JIT: Sign/zero-extend small primitive arguments for pinvokes on arm32 and Apple arm64

[jakobbotsch]

Fix #101046

Alternative to #106312 that keeps the behavior as is for the managed calling convention and only changes it for pinvokes. This requires changing fcalls in the VM to access small primitive typed arguments without assumptions about normalization.

[jkotas]

Delete?

[jakobbotsch]

Looks like it's still used as the return value for HandleTableRestrictedCallbackFunction.

[jkotas]

And TryGetTaggedMemoryCallback.

These correspond to the following managed methods:

• runtime/src/coreclr/nativeaot/System.Private.CoreLib/src/System/Runtime/InteropServices/ComWrappers.NativeAot.cs

  Lines 428 to 434 in 37284fc

  	[UnmanagedCallersOnly]
  	private static bool IsRootedCallback(IntPtr pObj)
  	{
  	// We are paused in the GC, so this is safe.
  	ManagedObjectWrapperHolder* holder = (ManagedObjectWrapperHolder*)&pObj;
  	return holder->_wrapper->IsRooted;
  	}

• runtime/src/coreclr/nativeaot/System.Private.CoreLib/src/System/Runtime/InteropServices/ObjectiveCMarshal.NativeAot.cs

  Lines 58 to 59 in b1968e7

  	[RuntimeExport("ObjectiveCMarshalTryGetTaggedMemory")]
  	private static bool TryGetTaggedMemory(IntPtr pObj, IntPtr* tagged)

CLR_BOOL is not quite right for neither of these, but I do not think that this mismatch is going to cause any real problem. It would be nice to clean this up in a follow up.

[JulieLeeMSFT]

@AndyAyersMS, PTAL. This is for RC1 snap.

[jakobbotsch]

cc @dotnet/jit-contrib PTAL @AndyAyersMS @jkotas

Minor diffs

[AndyAyersMS]

Jit changes LGTM.

[jkotas]

The comment around line 150 in this file still mentions CLR_BOOL.

[jakobbotsch]

Mind if I update that in a separate change tomorrow?

[jkotas]

Fine with me

[jakobbotsch]

Nevermind, updated it as well now due to those other places I missed the conversion.

[jkotas]

These places are missing the CLR_BOOL -> FC_BOOL_ARG conversion:

\runtime\src\coreclr\vm\comutilnative.h(43):static FCDECL3(StringObject *, StripFileInfo, Object *orefExcepUNSAFE, StringObject *orefStrUNSAFE, CLR_BOOL isRemoteStackTrace); - dead code, shoul be deleted
\runtime\src\coreclr\vm\debugdebugger.cpp(236):CLR_BOOL fNeedFileInfo,
\runtime\src\coreclr\vm\debugdebugger.h(155):CLR_BOOL fNeedFileInfo,

[jkotas]

LGTM otherwise. Thanks!

[JulieLeeMSFT]

cc @carlossanlop to include this in RC1 snap if CI tests pass.

[carlossanlop]

Thanks. I'm aware, it's in my list of PRs to wait. @jakobbotsch reached out earlier.

[jakobbotsch]

The runtime pipeline is done (some of the jobs didn't have their status updates, presumably because of the outage). Build analysis is happy. We can skip waiting for the superpmi jobs – they were green in the previous run and there hasn't been any JIT changes since then.