Update dependencies from dotnet/roslyn

[am11]

Testing #110084 with build fixes.

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-meta
See info in area-owners.md if you want to be subscribed.

[am11]

@jjonescz, without this change, the tests were consistently failing:

$ ./build.sh libs+libs.tests
$ ./dotnet.sh build -t:Test src/libraries/System.Collections.Concurrent/tests
...
    Discovering: System.Collections.Concurrent.Tests (method display = ClassAndMethod, method display options = None)
    Discovered:  System.Collections.Concurrent.Tests (found 901 of 909 test cases)
    Starting:    System.Collections.Concurrent.Tests (parallel test collections = on [10 threads], stop on fail = off)
      System.Collections.Concurrent.Tests.ConcurrentStackTests.Enumerator_MoveNext_AfterDisposal(count: 1) [FAIL]
        System.NullReferenceException : Object reference not set to an instance of an object.
        Stack Trace:
          /Users/am11/projects/runtime5/src/libraries/System.Collections.Concurrent/src/System/Collections/Concurrent/ConcurrentStack.cs(743,0): at System.Collections.Concurrent.ConcurrentStack`1.GetEnumerator(Node head)+MoveNext()
          /Users/am11/projects/runtime5/src/libraries/Common/tests/System/Collections/IEnumerable.Generic.Tests.cs(395,0): at System.Collections.Tests.IEnumerable_Generic_Tests`1.Enumerator_MoveNext_AfterDisposal(Int32 count)
             at InvokeStub_IEnumerable_Generic_Tests`1.Enumerator_MoveNext_AfterDisposal(Object, Span`1)
          /Users/am11/projects/runtime5/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodBaseInvoker.cs(95,0): at System.Reflection.MethodBaseInvoker.InvokeWithOneArg(Object obj, BindingFlags invokeAttr, Binder binder, Object[] parameters, CultureInfo culture)
      System.Collections.Concurrent.Tests.ConcurrentStackTests.Enumerator_MoveNext_AfterDisposal(count: 75) [FAIL]
        System.NullReferenceException : Object reference not set to an instance of an object.
        Stack Trace:
          /Users/am11/projects/runtime5/src/libraries/System.Collections.Concurrent/src/System/Collections/Concurrent/ConcurrentStack.cs(743,0): at System.Collections.Concurrent.ConcurrentStack`1.GetEnumerator(Node head)+MoveNext()
          /Users/am11/projects/runtime5/src/libraries/Common/tests/System/Collections/IEnumerable.Generic.Tests.cs(395,0): at System.Collections.Tests.IEnumerable_Generic_Tests`1.Enumerator_MoveNext_AfterDisposal(Int32 count)
             at InvokeStub_IEnumerable_Generic_Tests`1.Enumerator_MoveNext_AfterDisposal(Object, Span`1)
          /Users/am11/projects/runtime5/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodBaseInvoker.cs(95,0): at System.Reflection.MethodBaseInvoker.InvokeWithOneArg(Object obj, BindingFlags invokeAttr, Binder binder, Object[] parameters, CultureInfo culture)
    Finished:    System.Collections.Concurrent.Tests
  === TEST EXECUTION SUMMARY ===
     System.Collections.Concurrent.Tests  Total: 2460, Errors: 0, Failed: 2, Skipped: 0, Time: 74,172s
...

is it expected?

[jjonescz]

@jcouv is this due to dotnet/roslyn#75908?

[jcouv]

Yes, that is the cause.

Here's a sharplab repro.

In the "main" branch, we NRE.
In the "default" branch (ie. before the change to clear locals during disposal), we printed a result without crash. But it looks like that result was incorrect already...
We were incorrectly allowing enumeration to continue.

The correct behavior would be for disposal to set the enumerator state to "after" (thereby causing MoveNext to return false thereafter).
This issue becomes more observable after the change to clear local state (NRE in the present scenario).
Filed dotnet/roslyn#76078

I will discuss with @AlekseyTs @333fred on how best to proceed. I thinking to revert the change temporarily and fix the disposal bug first.

[jcouv]

@am11 There's a couple of different ways we can proceed. Let me propose one:
I'm thinking to keep the roslyn change and fix the unearthed bug asap. Due to Thanksgiving, this means I could probably have an updated build of roslyn by next Wednesday (Dec 4th).
So you would either have to put this PR on hold until then, or you'd have to temporarily skip the failing tests (that do Dispose then MoveNext on compiler-generated enumerators) to merge this PR.
Does that sound okay?

[am11]

@jcouv, thanks for the info. I think we can wait, since there are some other NRE failures in CI which appear to be related.

[jcouv]

That works, thanks.
FWIW, here's the fix in progress