DataContractSerializerTests.DCS_MyPersonSurrogate_Stress fails in CI

[runfoapp]

Test failures in last 30 days as of 8/26:

Day	Run	Details
8/12	PR 1939151	Alpine.314.Arm32.Open
8/8	PR 1929242	Alpine.314.Arm32.Open
8/4	Rolling run 1922602	Alpine.314.Arm32.Open
7/29	PR 1912310	Alpine.314.Arm32.Open

https://helixre107v0xdeko0k025g8.blob.core.windows.net/dotnet-runtime-refs-heads-main-eb585c6d02b64c4dad/System.Runtime.Serialization.Xml.Tests/1/console.048975d4.log?%3Fhelixlogtype%3Dresult

  Starting:    System.Runtime.Serialization.Xml.Tests (parallel test collections = on, max threads = 4)
    DataContractSerializerTests.DCS_MyPersonSurrogate_Stress [FAIL]
      Assert.Equal() Failure
      Expected: Person[Name=name,Age=30]
      Actual:   Person[Name=name,Age=30]
      Stack Trace:
        /_/src/libraries/System.Runtime.Serialization.Xml/tests/DataContractSerializerStressTests.cs(62,0): at DataContractSerializerTests.DCS_MyPersonSurrogate_Stress()
           at System.RuntimeMethodHandle.InvokeMethod(Object target, Void** arguments, Signature sig, Boolean isConstructor)
        /_/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodInvoker.cs(69,0): at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr)
  Finished:    System.Runtime.Serialization.Xml.Tests

Runfo Tracking Issue: datacontractserializertests.dcs_mypersonsurrogate_stress fails in CI

Build	Definition	Kind	Run Name	Console	Core Dump	Test Results	Run Client
1939151	runtime	PR 73734	net7.0-Linux-Release-arm-CoreCLR_checked-(Alpine.314.Arm32.Open)[email protected]/dotnet-buildtools/prereqs:alpine-3.14-helix-arm32v7-20210910135806-8a6f4f3	console.log			runclient.py
1929242	runtime	PR 73416	net7.0-Linux-Release-arm-CoreCLR_checked-(Alpine.314.Arm32.Open)[email protected]/dotnet-buildtools/prereqs:alpine-3.14-helix-arm32v7-20210910135806-8a6f4f3	console.log			runclient.py

Build Result Summary

Day Hit Count	Week Hit Count	Month Hit Count
0	0	2

[noahfalk]

@tlakollo - This test failed one other time earlier today in your build. It doesn't look like this arm leg runs in all builds so it could be your build introduced it, or the failure is non-deterministic and existed before, or it is determinstic but so many builds skip the arm leg we aren't seeing it much yet. I am trying to figure out which builds ran that leg.

[jkotas]

@noahfalk We have started getting intermittent arm32 crashes all over the place. E.g. #73433 #73247 #71624 (comment) . It looks like a GC hole that was introduced sometime this week. Could you please look into it? (I believe that you are still the test monitor.)

[filipnavara]

FWIW I tried to repro #73343. I setup an arm32 container on M1 MacBook Air with the exact same Helix image as used for the test run and the same downloaded payloads. All the xunit runs fail with NullReferenceException somewhere in generated MethodInvoker code when xunit tries to reflect over attributes. :-/

(may or may not be related but crashes with NREs and AVs all over the place)

Nevermind, apparently M1 chips don't support Arm32 natively so Docker probably uses QEMU or something, so it's unrelated issue.

[noahfalk]

Yep, I'll see what I can do.

[karelz]

Didn't happen for 2 weeks, removing blocking-clean-ci label

[jkotas]

This was likely fixed as part of #73405 (comment)