- 20 6月, 2020 3 次提交
-
-
由 Carlos Sanchez Lopez 提交于
* Update MicrosoftPrivateIntellisensePackage and XmlDocFileRoot * Update eng/Versions.props Co-authored-by: NSantiago Fernandez Madero <safern@microsoft.com> * Update eng/Versions.props Co-authored-by: NSantiago Fernandez Madero <safern@microsoft.com>
-
由 Aaron Robinson 提交于
-
由 Carol Eidt 提交于
Initialize method info early so that jitStress can use that information to determine when to enable stress. Fix #8889
-
- 19 6月, 2020 34 次提交
-
-
由 Eric Erhardt 提交于
* Fix Diagnostics.Tracing ILLinkTrim entries - PropertyValue/ReferenceTypeHelper's pattern is recognized by the ILLinker. Protected it with a new trimming test. - RuntimeEventSource doesn't need to be rooted. This class is used by IL and doesn't get trimmed. - EventPipe* and NativeRuntimeEventSource only need to be rooted in CoreCLR, since they aren't used in Mono. So they are moved to just CoreCLR's ILLinkTrim file. Contributes to #35199
-
由 Thays Grazia 提交于
* First version of QCalls implementation, using QCalls on ICU Globalization.
-
由 Stephen Toub 提交于
* Stop ignoring CS1573 in multiple projects * Fix interop.process.cs * Update ZipFileExtensions.ZipArchive.Create.cs Co-authored-by: NDan Moseley <danmose@microsoft.com>
-
由 Adeel Mujahid 提交于
-
由 Anton Firszov 提交于
-
由 Tomáš Rylek 提交于
IJW tests using a managed C++ component (TestLibrary) fail building on Crossgen2 as the MSIL uses a special form of PInvoke without specification of the target library (PInvoke into the unmanaged code within the module itself) and that is currently unsupported in the Crossgen2 metadata reader. Thanks Tomas
-
由 Kenneth Pouncey 提交于
-
由 Yuchong Pan 提交于
* Fixed type cast compilation errors * Temporary changes to trigger the IL interpreter * Added lock holder when invoking JIT from interpreter * Added get_IsSupported as an intrinsic * Fixed widening of narrow integral types * Added COMPlus_InterpreterDebuggable flag to trigger the IL interpreter in the Debug/Checked configurations * Added more accurate description of InterpreterJITThreshold * Fixed assertion failure in AsyncPromoteToTier1 * Renamed env var to ForceInterpreterTier0 * Fixed a bug in AsyncPromoteToTier1 * Fixed the correct semantics of the env var ForceInterpreterTier0 * Added env var ForceInterpreterAlways * Turned off FEATURE_INTERPRETER * Moved CodeVersionManager::LockHolder to minimize scope * Removed redundant check for ForceInterpreterAlways * Renamed COMPlus_ env vars to ForceJitOptimization and ForceInterpreter * Fixed a pre-existing bug for async promotion to tier 1 * Removed redundant annotations on CLR config values * Modified handling of get_IsSupported as a JIT intrinsic * This does not seems a correct fix for the pre-existing promotion bug * Added call to ScheduleBackgroundWork after AsyncPromoteToTier1 * Added more precise check for hardware intrinsics in interpreter * Converted C-style type casts to static_cast<...> * Renamed ForceJitOptimization to EnableJitOptimization * Added COMPlus environment variable to force get_IsSupported to return false for HW intrinsics * Update src/coreclr/src/vm/interpreter.cpp Co-authored-by: NJohn Salem <josalem@microsoft.com> * Removed #ifdef PAL_STDCPP_COMPAT when including 'type_traits' * Removed EnableJitOptimization variable Co-authored-by: NJohn Salem <josalem@microsoft.com>
-
由 Anton Lapounov 提交于
* Improve R2RDump output for ARM64. Resolve targets of ADRP+ADD instruction pairs and BL instructions. Also fix occasional \r\r\n in output for x86/x64. * Handle other PC-relative instructions: ADR, B, CBZ, CBNZ, TBZ, TBNZ
-
由 Kenneth Pouncey 提交于
* [browser][wasm] Add reference for those object that are created with ownsHandle false. * Remove unnecessary special treatment for Global as it is no longer need.
-
由 Maxim Lipnin 提交于
We disable TermInfo library tests because the respective functionality doesn't make sense on WASM.
-
由 Jan Vorlicek 提交于
* Enhance DAC to allow getting AssemblyLoadContext for objects This change adds support to DAC for getting AssemblyLoadContext for a MethodTable. The returned AsseblyLoadContext is the context into which the related type was loaded.
-
由 Jan Kotas 提交于
-
由 Jan Kotas 提交于
Also deleted some dead code
-
由 Tarek Mahmoud Sayed 提交于
* Change Activity Default IdFormat to W3C * Remove the prefix from the switch and fix a test * address some of the feedback * Scope the breaking change to .NET 5 * Fix running the test on NetFX * Changed the define NET5 to W3C_DEFAULT_ID_FORMAT * Add net5 and remove netstandard2.0 configuration in the package * Return back netstandard 2.0 targetting When removing it, caused a CI source build failure
-
由 Eric StJohn 提交于
-
由 Stephen Toub 提交于
* Tweak XML comments on new GCMemoryInfo APIs Various enum members and some new methods were missing doc comments. * Update src/libraries/System.Private.CoreLib/src/System/GCMemoryInfo.cs
-
由 Egor Chesakov 提交于
-
由 Carol Eidt 提交于
* Fix and cleanup dead indirections The backend doesn't support a dead `TYP_STRUCT` indirection. Also, if a target register is needed, a `GT_IND` is better. Fix #36237
-
-
由 Marek Safar 提交于
* Add more WASM Substitutions * Remove internal method
-
由 lindexi 提交于
-
由 Carol Eidt 提交于
If we need to insert a COPY node to signify a temporary copy of field of a multi-reg lclVar, we need to set the appropriate last use bit. Fix #37830
-
由 Tarek Mahmoud Sayed 提交于
* Optimize memory allocations in ActivitySource * revert back using anonymous methods * remove empty line * Address the feedback
-
由 Maxim Lipnin 提交于
-
由 Steve Harter 提交于
Perf improvements mostly for small or value-type POCOs
-
由 Maoni Stephens 提交于
In Server GC when we allocate UOH objects and decide whether we should retry when hardlimit is in effect we have an inconsistency between should_retry_other_heap and virtual_commit. The former uses current_total_committed - current_total_committed_bookkeeping while the latter uses current_total_committed. This causes us to be in a situation where we may have an infinite loop since the allocate_uoh says to retry but virtual_commit fails.
-
由 Anirudh Agnihotry 提交于
* moving coMarshal type down * add unix implementation * setting default value for out params in unix * remove unsafe
-
由 Maoni Stephens 提交于
* new GetGCMemoryInfo API This implements the new GetGCMemoryInfo API that provides much more info than the one we shipped in 3.0. Also added an overload to query GCs by kind - ephemeral, full blocking or BGC, to facilitate not having to keep a really small sampling interval but can still get the kinds of GCs that happen very infrequently. Most of the changes inside GC are the following - All the info is updated at the end of a GC with the exception of BGC which needs special care. We maintain 2 BGC info data structures and one could be in the process of getting updated and if the user requests either BGC info or the last GC info which is also a BGC, we would return the one that's complete. See comments in code for a more detailed explanation. Fill in the after size data for all generations instead of just the affected generations. Timing info recording - + Changed the GC time recording from ms to us. Previously we did the timing analysis with events which also carry their own high res timestamps so in GC using ms is good enough. But now since we are returning the timing info we want to record it in us instead of ms. + Consolidated when we get end_gc_time. It was not including all of the GC work before (eg, decommit eph seg). And moving it to do_post_gc makes it easier for getting the GC info. + Not including the EE restart time for blocking GCs - the reason being restart takes very little time and by the time we get to the actual RestartEE the GC settings could be changed and it's not worth the effort to wait till RestartEE at which point we have to spend more effort to figure out which GC to attribute it to. + Maintain the running % Pause time in GC info.
-
由 dotnet-maestro[bot] 提交于
* Update dependencies from https://github.com/dotnet/arcade build 20200616.1 Microsoft.DotNet.XUnitExtensions , Microsoft.DotNet.VersionTools.Tasks , Microsoft.DotNet.ApiCompat , Microsoft.DotNet.Arcade.Sdk , Microsoft.DotNet.Build.Tasks.Feed , Microsoft.DotNet.Build.Tasks.Packaging , Microsoft.DotNet.Build.Tasks.SharedFramework.Sdk , Microsoft.DotNet.Build.Tasks.TargetFramework.Sdk , Microsoft.DotNet.CodeAnalysis , Microsoft.DotNet.GenAPI , Microsoft.DotNet.GenFacades , Microsoft.DotNet.Helix.Sdk , Microsoft.DotNet.RemoteExecutor , Microsoft.DotNet.XUnitConsoleRunner From Version 5.0.0-beta.20309.1 -> To Version 5.0.0-beta.20316.1 * Update dependencies from https://github.com/mono/linker build 20200617.2 Microsoft.NET.ILLink.Tasks From Version 5.0.0-preview.3.20317.1 -> To Version 5.0.0-preview.3.20317.2 * Update dependencies from https://github.com/dotnet/xharness build 20200618.2 Microsoft.DotNet.XHarness.TestRunners.Xunit From Version 1.0.0-prerelease.20315.3 -> To Version 1.0.0-prerelease.20318.2 Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
-
由 Steve Harter 提交于
-
由 Sean Hall 提交于
-
由 Layomi Akinrinade 提交于
* Add trimming tests for System.Private.Xml changes * Address review feedback
-
由 Adeel Mujahid 提交于
* Delegate cmake version check to cmake when possible There are two kinds of old cmake versions: * <ins>v3.13.0 onwards that support `-B` command-line option:</ins> cmake will print the error message based on `cmake_minimum_required` specified. * <ins>< v3.13.0, which do not have `-B` option:</ins> cmake will fail before the version check with an obscure message, quite implicitly indicating that it doesn't understand `-B`. For that we have an up-front check just for `-B` option, covering the existence of cmake as well. * Update eng/native/build-commons.sh Co-authored-by: NJan Vorlicek <janvorli@microsoft.com> Co-authored-by: NJan Vorlicek <janvorli@microsoft.com>
-
- 18 6月, 2020 3 次提交
-
-
由 lindexi 提交于
-
由 Tomáš Rylek 提交于
The PE issues that were causing issues in the multifolder test on ARM64 have been fixed with https://github.com/dotnet/runtime/pull/37518 so I'm reenabling the temporarily disabled test.
-
由 Stephen Toub 提交于
If a window update is received from the server between the time when we were creating an Http2Stream (initializing its available credit with the current initial window size) and when we stored the stream into the streams dictionary, we might miss out on updating the stream's window size, which could in turn lead to stalls.
-