Make the unit test more resilient * Add trace to Negotiate authentication * Dispose connection instead of draining the response * Remove outdated ActiveIssue * Fix usage of GSS_KRB5_CRED_NO_CI_FLAGS_X (#70447) * Fix build detection of GSS_KRB5_CRED_NO_CI_FLAGS_X * Don't use GSS_KRB5_CRED_NO_CI_FLAGS_X on NTLM * Handle GSS_S_UNAVAILABLE error from gss_set_cred_option (NTLM wrapped in Negotiate) * Make the GSSAPI shim work with krb5 1.13 * Preserve the gss_acquire_cred minor status * Use IndexOf in WebUtility (#70700) The IndexOfHtmlDecodingChars method was iterating character by character looking for either a `&` or a surrogate, but then the slow path if one of those is found doesn't special-case surrogates. So, we can just collapse this to a vectorized `IndexOf('&')`, which makes the fast path of detecting whether there's anything to decode much faster if there's any meaningful amount of input prior to a `&`. (I experimented with also using `IndexOf('&')` in the main routine, but it made cases with lots of entities slower, and so I'm not including that here.) * Hoist the invariants out of multi-level nested loops (#68061) * first working version * Skip check of VN hoisting * Account for duplicate blocks * clean up * wip * isCommaTree && hasExcep * revert lsra changes * Update hoisting condition - Only update if node to be hoisted has side-effects and the sibling that is before that throws exception * Change to BasicBlockList * organize preheaders * update hoistedInCurLoop and hoistedInSiblingLoop * Reverse the loop order * cleanup and jit-format * Revert "Minor fix to display IG01 weight correctly" This reverts commit 757120e863b2da188db2593da1b7142fd1ecf191. * simplify code * Remove m_hoistedVNInSiblingLoop * Add back ResetHoistedInCurLoop Fix igWeight * Remove reversal of loop processing order * jit format * Experimental: Also generate PerfScore: * review feedback * fix the superpmi script * Revert superpmi asmdiffs change * Rename method * Add a comment * Delete GTF_ASSERTION_PROP_LONG (#70521) * Convert exception help context parsing to managed (#70251) * Move help link parsing to managed * Cleanup wtoi and nativeIsDigit * Fix metasig declaration * Guard GetHelpContext under FEATURE_COMINTEROP * Remove definition of GetHelpLink and guard more methods under cominterop * DWORD should be uint * Add help context marshaling test * Adjust marshaling test * Fix #ifdef with ILLink * Fix method signature mismatch * Specify string marshaling * Throwing test should not test successful HRESULT * Implement ISupportErrorInfo * Test interface in InterfaceSupportsErrorInfo * Delete tests for .NET Framework Xsltc.exe tool (#70706) * Use regex matching for Xsltc test baseline files * Delete xsltc.exe tests entirely * Reenable C4244 in repo (#70026) * Reenable 4244 warning for Mono. * Enable compiler errors for gcc/clang. * Fix issues enabling -Werror=implicit-int-conversion * Revert turning on implicit-int-conversion as error. This check was triggering failures on parts of the PAL that require additional scrutiny beyond the scope of this work. * Add SlidingWindow and FixedWindow to RateLimitPartition (#68782) * align GC memory load calculation on Linux with Docker and Kubernetes (#64128) * align memory load calculation in GC with Docker and Kubernetes * pass inactive file field name as a parameter * [wasm] Fix Debug configuration builds (#70683) * Fix cmake error ``` Manually-specified variables were not used by the project: CONFIGURATION_WASM_OPT_FLAGS ``` * Build the interpreter with -O1 on Wasm in Debug configs Otherwise `interp_exec_method` and `generate_code` can easily overflow the stack in some browsers with even a few recursive calls (for example during .cctor initializaiton) * [wasm] Make runtime_is_initialized promise callbacks one-shot (#70694) * [wasm] Make runtime_is_initialized promise callbacks one-shot Throw if runtime_is_initialized_resolve or runtime_is_initialized_reject is called more than once * Add a slightly generalized GuardedPromise<T> object Protects against multiple-resolve, multiple-reject, reject after resolve and resolve after reject. Does not protect against the executor throwing. * Do not use ExecutableMemoryAllocator on 64 bit platforms if default base address usage is requested (#70563) * [wasm][debugger] Implement get bytes from loaded_files using debugger protocol. (#69072) * Implement get bytes from loaded_files using debugger protocol. * fix pdb size == nul * Adressing @radical comments. * Fix build. * fix compilation * Addressing @radical comments. Co-authored-by: Ankit Jain <[email protected]> * Add doc on Unix temporary file security practice (#70585) * Add doc on Unix temporary file security practice * Update unix-tmp.md * Update unix-tmp.md * Add example permissions encoding * Update docs/design/security/unix-tmp.md Co-authored-by: Dan Moseley <[email protected]> * Update unix-tmp.md Co-authored-by: Dan Moseley <[email protected]> * ConfigurationBinder.Bind on virtual properties duplicates elements (#70592) * ConfigurationBinder.Bind on virtual properties duplicates elements * simplify GetAllProperties * return array instead of list * Obsolete AssemblyName.CodeBase, AssemblyName.EscapedCodeBase (#70534) * JIT: Enable JitConsumeProfileForCasts by default (#69869) * Fix AttributePresence in composite mode (#70737) According to my comparative perf measurements composite framework spends about 5M more instructions in the method coreclr.dll!CMiniMdTemplate<CMiniMd>::SearchTableForMultipleRows(CMiniColDef sColumn) I tracked this down to the R2R table AttributePresence which is used to speed up attribute queries against the ECMA metadata model. In composite mode we were putting the table in the image header, not component header, and so the runtime was unable to locate it. This change fixes generation of the table in Crossgen2; I have verified that with this change I no longer see the 5M outlier. Thanks Tomas * Handle mis-sized structs in ARM/64 `PUTARG_SPLIT` codegen (#70249) * Add a test * Fix out-of-bounds loads in ARM/64's "genPutArgSplit" * Split the test out * Refactor how implicit byrefs are morphed (#70243) * Refactor implicit by-refs morphing Move it to "fgMorphLocal". Enabled folding in "LocalAddressVisitor". * Add FEATURE_IMPLICIT_BYREFS And use it in all the necessary places. Resolves the TP regression on Unix x64, turning it into a TP improvement. * Add a zero-diff quirk * Do not change `gtRetClsHnd` in `impNormStructVal` (#70699) * Do not change "gtRetClsHdl" in "impNormStructVal" Doing so leads breaking the proper ABI handling for the call. * Add a test * Support `PUTARG_STK(STRUCT LCL_VAR/LCL_FLD)` on ARM/64 (#70256) * Separate out LowerPutArgStk Call it from "LowerArg" instead of "ContainCheckCallOperands", as it does more than just containment analysis. * Support "PUTARG_STK(STRUCT LCL_VAR/LCL_FLD)" on ARM/64 To test this, transform "OBJ(LCL_VAR|FLD_ADDR)" to "LCL_FLD" on lowering. This additionally simplified the codegen code as it doesn't have to support two representations of the same thing. * Support `PUTARG_STK(STRUCT LCL_VAR/LCL_FLD)` on x86/Unix x64 (#70702) * Add GenTree::GetLayout * x86/Unix x64: lowering * x86/Unix x64: "genConsumePutStructArgStk" * x86/Unix x64: "genStructPutArgUnroll" * x86/Unix x64: codegen asserts * x86: "genStructPutArgPush" * Unix x64: "genStructPutArgPartialRepMovs" * Allow `TYP_STRUCT` `LCL_FLD` on the RHS of block copies (#70633) * Delete the unused "GTF_USE_FLAGS" It used to indicate that a branch operation didn't need to materialize its operand and could just "use flags" instead, but that purpose has long been lost now that we have explicit SETCC nodes in lowering. * Make GTF_DONT_EXTEND a shared flag So that it can be used for "LCL_FLD" as well as "GT_IND". No diffs. * Enable TYP_STRUCT on the RHS * fgMorphBlockOperand * Tweak TYP_STRUCT LCL_FLD costs Model it as two load, like OBJ. Note we could be more precise here, by using the register type of the layout. For now, we defer. * Block CSE Preserve previous behavior to avoid diffs. * Revert "Fix usage of GSS_KRB5_CRED_NO_CI_FLAGS_X (#70447)" (#70747) This reverts commit 84f7cad00ad834c365b5cd1297e1166525146b50. * Add note about backward branch constraints to ECMA-335 augments (#70760) * [NativeAOT] Enabling return address hijacking on ARM64 Windows (#70740) * Enabling return address hijacking on ARM64 Windows * pass right flags to the GcInfoDecoder * Fix value numbering of HWI loads (#70621) * Fix numbering of HWI loads Take into account all operands and describe the exception sets precisely. * Add a test * Recast the check * Small MsQuicStream refactorings (#70433) * Inline state transition helpers Fixes #55437 * Add high level comments for HandleEventReceive and ReadAsync * Updating HWIntrinsicInfo::lookupId to not accelerate Vector256 APIs when AVX2 is unsupported (#70686) * Updating HWIntrinsicINfo::lookupId to not accelerate Vector256 APIs when AVX2 is unsupported * Fixing a check in lookupId to properly negate the condition * Ensure the special-cased EA_32BYTE constants only happen when AVX/AVX2 are supported * Fixing a bad assert in morph * Dsiable CertificateValidationRemoteServer.ConnectWithRevocation_WithCallback on Android (#70768) * NativeAOT Unlock assignability comparison in reflection-free (#70726) * NativeAOT Unlock assignability comparison in reflection-free Closes #69960 * Document support for `IsAssignableFrom` * Document support for `IsInstanceOfType` * Document `IsAssignableTo` * Add suggestion from Michal * Update src/coreclr/nativeaot/System.Private.DisabledReflection/src/Internal/Reflection/RuntimeTypeInfo.cs Co-authored-by: Michal Strehovský <[email protected]> * Remove whitespaces Co-authored-by: Michal Strehovský <[email protected]> * [Mono] Fix C4018 round I (#70417) * First round of change of fixing C4018 * Address part of review feedback * Change the type of idx to unsigned for `effective_table_slow` * Add idx range check after the type was changed * Address review feedback for `class-init.c` * Change the return type of `*table_num_rows*` to `guint32`. Deal with the consequence of return type change of `table_info_get_rows`. Correct the type of a few local variables which store the return of `mono_metadata_token_index`. * Update return type * Address review feedbacks of metadata.c * Fix native crash * Make counter private to for-loop * Address review feedbacks * Address review feedbacks * Fix race condition in LoaderAllocator::CompareExchangeValueInHandle (#70765) The method was not actually using interlocked operation. It made it possible for multiple different values to win. Fixes #70742 * Enable redirection on arm64 (#70769) * Handle HW exceptions on Windows without redirection (#70428) This change modifies the way hardware exceptions are propagated from the vectored exception handler. Until now, runtime was returning from the vectored exception handler with instruction pointer in the context set to an asm helper. That redirected the thread to that helper and we have raised an exception from its call chain. While working on CET support, it was found that propagating exceptions from the vectored exception handler directly works just fine. So this change makes it work that way for all Windows targets. * Predicate for valid Tracker Reference is inverted (#70709) * Predicate for valid Tracker Reference is inverted * Fix test to not inline wrapper creation function. * JIT: Enable addressing modes for gc types (#70749) * JIT: Faster long->int overflow checks (#69865) * Update .net version for asp.net spmi collection script (#70778) * Enable IlcTrimMetadata by default (#70201) Enables more aggressive metadata stripping by default. In this mode, the AOT compiler only generates reflection metadata for artifacts that were deemed reflection-visible by either dataflow analysis, or user inputs. Previously we would consider everything that got generated reflection-visible. * Change the defaults within the compiler (making the conservative mode opt-in). * Add tests for scenarios in https://github.com/dotnet/runtimelab/issues/656 (resolves dotnet/runtimelab#656). * Update tests * Do more marking where it was necessary * Remove GSS_KRB5_CRED_NO_CI_FLAGS_X code (#70772) The support for building with GSS_KRB5_CRED_NO_CI_FLAGS_X was broken for quite some time. Attempts to reenable it failed due to bug in the krb5 GSSAPI implementation resulting in invalid memory accesses. * Make ConnectWithRevocation_StapledOcsp more resilient to being slow * Upgrade SDK to Preview 5 (#70117) Upgrade to released Preview 5 build * Find&replace FastInterlock* with Interlocked* (#70784) FastInterlock* operations were #defined as aliases to regular Interlock* operations. Deleted the FastInterlock* aliases and replaced usage with Interlocked* operations. * JIT: rework optCanonicalizeLoop (#70792) Rework loop canonicalization in anticipation of extending it to handle non-loop edges like those seen in #70802 (details in #70569). The code to fix up flow out of `h` was not needed and has been removed. We now assert upstream that no such fix up will be needed. * Avoid infinite loop in case of heap corruption (#70759) Co-authored-by: Juan Hoyos <[email protected]> Co-authored-by: Noah Falk <[email protected]> * [wasm][debugger] Remove workaround to get pauseOnExceptions setting (#70748) * Add SubstitutionParser that uses XPath Navigator into NativeAOT (#64671) * Add end-to-end test for NTLM/Negotiate authentication against fake server (#70630) * Add end-to-end test for NTLM/Negotiate authentication against fake server * Simplify the test since HttpClientHandler is always SocketsHttpHandler for the test environment * Fix test condition * Remove extra comment * Enable ThreadAbort with CET enabled via the QueueUserAPC2 (#70803) This change makes ThreadAbort work when CET is enabled. Instead of thread redirection, it uses the new user mode APC mechanism to get a thread to a handler and then throws the ThreadAbortException from there if the thread was interrupted at a safe location. I have verified it works using mdbg tests and also by manual testing in the Visual Studio 2022 using a test app that creates an instance of classes with properties containing infinite loop, wait on a lock, wait on a handle, infinite loop inside of a lock and infinite loop in finally. For the testing, I've enabled this separately from the CET so that the missing support for CET in the debugger code doesn't cause troubles. So we could enable this without CET enabled too, but I'd prefer doing that separately. * Test additional NativeAOT Lib testing (#70212) * Test additional NativeAOT Lib testing * more libraries to un nativeaot rolling build * FB * FB2 * only write results file if specified in args * excluding failing tests * oops, missed pull before a forced push * excluding some recently added tests that fail * fix typo with end element * FB * Fix emitDispJumpList for Arm64 (#70542) * fix jitdump * Fix arm build * Another format * Fix regression in runtime-jit-experimental (#70797) The newly-introduced `emitRemoveJumpToNextInst` optimization caused a regression when hot/cold-splitting, where jumps from the last hot instruction to the first cold instruction were erroneously removed. This is fixed by disabling the `isRemovableJmpCandidate` flag for branches between hot/cold sections. On an unrelated note, a JIT dump message has been added to indicate stress-splitting is occurring. * JIT: relax fwd sub restriction on changing class handle (#70587) For HW SIMD types, at least. Fixes #64879. * Use LastIndexOf in ZipArchiveEntry.GetFileName_Unix (#70701) There's not particularly good reason to open-code the loop here. * Update dogfooding instructions (#70833) - Delete note about multilevel lookup. It is not relevant anymore. - Fix nightly feed url to net7 - Replace .NET Core with just .NET * Add icon to dotnet.exe on Windows (#70578) * Add two new stages to prepare for our new custom type marshalling design. (#70598) * Simplify `Environment.IsWindows8OrAbove`. (#70818) * Simplify Environment.IsWindows8OrAbove. Since .NET 5, the regular Windows version detecton code always returns the correct version. * Remove two unused interop files. * Delete redundant aliases for primitive types (#70805) * Add limited support for LocalCertificateSelectionCallback for QUIC (#70716) * Inline state transition helpers Fixes #55437 * Add high level comments for HandleEventReceive and ReadAsync * [QUIC] Call `LocalCertificateSelectionCallback` to get client certificate * Code review feedback * JIT: Optimize range checks for X >> CNS and for short/byte indices (#67141) Co-authored-by: SingleAccretion <[email protected]> * Tighten checks in `areArgumentsContiguous` (#70829) * Tighten checks in "areArgumentsContiguous" * Add a test * Narrow cast fix (#70518) * Added genActualTypeSize. Remove narrow cast if the actual type sizes are the same * Trying to fix build * Fixing build * Removed genActualTypeSize. Using genActualType instead * Added helper function * Comments * Moving back to morph * Removing part of the test * Added fgOptimizeCastOnAssignment * Fixing build * Removing extra bits * Removed fgIsSafeToRemoveIntToIntCastOnAssignment, inlining the implementation, adding extra logic when actual types are not the same * [wasm] Disable WBT test failing on windows (#70808) Issue: https://github.com/dotnet/runtime/issues/70675 * Enable IDE1005 (Delegate invocation can be simplified) (#70522) * Disable long running test (#70843) * [wasm] Enabled WriteToReadOnly tests - emscripten bug got fixed (#70814) Fixes #53021. Enabled tests from #53021 - they are passing now because issue reported by @radekdoulik got fixed: https://github.com/emscripten-core/emscripten/issues/14299. * Use IndexOf{Any} in a few more places (#70176) * Use IndexOf{Any} in a few more places * Address PR feedback * Fix corerun for assemblies with mismatched filename (#70800) corerun gracefully handled filename not matching the assembly name. Handling of this case regressed in #68186. This change is fixing the regression. Fixes #68455 * Fix a few Stream-related issues in System.Net.Http (#70731) * Fix a few Stream-related issues in System.Net.Http * Put back WriteTimeout * Fix tests * Catch (more) mismatched args in `fgMorphArgs` (#70846) * Catch (more) mismatched args in "fgMorphArgs" * Add a test * [coop] fix coop suspend timeout cue card (#70828) the BLOCKING_SUSPEND_REQUESTED state is treated as suspend in full coop mode (the thread keeps running, but by definition it's not allowed to access managed resources and it will self-suspend if it tries to enter GC Unsafe mode by calling a runtime API or managed code). It is bad in hybrid suspend mode (the thread should be preemptively suspended, but we timed out before the signal handler had a chance to run). The corresponding suspension logic in the code is: https://github.com/dotnet/runtime/blob/3fc61ebb562afc327a8fc6de5c82d76e86bf6f5d/src/mono/mono/utils/mono-threads.c#L1149-L1158 * Fixing the handling of Positive NaN in Math.Min for float/double (#70795) * Adding tests validating Positive NaN for Max, MaxMagnitude, Min, and MinMagnitude * Fixing the handling of Positive NaN in Math.Min for float/double * Fixing the Max/Min code comments to use greater and lesser * Adding a code comment clarifying the sign toggling behavior * JIT: break loop canonicalization into two stages (#70809) For a given loop, we need to separate out the true backedge, any non-loop backedges, and any inner loop backedges so that they all target distinct blocks. Otherwise, we may violate assumptions that the loop entry dominates all blocks in the loop and that all backedges that reach top come from within the loop. This seems simplest to do with two rounds of canonicalization, one that moves the non-loop edges, and another that moves the true backedge. Fixes #70802. * Use `Array.Empty()` in `System.Reflection.Metadata`. (#70862) It is now available in all frameworks it targets. * Delete StaticallyLinked property from Microsoft.NETCore.Native.Unix.props (#70854) This option has many issues. Anybody trying to experiment with static linking can add `<LinkerArg Include="-static" />` into the local file. * Restore ArchivingUtils.AttemptSetLastWriteTime (#70617) * Fix GC stress failure in LoaderAllocator::CompareExchangeValueInHandle (#70832) InterlockedCompareExchangeT has to be called on a raw Object* to make the GC stress infrastructure happy. * Small performance cleanups in S.S.Cryptography * ARM64 - Optimize `i % 2` (#70599) * Assign proper VNs to "shared constant" CSE defs (#70852) * Assign proper VNs to shared CSE defs They must be those of the original expression, not the "base" constant CSE creates. * Add a test * Enable IDE0054 (Use compound assignment) (#70564) * Enable IDE0054 (Use compound assignment) * Update src/libraries/System.Data.Common/src/System/Data/Common/StringStorage.cs Co-authored-by: Tanner Gooding <[email protected]> Co-authored-by: Tanner Gooding <[email protected]> * Use new byte[] span optimization in a few more places (#70665) * Use new byte[] span optimization in a few more places Separated out of larger change to use CreateSpan (these don't rely on that). * Address PR feedback * Improve TLS1.3 detection in registry for QUIC (#70730) * Improve TLS1.3 detection in registry for QUIC * Split client and server detection * Code review feedback * Generate method bodies for delegate Invoke methods (#70883) There is a dataflow warning suppression in System.Linq.Expressions that assumes we'll always have an invocable method body for delegate Invoke method. We need one in IL. We don't in native code. Emulate what IL Linker does and generate a method body. This is a size regression. Suppression: https://github.com/dotnet/runtime/blob/3b2883b097a773715ca84056885e0ca1488da36e/src/libraries/System.Linq.Expressions/src/System/Dynamic/Utils/TypeUtils.cs#L906-L912 Fixes #70880. * Avoid crashing on unresolved dependencies (#70871) Fixes #70815. * Enable IDE0020 (Use pattern matching) (#70523) * Enable IDE0020 (Use pattern matching) * Update src/libraries/System.Private.Xml/src/System/Xml/Serialization/XmlSerializationWriter.cs Co-authored-by: Buyaa Namnan <[email protected]> * Update variable naming Co-authored-by: Buyaa Namnan <[email protected]> * Fix failure building two libraries tests (#70881) Microsoft.CSharp was hitting an issue due to a vararg constructor. We don't support varargs. They shouldn't make it into the dependency graph. We are already checking in many places. We were erroneously thinking a vararg constructor with no mandatory arguments is a default constructor. Runtime.InteropServices were crashing because we got a MulticastDelegate type into a codepath that expects a MulticastDelegate descendant. * Use hash/hmac one shots in NTLM (#70857) Co-authored-by: Stephen Toub <[email protected]> * Unpin locals, attempt 2 (#70655) Contributes to #40553 * Replace a few instances of PtrToStructure with more efficient marshalling (#70866) * Fix for timer scheduling happening on the incorrect thread when wasm threading is enabled (#70863) * Add event to capture min/max threads (#70063) Event ThreadPoolMinMaxThreads added. Parameters are: ushort MinWorkerThreads ushort MaxWorkerThreads ushort MinIOCompletionThreads ushort MaxIOCompletionThreads ushort ClrInstanceID It is fired in the ThreadPool constructor and in the SetMinThreads/SetMaxThreads functions. * [wasm] Add support for per-project customization of helix work items (#70461) * [wasm] Add support for per-project customization of helix work items Currently, for sending tests to helix: 1. the project binaries are zipped, as part of the project build 2. then separate helix targets files build that adds helix items for those zip files. - for wasm, we test with multiple 'scenarios' - like v8, browser, and nodejs. - so, 3 helix work items are submitted per zip file - If a test project needs to have any customizations, for example, for testing crypto with a managed implementation, and subtlecrypto, which would require: - passing different arguments to xharness, in case of managed, and subtlecrypto - and this should be done only for the browser case - Currently, this would need this would need to be done in `sendtohelix-wasm.targets` special casing for the test project, and scenario. - We add support for importing `$(ProjectName).helix.targets`, and calling a special target in that to add helix items, as needed. - This targets file can be set in the test project like: ```xml <HelixTargetsFile Condition="'$(TargetOS)' == 'Browser'">wasm.helix.targets</HelixTargetsFile> ``` - it will get deployed next to the zip file, and picked up automatically ```xml <Project> <PropertyGroup> <_CryptoProjectName>System.Security.Cryptography.Tests</_CryptoProjectName> <System_Security_Cryptography_Tests_TargetName Condition="'$(Scenario))' == 'WasmTestOnBrowser'">System_Security_Cryptography_Tests_Targ et</System_Security_Cryptography_Tests_TargetName> </PropertyGroup> <Target Name="System_Security_Cryptography_Tests_Target"> <ItemGroup> <HelixWorkItem Include="$(Scenario)-managed-$(_CryptoProjectName)"> <PayloadArchive>$(TestArchiveTestsDir)$(_CryptoProjectName).zip</PayloadArchive> <Command>$(HelixCommand)</Command> <Timeout>$(_workItemTimeout)</Timeout> </HelixWorkItem> <HelixWorkItem Include="$(Scenario)-subtlecrypto-System.Security.Cryptography.Tests"> <PayloadArchive>$(TestArchiveTestsDir)$(_CryptoProjectName).zip</PayloadArchive> <Command>$(HelixCommand)</Command> <Timeout>$(_workItemTimeout)</Timeout> <PreCommands Condition="'$(OS)' == 'Windows_NT'">set &quot;WasmXHarnessArgs=--web-server-use-cors&quot;</PreCommands> <PreCommands Condition="'$(OS)' != 'Windows_NT'">export &quot;WasmXHarnessArgs=--web-server-use-cors&quot;</PreCommands> </HelixWorkItem> </ItemGroup> </Target> ``` - The targets file *must* have these: - a property named like `System_Security_Cryptography_Tests_TargetName`, for a test project named `System.Security.Cryptography.Tests` - if this property is not set, then the default behavior of adding work items will run - The target should add any items it needs to to `@(HelixWorkItem)`. - Examples of adding these can be seen in `sendtohelix*` project files - Remember that all these properties, and targets get imported into the msbuild *global* namespace, so make sure to use unique names to avoid conflicts with other test projects. Future work: this commit only enables it for wasm/library tests, but it should be easy to extract it out, but needs some testing. * [wasm] Helix: test with, and without subtle crypto * disable non-wasm builds * address review feedback * Address review feedback * Fix typo * Revert "disable non-wasm builds" This reverts commit 7ef99e81f82200189dd3f61eeaf00d6ca4ced6d4. * Update src/libraries/System.Security.Cryptography/tests/wasm.helix.targets Co-authored-by: Eric Erhardt <[email protected]> * Address review feedback * remove debug spew * Change the way helix extension targets are discovered. The new approach: To run a custom project specific target for adding/editing @(HelixWorkItem): - In the project add: `<HelixTargetsFile Condition="'$(TargetOS)' == 'Browser'">wasm.helix.targets</HelixTargetsFile>` - This file gets copied next to the test archive as $(MSBuildProjectName).helix.targets - In this `wasm.helix.targets` file, add to $(HelixExtensionTargets) to run your custom target ```xml <PropertyGroup Condition="'$(IsRunningLibraryTests)' == 'true' and '$(Scenario)' == 'WasmTestOnBrowser'"> <HelixExtensionTargets>$(HelixExtensionTargets);_AddHelixCrypoItems</HelixExtensionTargets> ``` - The extension target will be called after the default items are added to `@(HelixWorkItem)`. - Useful properties to condition on: $(Scenario), $(IsRunningLibraryTests) - And add to, change, or remove from @(HelixWorkItem) Example: ```xml <Target Name="_AddHelixCrypoItems"> <ItemGroup> <!-- remove the existing item --> <HelixWorkItem Remove="@(HelixWorkItem)" Condition="'%(OriginalFileName)' == '$(_CryptoProjectName)'" /> <!-- add two new ones - managed, and subtylecrypto --> <HelixWorkItem Include="$(WorkItemPrefix)managed-$(_CryptoProjectName)"> <PayloadArchive>$(TestArchiveTestsDir)$(_CryptoProjectName).zip</PayloadArchive> <Command>$(HelixCommand)</Command> <Timeout>$(_workItemTimeout)</Timeout> <OriginalFileName>$(_CryptoProjectName)</OriginalFileName> </HelixWorkItem> <HelixWorkItem Include="$(WorkItemPrefix)subtlecrypto-$(_CryptoProjectName)"> <PayloadArchive>$(TestArchiveTestsDir)$(_CryptoProjectName).zip</PayloadArchive> <Command>$(HelixCommand)</Command> <Timeout>$(_workItemTimeout)</Timeout> <OriginalFileName>$(_CryptoProjectName)</OriginalFileName> <PreCommands Condition="'$(OS)' == 'Windows_NT'">set &quot;WasmXHarnessArgs=%WasmXHarnessArgs% --web-server-use-cop&quot;</PreCommands> <PreCommands Condition="'$(OS)' != 'Windows_NT'">export &quot;WasmXHarnessArgs=$WasmXHarnessArgs --web-server-use-cop&quot;</PreCommands> </HelixWorkItem> <_CryptoHelixItem Include="@(HelixWorkItem)" Condition="$([System.String]::new('%(HelixWorkItem.Identity)').EndsWith('-$(_CryptoProjectName)'))" /> </ItemGroup> <Error Text="Something went wrong. Expected to have only two work items for $(_CryptoProjectName). But got @(_CryptoHelixItem)" Condition="@(_CryptoHelixItem->Count()) != 2" /> </Target> ``` * cleanup * Move WBT specific stuff into the target too. This will make it simpler to move it off into a targets file later * fix build * fix libtests * fix wbt * [wasm] Bump helix timeout to 90mins for debugger tests on windows Co-authored-by: Eric Erhardt <[email protected]> * Fix bug in Tar preventing extraction of hardlinks or entries starting with `.\` (#70853) * Add PlatformDetection.SupportsHardLinkCreation property. * Fix how paths are combined/joined and sanitized on extraction, to ensure paths with redundant segments get properly handled. * Add tests that verify archives with entries whose paths start with .\, including the root folder itself. * Re-enable the hardlink test, condition it to not run if platform does not support extraction of hardlinks. * Remove unnecessary test - This same code is already being tested by TarReader_ExtractToFile_Tests.ExtractEntriesWithSlashDotPrefix * Reuse test code that retrieves memory stream. * Bump test data package version * Add missing typeof(PlatformDetection) in ConditionalFact Co-authored-by: carlossanlop <[email protected]> * Remove #define Sleep (#70868) * Remove #define Sleep * Fix ICorDebugFunction2::GetVersionNumber for arm32. (#69901) * Replace the remaining uses of Marshal.PtrToStructure in core networking (#70900) * Rename ICMP interop to match Windows SDK names * Add support for delegate GDV and method-based vtable GDV (#68703) Add support for instrumenting delegate calls and vtable calls into method handle histograms. Use these histograms to do GDV for delegate calls and also support method-based GDV for vtable calls. For instrumentation we now support class probes at interface call sites, method probes at delegate call sites and both class probes and method probes at vtable call sites. For vtable calls, when turned on, instrumentation produces both histograms as PGO data so that the JIT can later make the choice about what is the best form of guard to use at that site. For guarding, there are some things to take into account. Delegate calls currently (practically) always point to precode, so this PR is just guarding on getFunctionFixedEntryPoint which returns the precode address, and this is generally quite cheap (same cost as class-based GDV). That's the case for delegates pointing to instance methods anyway, this PR does not support static methods yet -- those will be more expensive. For vtable calls the runtime will backpatch the slots when tiering, so the JIT guards the address retrieved from the vtable against an indirection of the slot, which is slightly more expensive than a class-based guard. Currently the instrumentation is enabled conditionally with COMPlus_JitDelegateProfiling=1 (for delegates) and COMPlus_JitVTableProfiling=1 (for vtable calls). Currently delegate profiling is turned on by default while vtable profiling is off by default. * Use ArgumentNullException.ThrowIfNull in a few more places (#70897) * Append the function pointer type (#70923) The issue here was the recursion into PrettyPrintSigWorkerInternal that would reset the buffer. Now, we pass in a new buffer and append it when we return. * [main] Update dependencies from dotnet/runtime dotnet/icu dotnet/xharness dotnet/runtime-assets dotnet/emsdk dotnet/roslyn-analyzers (#70476) * Update dependencies from https://github.com/dotnet/runtime-assets build 20220608.1 Microsoft.DotNet.CilStrip.Sources , System.ComponentModel.TypeConverter.TestData , System.Drawing.Common.TestData , System.Formats.Tar.TestData , System.IO.Compression.TestData , System.IO.Packaging.TestData , System.Net.TestData , System.Private.Runtime.UnicodeData , System.Runtime.Numerics.TestData , System.Runtime.TimeZoneData , System.Security.Cryptography.X509Certificates.TestData , System.Text.RegularExpressions.TestData , System.Windows.Extensions.TestData From Version 7.0.0-beta.22281.1 -> To Version 7.0.0-beta.22308.1 * Update dependencies from https://github.com/dotnet/emsdk build 20220608.2 Microsoft.NET.Workload.Emscripten.Manifest-7.0.100 From Version 7.0.0-preview.6.22281.1 -> To Version 7.0.0-preview.6.22308.2 * Update dependencies from https://github.com/dotnet/xharness build 20220610.1 Microsoft.DotNet.XHarness.CLI , Microsoft.DotNet.XHarness.TestRunners.Common , Microsoft.DotNet.XHarness.TestRunners.Xunit From Version 1.0.0-prerelease.22305.1 -> To Version 1.0.0-prerelease.22310.1 * Revert "Update dependencies from https://github.com/dotnet/emsdk build 20220608.2" This reverts commit bbb4e156ddbad8a2cb7b604246214272085b6622. * Update dependencies from https://github.com/dotnet/icu build 20220609.1 Microsoft.NETCore.Runtime.ICU.Transport From Version 7.0.0-preview.6.22306.1 -> To Version 7.0.0-preview.6.22309.1 * Update dependencies from https://github.com/dotnet/emsdk build 20220608.2 Microsoft.NET.Workload.Emscripten.Manifest-7.0.100 From Version 7.0.0-preview.6.22281.1 -> To Version 7.0.0-preview.6.22308.2 * Update dependencies from https://github.com/dotnet/runtime-assets build 20220610.1 Microsoft.DotNet.CilStrip.Sources , System.ComponentModel.TypeConverter.TestData , System.Drawing.Common.TestData , System.Formats.Tar.TestData , System.IO.Compression.TestData , System.IO.Packaging.TestData , System.Net.TestData , System.Private.Runtime.UnicodeData , System.Runtime.Numerics.TestData , System.Runtime.TimeZoneData , System.Security.Cryptography.X509Certificates.TestData , System.Text.RegularExpressions.TestData , System.Windows.Extensions.TestData From Version 7.0.0-beta.22281.1 -> To Version 7.0.0-beta.22310.1 * Update dependencies from https://github.com/dotnet/roslyn-analyzers build 20220610.1 Microsoft.CodeAnalysis.NetAnalyzers From Version 7.0.0-preview1.22302.1 -> To Version 7.0.0-preview1.22310.1 * Update dependencies from https://github.com/dotnet/runtime build 20220612.5 Microsoft.NET.Sdk.IL , Microsoft.NETCore.App.Runtime.win-x64 , Microsoft.NETCore.DotNetHost , Microsoft.NETCore.DotNetHostPolicy , Microsoft.NETCore.ILAsm , runtime.native.System.IO.Ports , System.Text.Json From Version 7.0.0-preview.6.22305.4 -> To Version 7.0.0-preview.6.22312.5 * Update dependencies from https://github.com/dotnet/icu build 20220613.1 Microsoft.NETCore.Runtime.ICU.Transport From Version 7.0.0-preview.6.22306.1 -> To Version 7.0.0-preview.6.22313.1 * Update dependencies from https://github.com/dotnet/xharness build 20220613.1 Microsoft.DotNet.XHarness.CLI , Microsoft.DotNet.XHarness.TestRunners.Common , Microsoft.DotNet.XHarness.TestRunners.Xunit From Version 1.0.0-prerelease.22310.1 -> To Version 1.0.0-prerelease.22313.1 * Update dependencies from https://github.com/dotnet/runtime-assets build 20220613.1 Microsoft.DotNet.CilStrip.Sources , System.ComponentModel.TypeConverter.TestData , System.Drawing.Common.TestData , System.Formats.Tar.TestData , System.IO.Compression.TestData , System.IO.Packaging.TestData , System.Net.TestData , System.Private.Runtime.UnicodeData , System.Runtime.Numerics.TestData , System.Runtime.TimeZoneData , System.Security.Cryptography.X509Certificates.TestData , System.Text.RegularExpressions.TestData , System.Windows.Extensions.TestData From Version 7.0.0-beta.22281.1 -> To Version 7.0.0-beta.22313.1 * Update dependencies from https://github.com/dotnet/emsdk build 20220613.1 Microsoft.NET.Workload.Emscripten.Manifest-7.0.100 From Version 7.0.0-preview.6.22281.1 -> To Version 7.0.0-preview.6.22313.1 * [wasm] Wasm.Build.Tests: Disable strict version checks for emcc Due to the way we have to update `dotnet/emsdk`, and get the update PRs in `runtime`, the emsdk version can be mismatched. For example, if `dotnet/emsdk` has already updated to `3.1.12`, but `dotnet/runtime` is still on `3.1.7`. This is an expected scenario while working on updating to a newer emscripten. The version mismatch will still show up, but as a warning now. * Update dependencies from https://github.com/dotnet/xharness build 20220614.1 Microsoft.DotNet.XHarness.CLI , Microsoft.DotNet.XHarness.TestRunners.Common , Microsoft.DotNet.XHarness.TestRunners.Xunit From Version 1.0.0-prerelease.22310.1 -> To Version 1.0.0-prerelease.22314.1 * [wasm] Wasm.Build.Tests: Update Skiasharp reference The skiasharp dependent tests fail with: `error : undefined symbol: _ZNKSt3__220__vector_base_commonILb1EE20__throw_length_errorEv`. Update the reference to the latest package which added a binary compiled for `3.1.7`. * [wasm] Workaround a test failure in WBT Currently, there is a mismatch between emsdk versions with the workload emscripten packs using `3.1.12`, and the runtime being built with `3.1.7`. And that is causing `Wasm.Build.Tests.NativeLibraryTests.ProjectUsingSkiaSharp` to fail with: ``` EXEC : error : undefined symbol: _ZNKSt3__220__vector_base_commonILb1EE20__throw_length_errorEv (referenced by top-level compiled C/C++ code) [/datadisks/disk1/work/B1420981/w/ACC90933/e/blz_nativeref_aot_Debug/blz_nativeref_aot_Debug.csproj] EXEC : warning : Link with `-sLLD_REPORT_UNDEFINED` to get more information on undefined symbols [/datadisks/disk1/work/B1420981/w/ACC90933/e/blz_nativeref_aot_Debug/blz_nativeref_aot_Debug.csproj] EXEC : warning : To disable errors for undefined symbols use `-sERROR_ON_UNDEFINED_SYMBOLS=0` [/datadisks/disk1/work/B1420981/w/ACC90933/e/blz_nativeref_aot_Debug/blz_nativeref_aot_Debug.csproj] EXEC : warning : __ZNKSt3__220__vector_base_commonILb1EE20__throw_length_errorEv may need to be added to EXPORTED_FUNCTIONS if it arrives from a system library [/datadisks/disk1/work/B1420981/w/ACC90933/e/blz_nativeref_aot_Debug/blz_nativeref_aot_Debug.csproj] EXEC : error : undefined symbol: _ZNKSt3__221__basic_string_commonILb1EE20__throw_length_errorEv (referenced by top-level compiled C/C++ code) [/datadisks/disk1/work/B1420981/w/ACC90933/e/blz_nativeref_aot_Debug/blz_nativeref_aot_Debug.csproj] EXEC : warning : __ZNKSt3__221__basic_string_commonILb1EE20__throw_length_errorEv may need to be added to EXPORTED_FUNCTIONS if it arrives from a system library [/datadisks/disk1/work/B1420981/w/ACC90933/e/blz_nativeref_aot_Debug/blz_nativeref_aot_Debug.csproj] EXEC : error : Aborting compilation due to previous errors [/datadisks/disk1/work/B1420981/w/ACC90933/e/blz_nativeref_aot_Debug/blz_nativeref_aot_Debug.csproj] emcc : error : '/datadisks/disk1/work/B1420981/w/ACC90933/e/dotnet-workload/packs/Microsoft.NET.Runtime.Emscripten.3.1.12.Node.linux-x64/7.0.0-preview.6.22313.1/tools/bin/node /datadisks/disk1/work/B1420981/w/ACC90933/e/dotnet-workload/packs/Microsoft.NET.Runtime.Emscripten.3.1.12.Sdk.linux-x64/7.0.0-preview.6.22313.1/tools/emscripten/src/compiler.js /datadisks/disk1/work/B1420981/t/tmp44tn7y2d.json' failed (returned 1) [/datadisks/disk1/work/B1420981/w/ACC90933/e/blz_nativeref_aot_Debug/blz_nativeref_aot_Debug.csproj] ``` This commit adds a temporary workaround to ignore the undefined symbols for this specific test. And this should be removed when runtime is updated to `3.1.12` . Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com> Co-authored-by: Alexander Köplinger <[email protected]> Co-authored-by: Larry Ewing <[email protected]> Co-authored-by: Ankit Jain <[email protected]> * Harden for null byrefs (#70317) * Remove enum_flag_Unrestored usage in boxing stubs. * Add AND instruction (21h) to x86 decoder. * Update mono for null ref in interpreter paths. * Disable test on llvmfullaot and wasm * Handle null destination for intrinsics. * Use u8 in a few more places (#70894) * Use u8 in a few more places * A few more u8s * Enable lib tests that used to fail with GC issues (#70831) * enable lib tests that used to fail with GC issues * excluding linq parallel test since that doesn't build * Threading.Thread tests have issues * Excluding new test failures * Remove some dead code / branches (#70146) * Remove some dead code / branches Some signal amidst the noise in an lgtm.com report. * Address PR feedback * superpmi.py: Add `-jitoption` for asmdiffs (#70939) The new `-jitoption` option passes the argument options to both baseline and diff compilers. This is a convenience option: there already is `-base_jit_option` and `-diff_jit_option` to specify passing options to either baseline or diff. The name of the option is the same as that used for `replay`. * EventLogException is missing the original win32 error code (#70629) * Enable IDE0071 (Simplify interpolation) (#70918) * Enable IDE0100 (Remove redundant equality) (#70896) * Enable IDE0065 (Misplaced using directive) (#70919) * Optimization for full range checks (#70145) (#70222) * Update C# compiler (#70947) Brings in support for required members * Allow ValueListBuilder to work with empty scratch spans (#70917) * Enable IDE0030 (Use coalesce expression) (#70948) * Do not reuse implicit byrefs for by-value args (#70955) * Do not reuse implicit byrefs for by-value args * Add a test * Enable `TYP_STRUCT` `LCL_VAR/LCL_FLD` call args on Windows x64 (#70777) * Fix forward sub * Enable folding in local morph * Morph: TYP_STRUCT LCL_FLD * Make SuperPMI more explicit about JIT in error messages (#70938) This makes it easier to figure out which JIT failed in asmdiffs scenarios. e.g., ``` [10:40:50] ERROR: Method 3673 of size 3107 failed to load and compile correctly by JIT2 (C:\gh\runtime\artifacts\tests\coreclr\windows.x64.Checked\Tests\Core_Root\clrjit_universal_arm64_x64.dll). [10:40:50] ERROR: Method 3673 of size 3107 failed to load and compile correctly by JIT1 (C:\gh\runtime2\artifacts\tests\coreclr\windows.x64.Checked\Tests\Core_Root\clrjit_universal_arm64_x64.dll). ``` * Fix illumos-x64 build (#70956) * Disable W^X on macOS under Rosetta emulation (#70912) * Disable W^X on macOS under Rosetta emulation Apple has informed us that double mapping doesn't work properly on Rosetta emulation. This change disables W^X if Rosetta is detected. * Reflect PR feedback * Add support for randomized guard stack cookies (#70806) The compiler was generating GS cookies, but the cookie was always an ASCII encoding of "Hi, mom!". Make the cookie properly per-process randomized. The randomization algorithm is same as in the CoreCLR VM. Fixes #70071. * Fixes #64159 - JsonSerializerContext source generation fails with keyword identifiers (#66876) * Fixes #64159 - initial implementation * Tidy up, and add test for ignored reserved-keyword-named property * PR feedback - use same approach as logging generator * PR feedback * Update src/libraries/System.Text.Json/gen/JsonSourceGenerator.Parser.cs Co-authored-by: Adeel Mujahid <[email protected]> * PR feedback - rename ExactNameSpecifiedInSourceCode * PR feedback - use extracted (and renamed) local variable * Remove commented code * Added `IsVerbatimName` as extension method * Support fields with verbatim names (@) * Use alternative method for checking for verbatim names * Uses `SyntaxFacts` to determine if escaping is needed * Remove extension method * Modified source generator test to include a verbatim field * Minor typo Co-authored-by: Adeel Mujahid <[email protected]> * [main] Update dependencies from dotnet/arcade (#70662) Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com> * Enable IDE0031 (Use null propagation) (#70965) * Fix nullable annotations on MailAddress.TryCreate (#70943) * [QUIC] Removes abstract providers and mock from S.N.Quic (#70421) * Removed abstract providers and mock from S.N.Quic * Removed provider and replaced with statics on Listener and Connection * Added assert * Stop wrapping SIMD nodes in OBJs (#70888) * Upstream coreclr and pal changes to support powerpc (ppc64le) architecture (#69105) * coreclr pal layer chanegs. * Updated the system arch for power architecture. * Fixed the failing PAL SXS exception_handling test case * replaced gint with gint64 to avoid overflow and this has fixed the segmentation fault issue. * coreclr pal layer chanegs. * Fixed the failing PAL SXS exception_handling test case * coreclr pal layer chanegs. * Updated the system arch for power architecture. * Fixed the failing PAL SXS exception_handling test case * replaced gint with gint64 to avoid overflow and this has fixed the segmentation fault issue. * coreclr pal layer chanegs. * Fixed the failing PAL SXS exception_handling test case * Removing intsafe.h file as main does not have this file now.(Already removed in commit 27195f670937c7e21ab68a806396f9d17c57231a) Co-authored-by: Alhad Deshpande <[email protected]> * Collections contain stale object IDs (#70924) It looks like the collections contain stale ObjectID values. These collection should be purged between GCs. * Minor refactoring for `UpgradeToRegexGeneratorAnalyzer` (#70767) * Fix arm64 funclet frame type 5 (#70922) * Fix arm64 funclet frame type 5 * Add code to stress arm64 funclet frame type 5 This can be forced using `COMPlus_JitSaveFpLrWithCalleeSavedRegisters=3`, or will be enabled under stress. * Add a regression test * Fix funclet frame type 5 alignment calculation * Fix a couple bugs; improve documentation 1. Improve the frame shape pictures, especially for arm64 funclet frames. Include the MonitorAcquired element and OSR pad. 2. Fix bug for `SP_to_PSP_slot_delta` for funclet frame type 4 for OSR: don't include osrPad. 3. Fix bug with funclet frame type 5 for `SP_to_FPLR_save_delta` using wrong aligned func size; would be incorrect if one and two SP adjustment full frame aligned sizes are different (e.g., if two SP adjustment required two alignment slots and one SP adjustment thus required none). * Fix default log color behavior when console is redirected (#70504) * JIT: Model string literal objects as invariant and non-GLOB_REF (#70986) * Model string literals as invariant/non GLOB_REF * String literals are never null * Add support for cross module inlining and cross module generic compilation to Crossgen2 (#68919) * Add support for cross module inlining and cross module generic compilation to Crossgen2 - Refactor Module into ModuleBase and Module - The goal is to have allow a subset version of Module which can only hold refs, this is to be used by the manifest module in an R2R image to allow for version resilient cross module references. - Update handling of ModuleBase so that its used everywhere that tokens are parsed from R2R - Remove ENCODE_MODULE_ID_FOR_STATICS and ENCODE_ACTIVE_DEPENDENCY - These were only used for NGEN, and conflict with easy impelmentation for the ModuleBase concept - Remove locking in ENCODE_STRING_HANDLE processing, and tweak comments. Comments applied to the removed ngen based code, and the lock was only necessary for the old ngen thing. - Adjust ComputeLoaderModuleWorker for locating loader module - Follow comment more accurately, to avoid putting every generic into its definition module. This will make R2R function lookup able to find compiled instantiations in some cases. This may be what we want long term, it may not. - Remove MemberRefToDesc map and replace with LookupMap like the other token types. We no longer make use of the hot table, so this is more efficient - Also reduces complexity of implementation of ModuleBase - Build fixup to describe a single method as a standalone blob of data - There are parallel implementations in Crossgen2 and in the runtime - They produce binary identical output - Basic R2RDump support for new fixup - Adjust module indices used within the R2R format to support a module index which refers to the R2R manifest metadata. This requires bumping the R2R version to 6.2 - Add a module index between the set of assembly refs in the index 0 module and the set of assembly refs in the R2R manifest metadata - Adjust compilation dependency rules to include a few critical AsyncStateMachineBox methods - Remove PEImage handling of native metadata which was duplicative - Do not enable any more devirtualization than was already in use, even in the cross module compilation scenario. In particular, do not enable devirtualization of methods where the decl method isn't within the version bubble, even if the decl method could be represented with a cross-module reference token. (This could be fixed, but is out of scope for this initial investigation) Make the compilation deterministic in this new model, even though we are generating new tokens on demand - Implement this by detecting when we need new tokens during a compile, and recompiling with new tokens when necessary - This may result in compiling the same code as much as twice Compile the right set of methods with cross module inlining enabled - Add support for compiling the called virtual methods on generic types - This catches the List<T> and Dictionary<TKey,TValue> scenarios - Add command line switches to enable/disable the new behavior - By default the new behavior is not enabled * Implement Vector128 version of System.Buffers.Text.Base64 DecodeFromUtf8 and EncodeToUtf8 (#70654) * Implement Vector128 version of System.Buffers.Text.Base64.DecodeFromUtf8 Rework the SS3 into a Vector128 version, and add Arm64 support. * SSE3 improvements * Remove superfluous bitwise And * Add comment to SimdShuffle * Inline SimdShuffle * Implement Vector128 version of System.Buffers.Text.Base64.EncodeToUtf8 * Ensure masking on SSE3 Change-Id: I319f94cfc51d0542ae4eb11a8d48b3eb8180553f CustomizedGitHooks: yes * Restore asserts and move zero inside the loop * Neater C# code Change-Id: I2cbe14f4228f8035e7d213b5b58815c4eee35563 CustomizedGitHooks: yes * Make SimdShuffle consistent across X64 and Arm64 * Better looking multiply * Use `gtEffectiveVal` for `GT_ADD` op1 in `optCreateAssertion` (#70228) * Initial work for comma hoisting in cse * Formatting * Adding more ops to comma hoisting * Set regnum * Update optcse.cpp * Using effectiveval instead * Convert fallback path of GetCommandLineArgs to managed (#70608) * P/Invoke definition * Use P/Invoke in managed code * Update managed call site * Add test using private reflection * Native command line should be superset of managed Co-authored-by: Jan Kotas <[email protected]> * [wasm] Enabled WasmTestOnBrowser run on Windows for Browser. (#70890) * Enabled tests - no timeout with new ems. Keeping disabled tests failing from other reasons. * Added @radical's suggestion. * Test CI- enable test that are throwing in Debug. * [wasm][debugger] Fix side effect on Firefox of getting bytes from loaded_files using debugger protocol (#70990) * Fix 70983 * fix typo * Fix compilation * Add AddSystemd() and AddWindowsService() IServiceCollection extension methods (#68580) * Add AddSystemd() IServiceCollection extension method * Add AddWindowsService() IServiceCollection extension method * Don't default to CWD if in C:\Windows\system32 - instead, when CWD is C:\Windows\system32 Hosting will use AppContext.BaseDirectory. This way Windows apps and services that are launched will work by default. HostApplicationBuilder.ContentRootPath can't be changed after construction, so setting it to a workable default for Windows apps. Co-authored-by: Eric Erhardt <[email protected]> Co-authored-by: Martin Costello <[email protected]> * Use RemoteExecutor * Update src/libraries/Microsoft.Extensions.Hosting/tests/UnitTests/HostTests.cs * Skip test on Windows nano server * Respond to PR feedback Co-authored-by: Eric Erhardt <[email protected]> Co-authored-by: Martin Costello <[email protected]> * [main] Update dependencies from dotnet/runtime dotnet/xharness dotnet/icu dotnet/emsdk (#70991) * Update dependencies from https://github.com/dotnet/runtime build 20220619.5 Microsoft.NET.Sdk.IL , Microsoft.NETCore.App.Runtime.win-x64 , Microsoft.NETCore.DotNetHost , Microsoft.NETCore.DotNetHostPolicy , Microsoft.NETCore.ILAsm , runtime.native.System.IO.Ports , System.Text.Json From Version 7.0.0-preview.6.22312.5 -> To Version 7.0.0-preview.6.22319.5 * Update dependencies from https://github.com/dotnet/xharness build 20220620.1 Microsoft.DotNet.XHarness.CLI , Microsoft.DotNet.XHarness.TestRunners.Common , Microsoft.DotNet.XHarness.TestRunners.Xunit From Version 1.0.0-prerelease.22314.1 -> To Version 1.0.0-prerelease.22320.1 * Update dependencies from https://github.com/dotnet/icu build 20220620.2 Microsoft.NETCore.Runtime.ICU.Transport From Version 7.0.0-preview.6.22313.1 -> To Version 7.0.0-preview.6.22320.2 * Update dependencies from https://github.com/dotnet/emsdk build 20220620.1 Microsoft.NET.Workload.Emscripten.Manifest-7.0.100 From Version 7.0.0-preview.6.22313.1 -> To Version 7.0.0-preview.6.22320.1 Co-authored-by: dotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com> * Windows/Arm64: Use 8.1 atomic instructions if they are available (#70921) * Use Fast compareexchange, acquire/release * working windows version * remove printf * some more #ifdef * fix some #ifdef * optimize EnterObjMonitorHelperSpin * Remove FastInterlockedCompareExchange64 for now * Add ZipArchiveEntry.IsEncrypted for password-protection detection (#70036) * Add ZipArchiveEntry.IsEncrypted for password-protection detection * Apply suggestions from code review Co-authored-by: David Cantú <[email protected]> * Run GenerateReferenceAssemblySource for System.IO.Compression * Revert WriteAsync default parameter value removal * Update DeflateStream.WriteAsync to include the default param value * Revert unrelated change from GenerateReferenceAssemblySource * Revert unrelated change after syncing with GenerateReferenceAssemblySource Co-authored-by: David Cantú <[email protected]> * Exposing IRootFunctions.Hypot and IRootFunctions.Root (#71010) * Exposing IRootFunctions.Hypot and IRootFunctions.Root * Adding tests for IRootFunctions.Hypot and IRootFunctions.Root * Add TarEntry conversion constructors (#70325) * ref: Conversion constructors * src: Conversion constructors * tests: Conversion constructors * [wasm] Wasm.Build.Tests: disable tests failing on CI (#71020) * [wasm] Wasm.Build.Tests: disable tests failing on CI .. due to being oomkill'ed. * [wasm] Don't fail the job when firefox tests fail, as they are known to be unstable * Remove allocation in ImmutableArray Builder Sort (#70850) * Remove allocation in ImmutableArray Builder Sort `ImmutableArray<T>.Builder.Sort(Comparison<T> comparison)` allocates a new instance of `Comparer<T>` using `comparison`. Then `ArraySort` allocates a new delegate from `comparer.Compare`. Both allocations can be eliminated by calling `Array.Sort<T>(T[] array, Comparison<T> comparison)` directly. * Sort only valid range of array * Use MemoryExtensions.Sort where available. * Update src/libraries/System.Collections.Immutable/src/System/Collections/Immutable/ImmutableArray_1.Builder.cs Fix spelling mistake. Co-authored-by: Eirik Tsarpalis <[email protected]> Co-authored-by: Eirik Tsarpalis <[email protected]> * [Android] Make sure AdditionalTimeZoneChecks trait is added to xunit-excludes (#70974) The trait is supposed to be included in System.Runtime tests by default, but wasn't because the _withoutCategories property was in tests.props. As a result, the property was resolved before the .csproj properties. This fix moves the property definition to tests.targets. Fixes #70482 * Use LastIndexOfAny in GetFileName_Windows (#71032) * Avoid unnecessary allocations in SanitizeEntryFilePath (#71034) * Fix Ordinal Ignore Case string compare (#71022) * Move allocation of command line argument array to C# (#71021) Also, avoid redundant allocation of managed string for each command line argument. * Add markdown readme for System.Reflection.MetadataLoadContext (#70610) * Create README.md * set red zone size 512 for powerpc64. (#70885) * Ensure consistent reflectability for generic methods (#70977) If we have a method body for `SomeMethod<Foo>` and `SomeMethod<T>` was decided to be reflection-visible, ensure `SomeMethod<Foo>` is also reflection-visible. * Implement NegotiateAuthentication API (#70720) * WIP: Add implementation of NegotiateAuthentication Switch System.Net.Http to use NegotiateAuthentication Fix IsCompleted in managed NTLM implementation * WIP: Update error code mapping * Spanify input of GetOutgoingBlob * Update comments * Move NegotiateStreamPal.Encrypt/Decrypt to shared sources. Unix implementation already had them and they get trimmed anyway. * Revert accidental change * Build fixes. * Fix error handling condition * Update error mapping based on HttpListener usage. * WIP: HttpListener test * Move workaround from HttpListener to low-level SSPI code * Fix build * Clean up * Revert "WIP: HttpListene…
1 parent 88efa3b commit 5776aad

File tree

2,950 files changed

+57573
-33405
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

2,950 files changed

+57573
-33405
lines changed

.config/dotnet-tools.json

+1-1
Original file line numberDiff line numberDiff line change
@@ -15,7 +15,7 @@
1515
]
1616
},
1717
"microsoft.dotnet.xharness.cli": {
18-
"version": "1.0.0-prerelease.22305.1",
18+
"version": "1.0.0-prerelease.22320.3",
1919
"commands": [
2020
"xharness"
2121
]

Directory.Build.props

+1
Original file line numberDiff line numberDiff line change
@@ -35,6 +35,7 @@
3535
<TargetArchitecture Condition="'$(TargetArchitecture)' == '' and '$(_hostArch)' == 'arm64'">arm64</TargetArchitecture>
3636
<TargetArchitecture Condition="'$(TargetArchitecture)' == '' and '$(_hostArch)' == 'loongarch64'">loongarch64</TargetArchitecture>
3737
<TargetArchitecture Condition="'$(TargetArchitecture)' == '' and '$(_hostArch)' == 's390x'">s390x</TargetArchitecture>
38+
<TargetArchitecture Condition="'$(TargetArchitecture)' == '' and '$(_hostArch)' == 'ppc64le'">ppc64le</TargetArchitecture>
3839
<TargetArchitecture Condition="'$(TargetArchitecture)' == '' and ('$(TargetOS)' == 'Browser' or '$(RuntimeIdentifier)' == 'browser-wasm')">wasm</TargetArchitecture>
3940
<TargetArchitecture Condition="'$(TargetArchitecture)' == '' and '$(TargetsMobile)' == 'true'">x64</TargetArchitecture>
4041
<TargetArchitecture Condition="'$(TargetArchitecture)' == ''">x64</TargetArchitecture>

THIRD-PARTY-NOTICES.TXT

+37
Original file line numberDiff line numberDiff line change
@@ -1099,3 +1099,40 @@ LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
10991099
ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
11001100
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
11011101
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
1102+
1103+
License notice for code from The Practice of Programming
1104+
-------------------------------
1105+
1106+
Copyright (C) 1999 Lucent Technologies
1107+
1108+
Excerpted from 'The Practice of Programming
1109+
by Brian W. Kernighan and Rob Pike
1110+
1111+
You may use this code for any purpose, as long as you leave the copyright notice and book citation attached.
1112+
1113+
License notice for amd/aocl-libm-ose
1114+
-------------------------------
1115+
1116+
Copyright (C) 2008-2020 Advanced Micro Devices, Inc. All rights reserved.
1117+
1118+
Redistribution and use in source and binary forms, with or without modification,
1119+
are permitted provided that the following conditions are met:
1120+
1. Redistributions of source code must retain the above copyright notice,
1121+
this list of conditions and the following disclaimer.
1122+
2. Redistributions in binary form must reproduce the above copyright notice,
1123+
this list of conditions and the following disclaimer in the documentation
1124+
and/or other materials provided with the distribution.
1125+
3. Neither the name of the copyright holder nor the names of its contributors
1126+
may be used to endorse or promote products derived from this software without
1127+
specific prior written permission.
1128+
1129+
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
1130+
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1131+
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
1132+
IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
1133+
INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
1134+
BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
1135+
OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
1136+
WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
1137+
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1138+
POSSIBILITY OF SUCH DAMAGE.

docs/coding-guidelines/clr-code-guide.md

+1-1
Original file line numberDiff line numberDiff line change
@@ -437,7 +437,7 @@ A GC_NOTRIGGER function cannot:
437437

438438
[1] With one exception: GCX_COOP (which effects a preemp->coop->preemp roundtrip) is permitted. The rationale is that GCX_COOP becomes a NOP if the thread was cooperative to begin with so it's safe to allow this (and necessary to avoid some awkward code in our product.)
439439

440-
**Note that for GC to be truly prevented, the caller must also ensure that the thread is in cooperative mode.** Otherwise, all the precautions above are in vain since any other thread can start a GC at any time. Given that, you might be wondering why cooperative mode is not part of the definition of GC_NOTRIGGER. In fact, there is a third thread state called GC_FORBID which is exactly that: GC_TRIGGERS plus forced cooperative mode. As its name implies, GC_FORBID _guarantees_ that no GC will occur on any thread.
440+
**Note that for GC to be truly prevented, the caller must also ensure that the thread is in cooperative mode.** Otherwise, all the precautions above are in vain since any other thread can start a GC at any time. Given that, you might be wondering why cooperative mode is not part of the definition of GC_NOTRIGGER. In fact, there is a third thread state called GC_FORBID which is exactly that: GC_NOTRIGGER plus forced cooperative mode. As its name implies, GC_FORBID _guarantees_ that no GC will occur on any thread.
441441

442442
Why do we use GC_NOTRIGGERS rather than GC_FORBID? Because forcing every function to choose between GC_TRIGGERS and GC_FORBID is too inflexible given that some callers don't actually care about GC. Consider a simple class member function that returns the value of a field. How should it be declared? If you choose GC_TRIGGERS, then the function cannot be legally called from a GC_NOTRIGGER function even though this is perfectly safe. If you choose GC_FORBID, then every caller must switch to cooperative mode to invoke the function just to prevent an assert. Thus, GC_NOTRIGGER was created as a middle ground and has become far more pervasive and useful than GC_FORBID. Callers who actually need GC stopped will have put themselves in cooperative mode anyway and in those cases, GC_NOTRIGGER actually becomes GC_FORBID. Callers who don't care can just call the function and not worry about modes.
443443

+253
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,253 @@
1+
# Mono code guidelines
2+
3+
This document is meant to capture guidelines for contributing code to
4+
the [src/mono/mono/](../../src/mono/mono),
5+
[src/native/public/mono](../../src/native/public/mono) areas of the
6+
dotnet/runtime repo.
7+
8+
In general this guide does not apply to:
9+
10+
1. Shared native code in [src/native](../../src/native)
11+
2. The Mono-specific C# code in [src/mono](../../src/mono) in System.Private.CoreLib or elsewhere
12+
13+
## Code style
14+
15+
Mono is written in C.
16+
17+
We follow the [Mono Coding guidelines](https://www.mono-project.com/community/contributing/coding-guidelines/) for the C code - in particular:
18+
19+
* tabs, not spaces
20+
* space between a function name and the open parenthesis
21+
* braces on the same line as `if`, `for`, `while` etc
22+
23+
## Naming
24+
25+
Mono reserves the following prefixes for symbols: `mono_`, `monovm_`, `m_`, `monoeg_` (used to remap [`eglib`](../../src/mono/mono/eglib)
26+
functions which in source code have a `g_` prefix).
27+
28+
All non-`static` symbols should use one of these prefixes. We generally use `mono_` for most
29+
functions. `m_` is used for some inline accessor functions and macros. `g_` (`mono_eg_`) is used
30+
for `eglib` functions.
31+
32+
Types in a single C file can use any name. Types in a header should use a `Mono` (or sometimes
33+
`mono_`) prefix.
34+
35+
Public API symbols and types *must* be prefixed.
36+
37+
For types Mono generally uses `typedef struct _MonoWhatever { ... } MonoWhatever`. Opaque types may
38+
define `typedef struct _MonoWhatever MonoWhatever` in a client header and define `struct
39+
_MonoWhatever {...}` in an implementation header. Occasionally we break `#include` cycles by adding
40+
forward declarations for some types.
41+
42+
## Macros
43+
44+
Mono derives from an autotools-style project so we generally write `HOST_XYZ` for the machine on
45+
which the runtime is executing (as opposed to the machine on which the runtime was compiled), and
46+
`TARGET_XYZ` for the machine that will be targeted by the JIT and AOT compilers. In the case of AOT
47+
compilation host and target might be different: the host might be Windows, and the target might be
48+
Browser WebAssembly, for example.
49+
50+
Macros generally use a `MONO_` prefix. Macros in public API headers *must* use the prefix.
51+
52+
## Types
53+
54+
Prefer the standard C sized types `int32_t`, `intptr_t`, etc over the eglib types `gint32`, `gsize` etc.
55+
56+
One exception is `gboolean` is prefered over C `bool`.
57+
58+
There are actually three boolean types to keep in mind:
59+
60+
* `gboolean` used internally in the runtime
61+
* `MonoBoolean` used as an interop type with C# bool in internal calls
62+
* `mono_bool` used by the public C API - generally new code shouldn't use it except when adding a new public API function.
63+
64+
## Utility and platform abstraction functions
65+
66+
Mono generally tries to fill in POSIX-like abstractions on platforms that lack them (for example, Windows).
67+
This is in contrast to the CoreCLR PAL which generally tries to add Windows API abstractions on top
68+
of POSIX.
69+
70+
If an existing `eglib` utility function is available, it should be used. New ones can be added.
71+
72+
Other platform specific code is in `src/mono/utils`
73+
74+
## Directory dependencies
75+
76+
For the code in `src/mono/mono`:
77+
78+
* `eglib` should not depend on other code from `src/mono`
79+
80+
* `utils` can depend on `eglib` and 3rd party code from `src/native/external`
81+
82+
* `sgen` depends on `eglib` and `utils`
83+
84+
* `metadata` depends on all of the above (but note that some Boehm-GC files in this directory should not depend on `sgen`)
85+
86+
* `mini` depends on all of the above
87+
88+
* `mini/interp` depends on all of the above
89+
90+
* `components` can use functions from all of the above provided they're marked with
91+
`MONO_COMPONENT_API`, see [../design/mono/components.md](../design/mono/components.md)
92+
93+
The main distinction between `metadata` and `utils` is that utils code should not assume that it is
94+
part of an executing .NET runtime - anything that loads types, creates objects, etc does not belong
95+
in utils but in metadata.
96+
97+
The `mini` directory contains execution engines. If the execution engine has to provide some
98+
functionality to `metadata` it generally does so by installing some callback that is invoked by
99+
`metadata`. For example, `mini` knows how to unwind exceptions and perform stack walks - while
100+
`metadata` decides *when* to unwind or perform a stackwalk. To coordinate, `metadata` exposes an
101+
API to install hooks and `mini` provides the implementations.
102+
103+
## Error handling
104+
105+
New code should prefer to use the `MonoError` functions.
106+
107+
* A non-public-API function should take a `MonoError *` argument.
108+
* In case of an error the function should call one of the `mono_error_set_` functions from [../../src/mono/mono/utils/mono-error-internals.h](../../src/mono/mono/utils/mono-error-internals.h)
109+
* Inside the runtime check whether there was an error by calling `is_ok (error)`, `mono_error_assert_ok (error)`, `goto_if_nok` or `return_if_nok`/`return_val_if_nok`
110+
* If there is an error and you're handling it, call `mono_error_cleanup (error)` to dispose of the resources.
111+
* `MonoError*` is generally one-shot: after it's been cleaned up it needs to be re-inited with `mono_error_init_reuse`, but this is discouraged.
112+
* Instead if you intend to deal with an error, use `ERROR_DECL (local_error)` then call the
113+
possibly-failing function and pass it `local_error`, then call `mono_error_assert_ok
114+
(local_error)` if it "can't fail" or `mono_error_cleanup (local_error)` if you want to ignore the
115+
failure.
116+
117+
## Managed Exceptions
118+
119+
New code should generally not deal with `MonoException*`, use `MonoError*` instead.
120+
121+
New code should avoid using `mono_error_set_pending_exception` - it affects a thread-local flag in
122+
a way that is not obvious to the caller of your function and may trample existing in-flight
123+
exceptions and make your code fail in unexpected ways.
124+
125+
There are two circumstances when you might need to call `mono_error_set_pending_exception`:
126+
127+
1. You're working with a public Mono API that sets a pending exception
128+
2. You're implementing an icall, but can't use `HANDLES()` (see the [internal calls](#internal-calls) section below)
129+
130+
## Internal calls
131+
132+
Prefer P/Invokes or QCalls over internal calls. That is, if your function only takes arguments that
133+
are not managed objects, and does not need to interact with the runtime, it is better to define it
134+
outside the runtime.
135+
136+
Internal calls generally have at least one argument that is a managed object, or may throw a managed exception.
137+
138+
Internal calls are declared in [`icall-def.h`](../../src/mono/mono/metadata/icall-def.h) See the comment in the header for details.
139+
140+
There are two styles of internal calls: `NOHANDLES` and `HANDLES`. (This is a simplification as there
141+
are also JIT internal calls added by the execution engines)
142+
143+
The difference is that `HANDLES` icalls receive references to managed objects wrapped in a handle
144+
that attempts to keep the object alive for the duration of the internal call even if the thread
145+
blocks or is suspended, while `NOHANDLES` functions don't. Additionally `HANDLES` functions get a
146+
`MonoError*` argument from the managed-to-native interop layer that will be converted to a managed
147+
exception when the function returns. `NOHANDLES` functions generally have to call
148+
`mono_error_set_pending_exception` themselves.
149+
150+
## Suspend Safety
151+
152+
See [Cooperative Suspend at mono-project.com](https://www.mono-project.com/docs/advanced/runtime/docs/coop-suspend/) and the [mono thread state machine design document](../design/mono/mono-thread-state-machine.md)
153+
154+
In general runtime functions that may be called from the public Mono API should call
155+
`MONO_ENTER_GC_UNSAFE`/`MONO_EXIT_GC_UNSAFE` if they call any other runtime API. Internal calls are
156+
already in GC Unsafe on entry, but QCalls and P/Invokes aren't.
157+
158+
When calling a blocking native API, wrap the call in `MONO_ENTER_GC_SAFE`/`MONO_EXIT_GC_SAFE`.
159+
160+
Prefer `mono_coop_` synchronization primitives (`MonoCoopMutex`, `MonoCoopCond`,
161+
`MonoCoopSemaphore`, etc) over `mono_os_` primitives (`mono_mutex_t`, `mono_cond_t`). The former
162+
will automatically go into GC Safe mode before doing operations that may block. The latter should
163+
only be used for low-level leaf locks that may need to be shared with non-runtime code. You're
164+
responsible for switching to GC Safe mode when doing blocking operations in that case.
165+
166+
## GC Memory Safety
167+
168+
We have explored many different policies on how to safely access managed memory from the runtime. The existing code is not uniform.
169+
170+
This is the current policy (but check with a team member as this document may need to be updated):
171+
172+
1. It is never ok to access a managed object from GC Safe code.
173+
2. Mono's GC scans the managed heap precisely. Mono does not allow heap objects to contain pointers into the interior of other managed obejcts.
174+
3. Mono's GC scans the native stack conservatively: any value that looks like a pointer into the GC
175+
heap, will cause the target object to be pinned and not collected. Interior pointers from the
176+
stack into the middle of a managed object are allowed and will pin the object.
177+
178+
In general one of the following should be used to ensure that an object is kept alive, particularly
179+
across a call back into managed from native, or across a call that may trigger a GC (effectively
180+
nearly any runtime internal API, due to assembly loading potentially triggering managed callbacks):
181+
182+
* The object is pinned in managed code using `fixed` (common with strings) and the native code gets a `gunichar2*`
183+
* a `ref` local is passed into native code and the native code gets a `MonoObject * volatile *` (the `volatile` is important)
184+
* a `Span<T>` is passed into native code and the native code gets a `MonoSpanOfObjects*`
185+
* an icall is declared with `HANDLES()` and a `MonoObjectHandle` (or a more specific type such as `MonoReflectionTypeHandle` is passed in).
186+
* a GCHandle is passed in
187+
188+
Generally only functions on the boundary between managed and native should use one of the above
189+
mechanisms (that is, it's enough that an object is pinned once). Callees can take a `MonoObject*`
190+
argument and assume that it was pinned by the caller.
191+
192+
In cases where an object is created in native code, it should be kept alive:
193+
194+
1. By assigning into a `MonoObject * volatile *` (ie: use `out` or `ref` arguments in C#)
195+
2. By creating a local handle using `MONO_HANDLE_NEW` or `MONO_HANDLE_PIN` (the function should then use `HANDLE_FUNCTION_ENTER`/`HANDLE_FUNCTION_RETURN` to set up and tear down a handle frame)
196+
3. By creating a GCHandle
197+
4. By assigning to a field of another managed object
198+
199+
In all cases, if there is any intervening internal API call or a call to managed code, the object
200+
must be kept alive before the call using one of the above methods.
201+
202+
### Write barriers
203+
204+
When writing a managed object to a field of another managed object, use one of the
205+
`mono_gc_wbarrier_` functions (for example, `mono_gc_wbarrier_generic_store`). It is ok to call the write
206+
barrier functions if the destination is not in the managed heap (in which case they will just do a normal write)
207+
208+
## Assertions
209+
210+
Mono code should use `g_assert`, `mono_error_assert_ok`, `g_assertf`, `g_assert_not_reached` etc.
211+
212+
Unlike CoreCLR, Mono assertions are always included in the runtime - both in Debug and in Release builds.
213+
214+
New code should try not to rely on the side-effects of assert conditions. (That is, one day we may want
215+
to turn off assertions in Release builds.)
216+
217+
## Mono Public API
218+
219+
Mono maintains a public API for projects that embed the Mono runtime in order to provide the ability
220+
to execute .NET code in the context of another application or framework. The current Mono API is
221+
`mono-2.0`. We strive to maintain binary ABI and API stability for our embedders.
222+
223+
The public API headers are defined in
224+
[`../../src/native/public/mono`](../../src/native/public/mono). Great care should be taken when
225+
modifying any of the functions declared in these headers. In particular breaking the ABI by
226+
removing these functions or changing their arguments is not allowed. Semantic changes should be
227+
avoided, or implemented in a way that is least disruptive to embedders (for example the runtime does
228+
not support multiple appdomains anymore, but `mono_domain_get` continues to work).
229+
230+
### Hidden API functions
231+
232+
In practice certain functions have been tagged with `MONO_API` even if they are not declared in the
233+
public headers. These symbols have been used by projects that embed Mono despite not being in the
234+
public headers. They should be treated with the same care as the real public API.
235+
236+
### Unstable API functions
237+
238+
Functions declared in `mono-private-unstable.h` headers do not need to maintain API/ABI stability.
239+
They are generally new APIs that have been added since .NET 5 but that have not been stabilized yet.
240+
241+
As a matter of courtesy, notify the .NET macios and .NET Android teams if changing the behavior of these functions.
242+
243+
### WASM
244+
245+
The WASM and WASI runtimes in [`src/mono/wasm/runtime`](../../src/mono/wasm/runtime) and
246+
`src/mono/wasi` are effectively external API clients. When possible they should use existing `MONO_API` functions.
247+
248+
As a matter of expedience, the wasm project has sometimes taken advantage of static linking by
249+
adding declarations of internal Mono functions in `src/mono/wasm/runtime/driver.c` and directly
250+
calling Mono internals.
251+
252+
In general new code should not do this. When modifying existing code, mysterious WASM failures may
253+
be attributed to symbol signature mismatches between WASM and the Mono runtime.

0 commit comments

Comments
 (0)