- 17 7月, 2022 6 次提交
-
-
由 Adeel Mujahid 提交于
* Fix typos * Cleanup trailing whitespaces in committed files * Revert a macro for win32 compat * Disambiguate test data method * Revert XMLPath test which rely on external assets * Revert whitespace change in Xml tests * Revert ClrEtwAl and ILLink.Shared * Revert crossgen2 props/targets and *.wxl
-
由 Mike McLaughlin 提交于
Properly calculate the ELF unwind info section size Uses the same method that windbg does to calculate the eh_frame section size by partial decoding the FDE/CIE's enough to figure out the total size. We can't use the .eh_frame section because ELF module sections are not in-memory. Change the ReadMemoryAdapter passed to the new PAL_GetUnwindInfoSize function to read memory directly and not add it to the memory region list
-
由 Adeel Mujahid 提交于
-
由 chrisnas 提交于
* Implement ICorProfilerInfo13 to support object handle creation/destruction/access * Add test for ICorProfilerInfo13 methods about handle management * Fix compilation issues * Remove TAB * Fix build issues * Complete handles test * Fix Linux compilation error * Fix test build issue * Improve handles test * Fix build error * Add logs in handles test * Add type information to handles test logs * Take review comments into account * Add logs to handles test * Fix test error
-
由 Michal Strehovský 提交于
Fixes dotnet/corert#3251. `Stream.HasOverriddenBeginEndRead`/`Stream.HasOverriddenBeginEndWrite` are magic methods that call into the runtime on both CoreCLR and Mono to find out whether `Stream.BeginRead`/`EndRead`/`BeginWrite`/`EndWrite` are overriden on the current class. Since we don't have a runtime in NativeAOT, implement this in the compiler. The answer to this question is expressible in IL: load function pointer to the method virtually and non-virtually and compare. I'm not calling into `FunctionPointerOps` because I can't imagine a scenario where this wouldn't work, but I can be persuaded to call to `FunctionPointerOps` to do the comparison. Should make us pass all System.IO libraries tests.
-
由 David Mason 提交于
-
- 16 7月, 2022 5 次提交
-
-
由 Jan Kotas 提交于
Fixes #72231
-
由 Aaron Robinson 提交于
There were scenarios where GCStress was trashing last error. This change attempts to avoid it ever trashing last error.
-
由 Qiao Pengcheng 提交于
* [LoongArch64] add crossgen2 for LoongArch64. * add `aot/ILCompiler.Compiler/Compiler/DependencyAnalysis/Target_LoongArch64` * amend the code format. * delete unused comments.
-
由 Bruce Forstall 提交于
Due to very odd, stress-induced flow graph, on x86, the fall-through predecessor block to an empty (but not unreachable) block was a "catch" handler that was not reachable from the block's target (since "catch" blocks don't have predecessors). However, the code in fgRemoveBlock that was calling optScaleLoopBlocks didn't make any sense, and removing it causes no diffs. So get rid of it. Fixes #64764
-
由 Jakob Botsch Nielsen 提交于
FixContextForEnC was treating an index that may be negative as an unsigned value. This would crash the runtime during EnC when running as 64-bit. Fix #70834
-
- 15 7月, 2022 7 次提交
-
-
由 Michal Strehovský 提交于
-
由 Michal Strehovský 提交于
This should root all non-trimmable assemblies, same as TrimmerDefaultAction=copy. We still support consuming ILCompiler through NuGet when targeting .NET 6, so recognize the old thing too. We might want to cut the .NET 6 support close to RC and at that time TrimmerDefaultAction handling can go away. Related: https://github.com/dotnet/linker/pull/2856.
-
由 Adeel Mujahid 提交于
-
由 Andrew Au 提交于
-
由 Andrew Au 提交于
Co-authored-by: NMaoni0 <maoni@microsoft.com>
-
由 Jan Kotas 提交于
Fixes #71884
-
由 SingleAccretion 提交于
* Generalize init block morphing a little Make it handle arbitrary local destinations. * Add a test
-
- 14 7月, 2022 9 次提交
-
-
由 Egor Bogatov 提交于
-
由 Egor Bogatov 提交于
-
由 Jakob Botsch Nielsen 提交于
We were constructing the TypeHandle before validating that the handle did not represent an unknown handle. This crashes in the constructor in debug builds due to some debug-only validation running there.
-
由 Michal Strehovský 提交于
Product fixes: * Don't try to be helpful in the exception when `MakeGeneric*` constraint validation fails. CoreCLR is not helpful and S.Linq.Expressions is testing for it. * We need to inject delegate `Invoke` method dependencies elsewhere. A test looking to reflection-invoke the `Invoke` method was failing with `MissingMetadataException` but we have no right to be failing on that per policy. Test fixes: * Consider NativeAOT always trimmed. We could potentially do what Mono is doing, but we currently don't support substitution files (only embedded substitutions), and NativeAOT will always be trimmed. * Mark a test that is fundamentally incompatible (and pointless) with trimming as such. * Add a bunch of RD.XML. Fixes #69944. Fixes #69928. Fixes #69943.
-
由 Isabel Serrato 提交于
* Implement capability for users to upload .dgml files in addition to ETW events. Allows program to be run without admin permissions if the user has a .dgml file representing the dependencies in their program. * Update UI to allow users to upload .dgml files from the file browser or the command line * Update GUI to allow resizing of pages, clean up layout of pages * Add test cases to ensure that graph is correctly replicated from sample .dgml files * Restructure graph class to allow multiple edges and update GUI to show all reasons that two nodes are connected on one line Co-authored-by: NAaron Robinson <arobins@microsoft.com>
-
由 Jeremy Koritzinsky 提交于
If the base type is a sequential zero-sized type, treat a derived auto-layout type as starting from field offset 0. (#72065)
-
由 Jakob Botsch Nielsen 提交于
Adding object dereferences was calling EnsureBlockConditions which would reallocate all the lists, removing previous array dereferences. This resulted in missing array null checks in some cases. Fix #71611
-
由 Jan Kotas 提交于
Fixes #72090 Co-authored-by: NStephen Toub <stoub@microsoft.com>
-
由 Jakob Botsch Nielsen 提交于
This was relying on the profile data containing the unboxing stub before. That really only works out for dynamic PGO. Static profiles or textual PGO data will end up with the exact method and hit a JIT assert passing a TYP_REF this arg to a TYP_BYREF this param.
-
- 13 7月, 2022 10 次提交
-
-
由 Stephen Toub 提交于
* More reliably clean up more SafeHandle instances This calls Dispose on SafeHandles (or things wrapping SafeHandles) that were otherwise being left for finalization: 1. Some of these are fixing finalization happening even on success paths (typically where the implementation isn't disposing of something that directly or indirectly wraps a SafeHandle). 2. Some of these are fixing finalization happening for invalid SafeHandles. 3. Some of these are fixing tests to finalize less. My goal with fixing the tests was to eliminate the noise in order to find instances of the other two cases. These were found primarily via two means: - Debug-instrumentation in SafeHandle to log when one is finalized. This instrumentation is built into debug/checked builds of SafeHandle and requires setting the DEBUG_SAFEHANDLE_FINALIZATION environment variable to "1". - Auditing use of SafeHandle.IsInvalid There's a lot more that can be cleaned up using the SafeHandle instrumentation, but I'm pausing here for now. The System.IO.Pipes, System.IO.FileSystem, and System.Security.Cryptography tests are clean on Windows.
-
由 Will Smith 提交于
-
由 Aaron Robinson 提交于
* Update details about ELEMENT_TYPE_TYPEDBYREF Details in siginfo.cpp were stale. Add some logging that helped track down the issue. * Fix byref reporting in crossgen2 Co-authored-by: NAaron Robinson <arobins@microsoft.com> Co-authored-by: NJan Kotas <jkotas@microsoft.com>
-
由 Michal Strehovský 提交于
The newstyle calling convention handling being exclusive to JitInterface is causing problems in other spots of the AOT compiler where we would need to replicate the rules of when to read custom modifiers, which ones matter, what's their relationship to `MethodSignatureFlags`, and what combinations of the modifiers makes sense. I'm adding a single `UnmanagedCallingConventions` enum that captures all the calling conventions we support and add means to get to this from p/invoke, `UnmanagedCallersOnly` method and from a calli signature.
-
由 Egor Bogatov 提交于
-
由 Tomáš Rylek 提交于
* Regression test for GitHub issue #71319 * Fix runtime assertion failure due to wrong GC mode * Disable the new regression test on Mono with issue #71910
-
由 Manish Godse 提交于
-
由 SingleAccretion 提交于
Reuse lists (now hashtables) of outstanding reads in "CheckLclVarSemanticsHelper".
-
由 Drew Scoggins 提交于
* Add logic for selecting the right version of pgort140 * Address PR feedback * Add logic for x86 pgort * Remove dead code
-
由 Jakob Botsch Nielsen 提交于
Fix #71987
-
- 12 7月, 2022 3 次提交
-
-
由 Mike McLaughlin 提交于
Various createdump fixes. Smaller MacOS dump size, better logging/stats, misc cleanup, etc. (#71569) * Fix where CombineMemoryRegions is called and perf fix for PAGE_SIZE on MacOS M1 * Remove MEMORY_REGION_FLAG_MEMORY_BACKED flags because it was always set now * Add better memory tracing and memory region stats Fix MacOS native module regions when overlapping with existing. Fix MacOS adding the managed modules to the module mapping list before the "other mappings" is built. * Don't add share_mode == SM_EMPTY regions * Add crashreport success status message for VS4Mac * Launch createdump from SIGTERM handler directly to reduce the time it takes to get the crash report/dump for VS4Mac * Fix issue https://github.com/dotnet/runtime/issues/71561 * Add target process terminated/alive message * Add total createdump elapsed time message
-
由 Bruce Forstall 提交于
If we don't do SSA, we don't do liveness, but Lowering depends on liveness-created lvDoNotEnregister, so if we're not doing SSA, at least run local var liveness. Note that in the MinOpts case, `lvSetMinOptsDoNotEnreg()` sets all locals to do-not-enregister. Fixes #62251
-
由 SingleAccretion 提交于
We cannot assume the underlying local has a layout.
-