[NativeAOT, win-x86] Bad EH info generated for nested trys

[MichalStrehovsky]

Hit with JIT\Regression\CLR-x86-JIT\V1-M10\b08944\b08944b test.

JitDump of the method with bad EH info:
jitdump.zip

Notice:

2 EH table entries, 0 duplicate clauses, 0 cloned finallys, 2 total EH entries reported to VM
setEHcount(cEH=2)
EH#0: try [0029..0032) handled by [008D..00B2) (class: 1000014)
EH#1: try [0028..0034) handled by [00B2..016D) (class: 1000014)

The 29..32 try block looks okay, but the 28..34 is too short and doesn't cover the handler that got moved out-of-line.

Wondered if it's related to a // TODO: Is this really UNIX_X86_ABI specific? Should we guard with compiler->UsesFunclets() instead? comment but that didn't seem to fix it.

[dotnet-policy-service]

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

[EgorBo]

@MichalStrehovsky is this important for 9.0 milestone or it can wait for 10.0? And is naot-win-32 now an officially supported target?

[jkotas]

Yes, naot-win32 is officially supported target.

[JulieLeeMSFT]

@jkotas, we will not have time to work on this before RC1 snap. Do you want to see this for RC2 snap or can we do it in .NET 10?

Yes, naot-win32 is officially supported target.

[jkotas]

I do not think that this issue is .NET 9 must have.