- 19 2月, 2022 21 次提交
-
-
由 Tomáš Rylek 提交于
When building tests in merged mode, we need to copy the native references along with the component test dlls to the output folder to make them discoverable by the OS loader. Thanks Tomas
-
由 Tomáš Rylek 提交于
Turns out this particular test doesn't actually need process isolation, I overzealously marked it as such in a previous wave just because it's a static initialization test. This lets us postpone dealing with out-of-process ilproj tests to a later phase. Thanks Tomas
-
由 Jan Vorlicek 提交于
While running managed debugger tests with x86 runtime debug build, I was hiting an assert in MachineState::pRetAddress() method in a couple of tests. It turns out there was a bug in the HelperMethodFrame::UpdateRegDisplay introduced about 5 years ago that only happens in DAC. The m_MachState can be uninitialized when we enter the method and we were calling m_MachState.pRetAddr() before performing the initialization. This change fixes it.
-
由 Jeremy Koritzinsky 提交于
* Update conversion code fix to offer multiple fixes for appending each previously probed entry-point suffix when ExactSpelling is false. * Remove ExactSpelling property and move properties that were relying on ExactSpelling = false to use explicit entrypoint names. * Emit ExactSpelling = true for all of our emitted target DllImports. * Fix a few entry points * Add a comment around CharSet.Auto * Update compatibility doc. * Update docs/design/libraries/DllImportGenerator/Compatibility.md Co-authored-by: NElinor Fung <elfung@microsoft.com> Co-authored-by: NElinor Fung <elfung@microsoft.com>
-
由 Ankit Jain 提交于
* [wasm] Trigger all wasm jobs, if eng/Version.Details.xml has any changes, so we run a full check on emsdk updates * fix yml * Fix merge * [wasm] Pass isRollingBuild to runtime-extra-platforms-wasm template, to correctly trigger the EAT/AOT builds * .. and pass isWasmOnlyBuild also
-
由 Tarek Mahmoud Sayed 提交于
-
由 Larry Ewing 提交于
* Make the sanitize pass less aggressive
-
由 Elinor Fung 提交于
-
由 Marek Fišera 提交于
- Read first directory name under node instead of hardcoded version.
-
由 Bruce Forstall 提交于
1. Be more robust to temp directory removal failure If we fail to remove a temporary directory, don't crash. Log the failure and the set of directories and files still remaining, and continue. We have seen this failure in at least one Linux x64 PMI coreclr_tests SuperPMI collection: ``` [Errno 39] Directory not empty: '/datadisks/disk1/work/B18E0979/t/tmpov3b4_qa' ``` 2. Limit the length of file names created by `make_safe_filename`. We were creating file names out of full path names, where those paths contained long temporary directory paths, and thus we were exceeding the maximum allowed file name component.
-
由 Viktor Hofer 提交于
* Clean-up slngen related logic * Update dotnet-tools.json Co-authored-by: NAlexander Köplinger <alex.koeplinger@outlook.com>
-
由 Stephen Toub 提交于
* Skip RegexOptions.NonBacktracking phase 3 for some patterns If a pattern doesn't have any captures and if any match of that pattern will always be the same length, we can skip the Phase 3 computation as, given the computed starting position of the match, we know exactly where it's going to end. * Add example to the phase comments
-
由 Will Smith 提交于
* Initial work * Added a comma to display * Cleanup * Fixing build * More cleanup * Update comment * Update comment * Added CompareEqual Vector64/128 with Zero tests * Do not contain op1 for now * Wrong intrinsic id used * Removing generated tests * Removing generated tests * Added CompareEqual tests * Supporting containment for first operand * Fix test build * Passing correct register * Check IsVectorZero before not allocing a register * Update comment * Fixing test * Minor format change * Fixed formatting * Renamed test * Adding AdvSimd_Arm64 tests: * Adding support for rest of 'cmeq' and 'fcmeq' instructions * Removing github csproj * Minor test fix * Fixed tests * Fix print * Minor format change * Fixing test * Initial commit for Vector.Create to Vector.Zero normalization * Added some emitter tests * Feedback * Update emitarm64.cpp * Feedback * Handling variations of Vector.Create * Use Operands iterator instead of edges * Fix condition * Simplify * format * Fixed IsFloatPositiveZero * Uncomment * Updated tests to include Vector64.Create/Vector128.Create for ARM64 * Making implementation of IsFloatPositiveZero explicit * Update src/coreclr/jit/gentree.cpp Co-authored-by: NSingleAccretion <62474226+SingleAccretion@users.noreply.github.com> * Feedback * Update comment * Update comment * Do not perform optimization when VN CSE phase * use ResetHWIntrinsicId * Assert !optValnumCSE_phase * Simplify IsVectorZero * Simplify IsVectorZero * Simplify some uses of Vector*_get_Zero * Added another test * Fixed formatting * Revert lowering removal * Initial work for optimizations on VectorZero value numbering * Allowing all Vector.Zero to be constant prop'ed. Added VNFuncSimdTypeInfo. * Update gentree.h * Quick rename * Removed extra variable * Added default case * Format * Fixed vnDumpSimdType to take into account CorInfoType * Fixed gtNewSimdZeroNode to produce the right Vector*_get_Zero based on simdSize * Formatting * Feedback and a loop test * Added another test. Formatting fixes * Added GetSimdBaseJitPreciseType * Feedback * Minor fix * Minor comment update * Added another comment * Added another comment * Added another comment * Update comment * Formatting * Feedback * Fixing build * Feedback * Update assertionprop.cpp * Formatting Co-authored-by: NSingleAccretion <62474226+SingleAccretion@users.noreply.github.com>
-
由 David Wrighton 提交于
* Fix Vector256.IsHardwareAccelerated in R2R binaries - If a function had a dependency on Vector256<T> but not on any Avx2 functionality then we would fail to produce an R2R image using a debug version of the compiler, and would generate an incorrect function where the result of calling Vector256.IsHardwareAccelerated would produce the wrong result. - To fix this we now annotate the VectorXXX instruction sets so that if there is a negative depenedency on them during R2R generation, that the function is not available for use if the underlying actual instruction set is available to use at runtime. - Adjust test Runtime_34587 to test instruction set manner in a way that allows testing crossgen2 instruction set variation - Move instruction set issupported checks into standalone functions to isolate the tests in the JIT/Crossgen2 compiler to better see exactly what is happening - Add diagnostic output to test, for better diagnosis of future failures.
-
由 Jeremy Koritzinsky 提交于
-
由 Jeremy Koritzinsky 提交于
-
由 SingleAccretion 提交于
* Add checking for zero-offset FldSeq addition * Add a test * Fix ADDR(LCL_VAR) Zero [FldSeq] * Always add NotAField field sequences * NotAField printing improvements
-
由 SingleAccretion 提交于
We only require the "greater than" condition on x86.
-
由 Nikola Milosavljevic 提交于
-
由 Marek Fišera 提交于
-
由 Ankit Jain 提交于
* Rationalize wasm jobs run on CI Essentially: - For rolling builds, we want to run all the wasm jobs, unconditionally - And `runtime-extra-platforms` runs the tests that were not in `runtime`, and `runtime-staging` - For PRs, we use the various path checks Changes: - split wasm jobs into various templates - combine, and simplify some jobs * `runtime-extra-platforms`, and `runtime-wasm` run only when manually triggered with a comment - `/azp run <pipeline-name>` * `runtime`, and `runtime-staging`, run jobs only when relevant paths change. And for `EAT`, and `AOT`, only smoke tests are run. * And when `runtime-wasm` is triggered manually, it runs *all* the wasm jobs completely | . | runtime | runtime-staging | runtime-extra-platforms(manual only) | runtime-wasm (manual only) | | ----------------- | -------------------- | --------------- | ------------------------------------ | ------- | | libtests | linux: all, only-pc | windows: all, only-pc | linux+windows: all, only-pc | linux+windows: all, always | | libtests eat | linux: smoke, only-pc | - | linux: all, only-pc | linux: all, always | | libtests aot | linux: smoke, only-pc | windows: smoke, only-pc | linux+windows: all, only-pc | linux+windows: all, always | | | | | | | | Wasm.Build.Tests | linux: only-pc | windows: only-pc | linux+windows: only-pc | linux+windows | | Debugger tests | - | linux+windows: only-pc | linux+windows: only-pc | linux+windows | | Runtime tests | linux: only-pc | - | linux: only-pc | linux | * `runtime`, and `runtime-staging`, run all the wasm jobs unconditionally, but `EAT`, and `AOT` still run only smoke tests. * `runtime-extra-platforms` also runs by default. And it runs only the cases not covered by the above two pipelines. * jobs w/o `only-pc` are always run | . | runtime | runtime-staging | runtime-extra-platforms (always run) | runtime-wasm (manual only) | | ----------------- | ------------- | --------------- | ------------------------------------ | ------ | | libtests | linux: all(v8/chr/node) | windows: all | linux+windows: all with node | N/A | | libtests eat | linux: smoke | - | linux: all | | | libtests aot | linux: smoke | windows: smoke | linux+windows: all | | | | | | | | | Wasm.Build.Tests | linux: always | windows: always | none | | | Debugger tests | - | linux+windows: always | none | | | Runtime tests | linux: always | - | none | | * [wasm] Run tests with node only on runtime-wasm
-
- 18 2月, 2022 19 次提交
-
-
由 Alexander Köplinger 提交于
Allocate the buffer outside of the loop so we don't hit OOM issues. While looking at the test I noticed that it actually had a bug too: Nothing was asserting that we indeed throw an InvalidOperationException when the writer is completed. Adding `Assert.ThrowsAsync` revealed that the current loop iteration count was too low to hit the exception reliably within the 10ms delay so instead check for the execution time. Fixes https://github.com/dotnet/runtime/issues/64930
-
由 Eirik Tsarpalis 提交于
* Support resumable serialization in NullableConverter<T> * use null instead of default
-
由 Krzysztof Wicher 提交于
* Fix XNodeReaderFunctionalTests * cleanup Flush/Dispose * cleanup few more flushes
-
由 Qiao Pengcheng 提交于
* [LoongArch64] add ToolBox directory about jitinterace for getting ABI-info. (#59561) Co-authored-by: Loongson's .NET-teams * [LoongArch64] add new interace for getting ABI-info. (#59561) * [LoongArch64] add the linking page for LoongArch64 ABI-info. (#59561) * [LoongArch64] moved ThunkInput.txt to #62885. * [LoongArch64] moved vm/jitinterface.cpp to #62885. * remove the JIT/EE interface back from #62885.. * [LoongArch64] Fix the compiling error after merge. * [LoongArch64] add comments for the returned value of `getFieldTypeByHnd`. * [LoongArch64] rename getFieldTypeByHnd to getFieldSizeClassificationByHnd. Also add macro define for returned value of `getFieldSizeClassificationByHnd`. * [LoongArch64] Delete the interface `getArgType2`. And refactor the returned values of `getFieldSizeClassificationByHnd`. * [LoongArch64] delete `GetArgType` within `ToolBox/superpmi`. * [LoongArch64] rename `getFieldSizeClassificationByHnd` to `getLoongArch64PassStructInRegisterFlags`. * [LoongArch64] amend the floating-ABI for native-struct. * [LoongArch64] update all related `GetFieldSizeClassificationByHnd` by `GetLoongArch64PassStructInRegisterFlags` and amend some comments. * [LoongArch64] replace `LookupApproxFieldTypeHandle()` by `GetFieldTypeHandleThrowing()`. * [LoongArch64] implements the crossgen2 for LoongArch64. * Revert "[LoongArch64] implements the crossgen2 for LoongArch64." This reverts commit b05a2b90e8d8a2f6d1cf7f101ddfc9d4ed8d984e. The crossgen2 for LoongArch64 will be submitted by a new PR. * [LoongArch64] update the `GUID JITEEVersionIdentifier`. Also delete some unused comments. Co-authored-by: Nqiaopengcheng <qiaopengcheng-hf@loongson.cn>
-
由 Marek Fišera 提交于
-
由 Tom Deseyn 提交于
* FileSystemWatcher.Linux: handle races while adding child directories. This handles the child directory getting removed or replaced by a file while we are adding a watch for it or enumerating its subdirectories. * Pass other enumeration exceptions to the user.
-
由 Vladimir Sadov 提交于
* Use CopyContext to restore saved context on X86 * PR feedback * more PR feedback
-
由 Steve Pfister 提交于
This change improves the inner devloop experience by making sure to cache the .aotdata file when it's generated. Otherwise, if you clean your artifacts directory and no aot changes have been made, it will not be generated again. The result being an error when building an app. I believe this mostly impacts iOS configurations.
-
由 Zoltan Varga 提交于
Enable it using a define instead of generating code into driver-gen.c. Co-authored-by: NLarry Ewing <lewing@microsoft.com>
-
由 Radek Doulik 提交于
* [wasm] Rollback emscripten to 2.0.23 * Enable disabled tests
-
由 Immo Landwerth 提交于
-
由 Elinor Fung 提交于
-
-
由 Egor Bogatov 提交于
-
由 Egor Bogatov 提交于
Co-authored-by: NAndy Ayers <andya@microsoft.com>
-
由 Bruce Forstall 提交于
Move about 3300 lines of architecture-specific code to the architecture-specific codegen files (codegenarm.cpp, codegenarm64.cpp, codegenarmarch.cpp, codegenxarch.cpp). Much of this is prolog/epilog generation. A new set of arch-specific functions, genZeroInitFrameUsingBlockInit, is added to extract a huge amount of arch-specific code from genZeroInitFrame. The moved code wasn't altered except to remove unnecessary ifdefs. Minor: - inlined and removed getAdjustSP - removed unused Compiler versions of GetHasTailCalls/SetHasTailCalls
-
由 Aaron Robinson 提交于
* Update CrossGen2 to support ByRef fields in all scenarios * Update tests to help uncover CrossGen2 scenario validation * Add support in Mono for the ByReference`1 being considered a ref field during Explicit offset validation.
-
由 Immo Landwerth 提交于
-
由 Egor Bogatov 提交于
-