- 20 8月, 2022 16 次提交
-
-
由 Alexander Köplinger 提交于
It was only disabled on interpreter but it fails on other configs too. Closes https://github.com/dotnet/runtime/issues/74049
-
由 Tomáš Rylek 提交于
The field offset mismatch in the test JIT/Regression/JitBlue/Runtime_60035/Runtime_60035.csproj specific to Crossgen2 composite mode is due to the recent proliferation of 16-byte alignment specific to Vector across the runtime repo. Previously, the largest supported alignment was 8 and X86 was the only architecture not respecting it. In general the problem is caused by the fact that the native CoreCLR runtime method table builder calculates field offsets without taking the method table pointer into account; in the particular case of the Runtime_60035 test, the class System.Text.Encodings.Web.OptimizedInboxTextEncoder has a field named _allowedAsciiCodePoints that is 16-aligned that exposes this inconsistency. Thanks Tomas
-
由 Robert Henry 提交于
Co-authored-by: NJan Kotas <jkotas@microsoft.com>
-
由 Marek Fišera 提交于
-
由 Anton Lapounov 提交于
-
由 Vladimir Sadov 提交于
* enable test * flat layout is ok if it is not a file * do not convert non-file layouts on OSX * run some code when loading from byte array * use flat layout on all platforms * Suggestion from PR feedback Co-authored-by: NJan Vorlicek <jan.vorlicek@volny.cz> * typo Co-authored-by: NJan Vorlicek <jan.vorlicek@volny.cz>
-
由 Tomáš Rylek 提交于
In a recent change DavidWr modified the Crossgen2 manifest metadata emitter by adjusting the assembly index offset - before his change, manifest assembly indices started at 1, now they start at 2 as 1 newly represents the manifest metadata itself. This change fixes the counterpart assembly index calculations in R2RDump. Thanks Tomas
-
由 Andy Ayers 提交于
* Rework the post importation phase * Rework the morph init phase * Rework fgAddInternal * Rework flow graph update phases, compute preds, and update finally flags * Rework struct promotion phase * Rework local morph * Rework morph implicit byrefs * Rework GS cookie phase * Rework compute block and edge weights * Rework create funclets * no funclets on x86, so no phase * account for some jit stress changes * review feedback
-
由 David Wrighton 提交于
Unfortunately, attempting to actually fix all the ABI issues is probably too complex for this time in the release cycle. - Mark `Int128` as being ABI unstable in Crossgen2. This will prevent it from appearing in function signatures - Adjust layout so that crossgen2 and the runtime agree about layout - Arm32 - Alignment of 8 bytes - Everywhere else - Alignment of 16 bytes - Disable use of `Int128` in pinvokes as a by value parameter or return value. (Match behavior of `Vector128<T>`) (Unlike Vector128<T> handle scenarios such as having fields of `Int128` types.) - Disable tests that will fail now that pinvokes are disabled - Build a test that will succeed now that pinvokes are disabled - Add test that alignment matches OS behavior - Add unit tests for alignment behavior - Update R2R version Fixes #72206
-
由 Jeremy Koritzinsky 提交于
**BYPASS_SECRET_SCANNING**
-
由 Katelyn Gadd 提交于
* Omit the JS sync context if threads are not enabled * Define FEATURE_WASM_THREADS in the InteropServices project if threads are enabled Co-authored-by: NAleksey Kliger <aleksey@lambdageek.org>
-
由 Tom Deseyn 提交于
* Directory.CreateDirectory: create missing parents using default UnixFileMode. * Simplify TarHelpers.CreateDirectory. * PR feedback. * AssertFileModeEquals: don't skip on Android. * Always overwite directory metadata. * TarHelpers.Windows: remove CreateDirectory overwriteMetadata. * Fix Assert indentation.
-
由 Kevin Jones 提交于
-
由 Simon Rozsival 提交于
This PR only disables tests that are failing on Android. iOS, Mac Catalyst, and tvOS reported in #74076 Issues to track disabled tests: #74244, #74245, #74247
-
由 Isabel Serrato 提交于
* Create README for DependencyGraphViewer, edit help messaging within app Co-authored-by: NSven Boemer <sbomer@gmail.com>
-
由 Andy Ayers 提交于
Simplistic version where we just look for identical operands.
-
- 19 8月, 2022 24 次提交
-
-
由 Adeel Mujahid 提交于
-
由 Jeremy Barton 提交于
-
由 Eric Erhardt 提交于
* Revert 6a02d5d3 * Revert relevant parts of 4222e699 and AES pieces of 8f75cc93 * Remove Subtle Crypto interop code Fix #73858
-
由 Meri Khamoyan 提交于
-
由 Alhad Deshpande 提交于
* Avoid transformation from multiplication to left shift in case of 64 bit value * Fixed System.Collections.Concurrent.Tests timeout issue * [ppc64le] Fixed test case failures * [ppc64le] Incorporated code review comments Co-authored-by: NSapana Khemkar <sapana.khemkar@ibm.com> Co-authored-by: NSapana-Khemkar <94051076+Sapana-Khemkar@users.noreply.github.com>
-
由 Marek Fišera 提交于
-
由 Yusuke Yamada 提交于
-
由 Michal Strehovský 提交于
Lets us make #72645 a non-blocking-release issue. We also set NativeAotSupported to false for Mac on the line above. Crossgen2 will still ship NativeAOT compiled on x64 Linux and Windows. R2R+SingleFile+Trimmed elsewhere.
-
由 Michal Strehovský 提交于
Fixes #73926. Also adding AOT analysis suppression for parity. I didn't port the warning level support because the compat kind of goes beyond that. We can revisit later if needed.
-
由 Bruce Forstall 提交于
Disable noway_assert testing, which falls back to MinOpts. This test specifically disables MinOpts and fails if code is not optimized. Fixes #57910
-
由 Michal Strehovský 提交于
#74045 has these legs failing because the Helix queue wait time was two hours and forty minutes...
-
由 Michal Strehovský 提交于
Port of https://github.com/dotnet/linker/pull/2979 Fixes #73048.
-
由 Kunal Pathak 提交于
* Revert "Ensure that GT_CNS_VEC is handled in LinearScan::isMatchingConstant (#70171)" This reverts commit 24f5de4b. * Revert "Revert "Ensure that GT_CNS_VEC is handled in LinearScan::isMatchingConstant (#70171)"" This reverts commit 984120f0cde0d26f9768112296e68164cfe76c8e. * Add partial save check * Use the correct register type
-
由 Anton Lapounov 提交于
-
由 SRV 提交于
* Fixed #70330 * Removed pessimistic buffer cleanup
-
由 Kunal Pathak 提交于
* Enable linux arm/arm64 superpmi collect * Log build output * Remove comment
-
由 Kevin Jones 提交于
-
由 Larry Ewing 提交于
-
由 Jo Shields 提交于
* Switch to runtime.yml model - ALL release branch commits get built
-
由 Bruce Forstall 提交于
* Add support for JitStress with random stress number There are two new pipelines: 1. runtime-coreclr jitstress-random (https://dev.azure.com/dnceng/public/_build/results?buildId=1950662&view=results) 2. runtime-coreclr libraries-jitstress-random (https://dev.azure.com/dnceng/public/_build/results?buildId=1950664&view=results) The first runs coreclr tests, the second runs libraries tests. Each individual job gets its own JitStress random number. The random number is set in src/tests/Common/testenvironment.proj such that its value is visible in the log file. Currently, the random number is the milliseconds value of the current time, so it is between 0 and 999. The pipelines are each configured to run once per week, on weekends. * Remove unused jitstress_random_*
-
由 Parker Bibus 提交于
* Try getting the maui-version from the json and using the rc1 rollback versions. * Update the rollback and maui-version links to their longterm versions. * Add -L option to curl to follow the aka.ms link for the maui-version.json. * Renamed maui-version.json to maui-supported-sdk-version.json to clarify it's use.
-
由 Dan Moseley 提交于
-
由 Kunal Pathak 提交于
* Add bunch of shortcircuit checks * shortcircuit varTypeIsStruct() * optCanPropBndsChk * Add some shortcircuit code * jit format * Fix a bug * Added helper methods to have consistent checks
-