JIT: fix issue with EH clause class types for fault/filters from C++/CLI

[AndyAyersMS]

C++/CLI appears to leave the CORINFO_EH_CLAUSE ClassToken/FilterOffset union set to some nonzero value for fault/filter clauses. The JIT currently just passes this value along to the runtime.

If a method with such a nonzero field is inlined into a dynamic method, this trips up a check in the runtime where a nonzero entry for such a field is interpreted as a class handle, even for fault/filter clauses where it should be ignored.

Tolerate this by zeroing the field in the JIT.

Note this could not have happened in pre .NET10 as methods with EH could not be inlined, so a dynamic method would never see such an EH clause, and in non-dynamic methods this field is ignored for faults and filters.

Fixes #118837.

[Copilot]

Pull Request Overview

This PR fixes an issue where C++/CLI generated code leaves nonzero values in the ClassToken/FilterOffset union field for fault/filter exception handling clauses. When such methods are inlined into dynamic methods in .NET 10+, the runtime incorrectly interprets these nonzero values as class handles, causing validation failures.

• Adds explicit zeroing of the ebdTyp field for finally and fault exception handling clauses
• Prevents runtime validation errors when C++/CLI methods with EH are inlined into dynamic methods
• Addresses a new issue introduced in .NET 10 when method inlining with exception handling was enabled

[AndyAyersMS]

@jakobbotsch PTAL
cc @dotnet/jit-contrib

I could (and perhaps should) also fix this in the jit interface (to ignore the value), and/or get the C++/CLI output fixed.

We will want to propose this for a .NET 10 backport.

Also I haven't added a test case; if anyone can point me at the right way to have a dependent C++/CLI library I will do so (many JIT tests just check in IL instead, which I could also do).

[jkotas]

this trips up a check in the runtime

I assume that it happened here: https://github.com/dotnet/runtime/blob/main/src/coreclr/vm/jitinterface.cpp#L12736-L12738 . We can change the condition to
(pEHClause->Flags & {COR_ILEXCEPTION_CLAUSE_FILTER | COR_ILEXCEPTION_CLAUSE_FINALLY | COR_ILEXCEPTION_CLAUSE_FAULT) == 0

[jkotas]

We can change the condition to

And also in CorInfoImpl.ReadyToRun.cs

[jkotas]

many JIT tests just check in IL instead, which I could also do

I am not sure you can convince ilasm to leave bogus value in unused EHInfo fields.

If you wanted to deterministically exercise a situation with a bogus value in unused EHInfo fields, I think you would have to emit the binary using System.Reflection.Metadata. I do not think it is it worth it.

[dotnet-policy-service]

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

[jakobbotsch]

Can we always zero out the entirety of compHndBBtab when we allocate/reallocate it?

[AndyAyersMS]

/ba-g System.Security.Cryptography.Tests dead-lettered

[AndyAyersMS]

Can we always zero out the entirety of compHndBBtab when we allocate/reallocate it?

Seems prudent. Let me add that.

[AndyAyersMS]

Seeing unrelated failures in OSX host activation tests

[xUnit.net 00:02:30.00]       --- Resolving libhostpolicy.dylib version from deps json [/private/tmp/helix/working/B1FB09EB/w/C75E0A16/e/test_artifacts/iet2mkkp.fd2/App/App.deps.json]
[xUnit.net 00:02:30.00]       Dependency manifest /private/tmp/helix/working/B1FB09EB/w/C75E0A16/e/test_artifacts/iet2mkkp.fd2/App/App.deps.json does not contain an entry for runtime.osx-x64.Microsoft.NETCore.DotNetHostPolicy
[xUnit.net 00:02:30.00]       The expected libhostpolicy.dylib directory is [/private/tmp/helix/working/B1FB09EB/w/C75E0A16/e/test_artifacts/iet2mkkp.fd2/App/]
[xUnit.net 00:02:30.00]       Tracing enabled @ Tue Aug 19 19:01:40 2025 UTC
[xUnit.net 00:02:30.00]       Reading from host interface version: [0x16041101:248] to initialize policy version: [0x16041101:248]
[xUnit.net 00:02:30.00]       ' could not be formatted with string.Format
[xUnit.net 00:02:30.00]          at System.String.FormatHelper(IFormatProvider provider, String format, ReadOnlySpan`1 args)
...
[xUnit.net 00:02:30.01]            at System.Reflection.RuntimeMethodInfo.Invoke(Object obj, BindingFlags invokeAttr, Binder binder, Object[] parameters, CultureInfo culture)
  Failed Microsoft.DotNet.CoreSetup.Test.HostActivation.DependencyResolution.LocalPath.RuntimeAssemblies_SelfContained(useLocalPath: True) [28 s]
  Error Message:
   **WARNING** failure message 'Expected command to pass but it did not.

Opened #118904

[AndyAyersMS]

/backport to release/10.0-rc1

[github-actions]

Started backporting to release/10.0-rc1: https://github.com/dotnet/runtime/actions/runs/17081234351