Test failure: ValueFactory attempted to access the Value property of this instance.

[v-wenyuxu]

Failed in: runtime-coreclr jitstress2-jitstressregs 20240331.1

Failed tests:

coreclr osx arm64 Checked jitstress2_jitstressregs0x80 @ OSX.1200.ARM64.Open
    - JIT/Methodical/eh/basics/throwinclassconstructor_d/throwinclassconstructor_d.dll

Error message:

 ValueFactory attempted to access the Value property of this instance.

Stack trace:

   at System.Threading.ThreadLocal`1.GetValueSlow()
   at Program.<<Main>$>g__TestExecutor273|0_274(StreamWriter tempLogSw, StreamWriter statsCsvSw, <>c__DisplayClass0_0&)

[v-wenyuxu]

Failed in: runtime-coreclr jitstress2-jitstressregs 20240406.1

Failed tests:

coreclr osx arm64 Checked jitstress2_jitstressregs0x80 @ OSX.1200.ARM64.Open
    - JIT/Methodical/eh/basics/throwinclassconstructor_d/throwinclassconstructor_d.dll

Error message:

 ValueFactory attempted to access the Value property of this instance.

Stack trace:

   at System.Threading.ThreadLocal`1.GetValueSlow()
   at Program.<<Main>$>g__TestExecutor273|0_274(StreamWriter tempLogSw, StreamWriter statsCsvSw, <>c__DisplayClass0_0&)

[kunalspathak]

@janvorli - This test passes with DOTNET_LegacyExceptionHandling=1. Assigning to you.

[v-wenyuxu]

Failed in: runtime-coreclr jitstress2-jitstressregs 20240414.1

Failed tests:

coreclr osx arm64 Checked jitstress2_jitstressregs0x80 @ OSX.1200.ARM64.Open
    - JIT/Methodical/eh/basics/throwinclassconstructor_d/throwinclassconstructor_d.dll

Error message:

 ValueFactory attempted to access the Value property of this instance.

Stack trace:

   at System.Threading.ThreadLocal`1.GetValueSlow()
   at Program.<<Main>$>g__TestExecutor273|0_274(StreamWriter tempLogSw, StreamWriter statsCsvSw, <>c__DisplayClass0_0&)

[janvorli]

@kunalspathak I am unable to repro it on my Apple M1. I've set COMPlus_TieredCompilation=0 COMPlus_JitStress=2 COMPlus_JitStressRegs=0x80 and ran the ./artifacts/tests/coreclr/osx.arm64.Checked/JIT/Methodical/Methodical_d1/Methodical_d1.sh
Were you able to repro the issue locally?

[janvorli]

Actually, the merged tests don't return any error code from the test run, so I've missed that the test is failing. But for me, it is failing quite differently:

• It reproes with legacy exception handling too
• It doesn't fail in the test itself, but in the merged tests infrastructure (that actually matches this issue)
• It doesn't repro if I remove the COMPlus_JitStressRegs=0x80
• It doesn't repro if I remove the COMPlus_JitStress=2
• It doesn't repro if I remove the COMPlus_TieredCompilation=0
• In my case, it fails with:
  System.NullReferenceException: Object reference not set to an instance of an object.
  at Program.<$>g__TestExecutor273|0_274(StreamWriter tempLogSw, StreamWriter statsCsvSw, <>c__DisplayClass0_0&) in /Users/janvorli/git/runtime/artifacts/tests/coreclr/obj/osx.arm64.Checked/Managed/JIT/Methodical/Methodical_d1/generated/XUnitWrapperGenerator/XUnitWrapperGenerator.XUnitWrapperGenerator/FullRunner.g.cs:line 732

[janvorli]

Interestingly, without COMPlus_JitStress=2, it doesn't call the
CORINFO_HELP_GETSHARED_GCTHREADSTATIC_BASE_DYNAMICCLASS that is getting the address that we later end up reading "1" from and then attempt to read at an offset from this value, which obviously results in the null reference (in my local run)
It seems that (if I correlated the JIT dumps correctly), without COMPlus_JitStress=2, it ends up calling
System.Threading.ThreadLocal``1[System.__Canon]:get_Value():System.__Canon:this instead.

[kunalspathak]

Interestingly, without COMPlus_JitStress=2, it doesn't call the CORINFO_HELP_GETSHARED_GCTHREADSTATIC_BASE_DYNAMICCLASS that is getting the address that we later end up reading "1" from and then attempt to read at an offset from this value, which obviously results in the null reference (in my local run) It seems that (if I correlated the JIT dumps correctly), without COMPlus_JitStress=2, it ends up calling System.Threading.ThreadLocal1[System.__Canon]:get_Value():System.__Canon:this `` instead.

and what happens when you set DOTNET_LegacyExceptionHandling=1?

[v-wenyuxu]

Failed in: runtime-coreclr jitstress2-jitstressregs 20240421.1

Failed tests:

coreclr osx arm64 Checked jitstress2_jitstressregs0x80 @ OSX.1200.ARM64.Open
    - JIT/Methodical/eh/basics/throwinclassconstructor_d/throwinclassconstructor_d.dll

Error message:

 ValueFactory attempted to access the Value property of this instance.

Stack trace:

   at System.Threading.ThreadLocal`1.GetValueSlow()
   at Program.<<Main>$>g__TestExecutor273|0_274(StreamWriter tempLogSw, StreamWriter statsCsvSw, <>c__DisplayClass0_0&)

[janvorli]

and what happens when you set DOTNET_LegacyExceptionHandling=1

Exactly the same as without it.

[JulieLeeMSFT]

It is still failing in runtime-coreclr jitstress2-jitstressregs.
@janvorli, so what would be the next step?

[JulieLeeMSFT]

link: https://dev.azure.com/dnceng-public/public/_build/results?buildId=660114&view=ms.vss-test-web.build-test-results-tab&runId=16358166&resultId=107955&paneView=debug

[kunalspathak]

I pulled in the latest artifacts from https://dev.azure.com/dnceng-public/public/_build/results?buildId=664183&view=ms.vss-test-web.build-test-results-tab&runId=16472434&resultId=105819&paneView=debug and was able to repro it on my Mac M2:

[janvorli]

I can repro it with runfo using the build from your recent link.