Fix ABIStress test hashing padding

[jakobbotsch]

ABIStress was storing data inside of padding and hashing it when trying to determine if all values made it through correctly. Make sure we only hash the fields.

With physical promotion enabled this causes the test to fail. That's because of this type:

runtime/src/tests/JIT/Stress/ABI/Types.cs

Line 65 in 19d2e90

publicstructI128_2{publicbyteF0;publicInt128F1;publicI128_2(bytef0,Int128f1)=>(F0,F1)=(f0,f1);}

which has 15 bytes of padding in it. When the constructor is invoked, a temp is created, and physical promotion ends up promoting that temp and only writing the significant "field" parts.

I verified that regular promotion also causes the test to fail if a type is added that is regularly promoted, e.g. with struct S { public byte F0; public long F1; } the same problem is hit on main.

[ghost]

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

Issue Details

---

ABIStress was including padding in its hashing when trying to determine if all values made it through correctly. Make sure we only hash the fields that actually store data.

With physical promotion enabled this causes the test to fail. That's because of this type:

runtime/src/tests/JIT/Stress/ABI/Types.cs

Line 65 in 19d2e90

publicstructI128_2{publicbyteF0;publicInt128F1;publicI128_2(bytef0,Int128f1)=>(F0,F1)=(f0,f1);}

which has 15 bytes of padding in it.

I verified that regular promotion also causes the test to fail if a type is added that is regularly promoted, e.g. with struct S { public byte F0; public long F1; } the same problem is hit on main.

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

[jakobbotsch]

cc @dotnet/jit-contrib PTAL @AndyAyersMS