Fix mach-o file vs segment size

[jkoritzinsky]

Make sure that segment vmsize is never smaller than segment file size

Fixes #124982

cc: @filipnavara

[Copilot]

Pull request overview

This PR addresses Mach-O validity issues in Crossgen2 composite R2R output by ensuring the first LC_SEGMENT_64 load command reports consistent segment sizes so Mach-O inspection tools (e.g., llvm-objdump, otool) don’t reject the file.

Changes:

• Updates Mach-O section layout logic to ensure the segment vmsize is not smaller than the segment filesize.

[filipnavara]

This doesn't look right. However, I admit that the calculation for segmentFileSize also doesn't look right to me since it never subtracts the initial file offset (of the first section that is marked as InFile).

[filipnavara]

I suppose the correct fix would look something like this:
filipnavara@3bbbd9c229a

(on second thought, maybe the initial offset should be the aligned one, so maybe more like filipnavara@c48520c4d77)

[jkoritzinsky]

I suppose the correct fix would look something like this: filipnavara@3bbbd9c229a

(on second thought, maybe the initial offset should be the aligned one, so maybe more like filipnavara@c48520c4d77)

I took your second fix

[filipnavara]

This is almost correct. I just forgot that now we also need to update the segmentFileOffset (https://github.com/jkoritzinsky/runtime/blob/6b21a40967681ec4cb792600d80e0f0943cb1928/src/coreclr/tools/Common/Compiler/ObjectWriter/MachObjectWriter.cs#L205) with the correctly aligned value that we computed in the initialFileOffset variable.

I can submit a full tested fix tomorrow unless you beat me to it. Sorry for the hassle and not catching this right away.