[RuntimeAsync] Feedback from the merging PR in the runtime repo.

[VSadov]

This is to keep track of the changes done to address PR feedback in dotnet/runtime#114675

[VSadov]

I think it is time to merge this as we need to update to the latest API shape (i.e. AsyncHelpers)

[jakobbotsch]

Does it mean the EH test fails on x86 now? Can we disable it somehow?

[VSadov]

x86 does not do FEATURE_EH_FUNCLETS right now so it will work. Thus we can postpone introducing IL_ThrowExact_x86.

When x86 is moved onto FEATURE_EH_FUNCLETS plan, which I understand will happen at some point, we can remove HELPER_METHOD_FRAME_BEGIN_ATTRIB_NOPOLL path, which is desirable, but will also need to implement IL_ThrowExact_x86. It takes an extra parameter, I have only vague idea why.

[VSadov]

All runtime tests should be passing right now, including async and including EH - that is on win-x64 and win-x86, where it is easy to test once in a while.

[jkotas]

When x86 is moved onto FEATURE_EH_FUNCLETS plan, which I understand will happen at some point,

It is expected to happen very soon. The changes are in dotnet/runtime#113576. The last remaining item before this gets enabled is private diagnostic tests run. If you would like to get ahead of this, you can apply the 10 line change locally and do the required fixes to make things work.

[am11]

Context about last testing item: dotnet/runtime#113985 (comment).

[filipnavara]

It takes an extra parameter, I have only vague idea why.

win-x86 doesn't have unwinding information for native code. Thus we need to capture the precise register context in assembly code before giving control to the C code since we cannot run VirtualUnwind to get it.

This should be fairly mechanical change so I can make a PR for the code.

[filipnavara]

#3100

[jakobbotsch]

LGTM

[VSadov]

Thanks!