Fix layout and GCInfo for ByRef fields.

[AaronRobinsonMSFT]

Update tests to use recursion and validate GCStress scenarios.

Related NativeAOT/CrossGen2 change - #64366

Fixes #64465
Fixes #64466

/cc @MichalStrehovsky @dotnet/interop-contrib

[MichalStrehovsky]

Tangentially related to the topic of GC holes, is there a similar problem in FindByRefPointerOffsetsInByRefLikeObject? That one is used in stack walking. Not sure how to hit that one... might be something along the lines of an interface call that takes a struct with ref fields as a parameter? It might be a good strategy to search for all occurrences of g_pByReferenceClass.

Found that one by accident when I wanted to point you to #12842 (that we should also do for byref fields due to GC tracking).

Sorry for dumping tangentially related stuff on this pull request - it should have probably gone to #63985.

[AaronRobinsonMSFT]

Tangentially related to the topic of GC holes, is there a similar problem in FindByRefPointerOffsetsInByRefLikeObject?
It might be a good strategy to search for all occurrences of g_pByReferenceClass.

@MichalStrehovsky That is exactly what I've been doing for the past 40 minutes. I agree there are some issues here. Based on my searching, I think FindByRefPointerOffsetsInByRefLikeObject is the only other thing that needs to be changed. I was going to defer until we remove ByReference<T> but since you brought it up, I will apply the fixes to the other places.

it should have probably gone to #63985.

Agree.

[jkotas]

LGTM

[AaronRobinsonMSFT]

@MichalStrehovsky I am checking this in now since it is impacting the outer loop tests. Please still give it a once over in case there is something else I'm not considering.

[MichalStrehovsky]

Looks good! I filed #64520 for one of the tangentially related issues.