Test failure: JIT/Regression/JitBlue/GitHub_65690/GitHub_65690/GitHub_65690.cmd

[v-wenyuxu]

Failed in: runtime-coreclr jitstress-isas-arm 20240615.1

Failed tests:

coreclr windows arm64 Checked jitstress_isas_nohwintrinsic @ Windows.11.Arm64.Open
    - JIT/Regression/JitBlue/GitHub_65690/GitHub_65690/GitHub_65690.cmd

Error message:

 Fatal error. 0xC0000005
   at System.Threading.Thread..cctor()
   at System.GC.RunFinalizers()

Return code:      1
Raw output file:      C:hwB1F10977wBD980A5AuploadsJitBlueGitHub_65690GitHub_65690output.txt
Raw output:
BEGIN EXECUTION
 "C:hwB1F10977pcorerun.exe" -p "System.Reflection.Metadata.MetadataUpdater.IsSupported=false" -p "System.Runtime.Serialization.EnableUnsafeBinaryFormatterSerialization=true"  GitHub_65690.dll 
Expected: 100
Actual: -1073741819
END EXECUTION - FAILED
FAILED
Test failed. Trying to see if dump file was created in C:cores since 6/15/2024 8:38:41 PM
Processing C:corescorerun.exe.1916.dmp
Unable to find cdb.exe at C:Program Files (x86)Windows Kits10Debuggers�rm64cdb.exe
Test Harness Exitcode is : 1
To run the test:
Set up CORE_ROOT and run.
> C:hwB1F10977wBD980A5AeJITRegressionRegression_3../JitBlue/GitHub_65690/GitHub_65690/GitHub_65690.cmd

Stack trace:

   at Xunit.Assert.True(Nullable`1 condition, String userMessage) in /_/src/Microsoft.DotNet.XUnitAssert/src/BooleanAsserts.cs:line 146
   at TestLibrary.OutOfProcessTest.RunOutOfProcessTest(String assemblyPath, String testPathPrefix)
   at Program.<<Main>$>g__TestExecutor346|0_347(StreamWriter tempLogSw, StreamWriter statsCsvSw, <>c__DisplayClass0_0&)

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[jakobbotsch]

This fails under normal jitstress as well, example log: https://helixre107v0xdcypoyl9e7f.blob.core.windows.net/dotnet-runtime-refs-pull-103559-merge-65fba342677a469aac/Regression_2/1/console.ff49ede3.log?helixlogtype=result

[jakobbotsch]

Bisected to range f69972f...d2dbdd0. #103501 seems most likely given the call stack, cc @jkotas

[jakobbotsch]

The repro seems to require

$env:DOTNET_JitDoValueNumber=0
$env:DOTNET_ReadyToRun=0
$env:DOTNET_TieredCompilation=0

so not sure whether or not there is a JIT bug lurking here.

[jakobbotsch]

The particular method that is sensitive to JIT optimization in the test seems to be System.Threading.ProcessorIdCache:ProcessorNumberSpeedCheck(). The failure does not repro with

$env:DOTNET_JitOnlyOptimizeRange="0-0"

but does repro with

$env:DOTNET_JitOnlyOptimizeRange="45e1fa53-45e1fa53"

(the hash of that method).

[jakobbotsch]

Yeah it looks like a JIT bug when VN is disabled, exposed by the new managed code introduced in that PR which we are now compiling with VN disabled in this test.