SPMI: Spurious diffs when stressing profiler ELT callbacks

[jakobbotsch]

When stressing ELT callbacks we generate code that points to a function that resides within clrjit.dll:

runtime/src/coreclr/jit/compiler.cpp

Line 3114 in 3a36920

compProfilerMethHnd = (void*)DummyProfilerELTStub;

This is not compatible with SPMI replay/diffs.
Likely the easiest fix is to introduce a new CORINFO_HELP_PROF_FCN_TEST_CALLBACK that is a NOP and defined in the runtime, similar to how CFG is stress tested today.

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

Issue Details

---

When stressing ELT callbacks we generate code that points to a function that resides within clrjit.dll:

runtime/src/coreclr/jit/compiler.cpp

Line 3114 in 3a36920

compProfilerMethHnd = (void*)DummyProfilerELTStub;

This is not compatible with SPMI replay/diffs.
Likely the easiest fix is to introduce a new CORINFO_HELP_PROF_FCN_TEST_CALLBACK that is a NOP and defined in the runtime, similar to how CFG is stress tested today.

Author:	jakobbotsch
Assignees:	-
Labels:	area-CodeGen-coreclr, untriaged
Milestone:	-

[jakobbotsch]

cc @dotnet/jit-contrib -- if you see spurious same-size diffs in 2 contexts in coreclr_tests, then it's most likely because of this issue.

[jakobbotsch]

As a temporary workaround we can also change the test in #81725 to avoid specifying DOTNET_JitStress=1, and to specify DOTNET_JitStressNames with the necessary stress modes to reproduce the issue.

[jakobbotsch]

Fixed by #86132