[mono] Default Interface Implementation of Static Virtual Methods

[lambdageek]

See the CoreCLR PR #66887 that adds the svm_diamondshape test case.

Mono should support default interface implementations of static virtual methods (including overrides).

[lambdageek]

/cc @SamMonoRT @BrzVlad

[lambdageek]

Related #80350 (comment) and #54399

[BrzVlad]

Some more context. This issue and the ones @lambdageek mentioned are probably duplicates for the svm_diamondshape runtime test failing on mono. The test in question passes if we mark the reabstraction of Func in I4Reabstract as final. From what I've seen, Roslyn also marks these as final and other IL runtime tests do the same. Mono currently considers these static virtual methods as being reabstracted if they are both abstract and final, in which case we don't allocate a newslot for them. Discussion around this was started in #83189, but to me it is still not exactly clear what needs to be done on mono side.

From what I could find, there is this comment in CoreCLR code "Re-abstractions are virtual static abstract with a MethodImpl". Does this mean that, for reabstraction, rather than checking for final flag, we should check in the metadata if the static virtual abstract method has an explicit .override instead ? @lambdageek any idea how easy would be to add this check, do we have anything cached somewhere ? @davidwrighton Could you confirm what we should do here so we can close these issues ?

[BrzVlad]

Doesn't look like we'll get this addressed in .net 8, moving milestone. Also this issue is present only when writing IL code, not from C#, so it is not critical.