JIT: Assert failure: "Inconsistent profile data" during 'Lowering nodeinfo' in OSR stress

[AndyAyersMS]

https://dev.azure.com/dnceng-public/public/_build/results?buildId=1118896&view=ms.vss-test-web.build-test-results-tab&runId=30732224&resultId=218120&paneView=dotnet-dnceng.dnceng-build-release-tasks.helix-test-information-tab

linux x64 osr stress

DOTNET_TC_OnStackReplacement=1
DOTNET_OSR_HitLimit=2
DOTNET_JitRandomOnStackReplacement=15
DOTNET_TieredCompilation=1
DOTNET_DbgMiniDumpName=/home/helixbot/dotnetbuild/dumps/coredump.%d.dmp
DOTNET_TC_OnStackReplacement_InitialCounter=1
DOTNET_TC_QuickJitForLoops=1
DOTNET_DbgEnableMiniDump=1
DOTNET_EnableCrashReport=1

  Discovering: System.Text.RegularExpressions.Tests (method display = ClassAndMethod, method display options = None)
  Discovered:  System.Text.RegularExpressions.Tests (found 347 of 378 test cases)
  Starting:    System.Text.RegularExpressions.Tests (parallel test collections = on [2 threads], stop on fail = off)

Assert failure(PID 7406 [0x00001cee], Thread: 7421 [0x1cfd]): Assertion failed '!"Inconsistent profile data"' in 'Microsoft.CodeAnalysis.CSharp.Syntax.InternalSyntax.SyntaxToken:get_Value():System.Object:this' during 'Lowering nodeinfo' (IL size 54; hash 0xca8b4ccf; Tier1)

    File: /__w/1/s/src/coreclr/jit/fgprofile.cpp:4750
    Image: /datadisks/disk1/work/B48C09BB/p/dotnet

[AndyAyersMS]

FYI @dotnet/jit-contrib

[dotnet-policy-service]

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

[MihuBot]

Possible related and/or duplicate issues:

• [jitstress] Assertion failed '!"Inconsistent profile data"' #97892

[amanasifkhalid]

This assert hit in Tier 1, and assuming this requires OSR stress to hit, I suspect we need a specific profile to be collected before OSR compilation starts, which makes reproducing this tricky. This assert is harmless in product builds, so it's not the most pressing issue, and the only subphase of lowering that could introduce profile inconsistency is fgUpdateFlowGraph, so I at least know where to look.

[AndyAyersMS]

Are you suggesting we move this out of 10.0? I agree there's likely not much product impact here.

[amanasifkhalid]

Are you suggesting we move this out of 10.0? I agree there's likely not much product impact here.

I was hoping to get an easy repro locally and fix this in .NET 10, but that doesn't seem to be the case. I think we ought to move it.

[AndyAyersMS]

Sounds good.