JIT: Don't move EH entry block to enable branch reversal during flow opts

[amanasifkhalid]

Fixes #117581 (comment). cc @dotnet/jit-contrib, @AndyAyersMS PTAL. Thanks!

[Copilot]

Pull Request Overview

This PR refines flow optimization by preventing the movement of exception-handling entry blocks when enabling branch reversal.

• Inlines and extends the EH-region check in fgopt.cpp to disallow moving try/handler entries in addition to same-region blocks.
• Removes the now-unneeded helper fgEhAllowsMoveBlock from fgbasic.cpp and its declaration in compiler.h.

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
src/coreclr/jit/fgopt.cpp	Inlined EH checks and added disallow conditions for EH entry blocks
src/coreclr/jit/fgbasic.cpp	Removed fgEhAllowsMoveBlock implementation
src/coreclr/jit/compiler.h	Removed fgEhAllowsMoveBlock declaration

Comments suppressed due to low confidence (2)

src/coreclr/jit/fgopt.cpp:4520

• The comment mentions EH invariants but doesn't reflect the newly added checks for try/handler entry blocks. Consider updating it to enumerate these conditions or reference the specific invariants for clarity.

                        // However, we cannot move bDest if it will break EH invariants.

src/coreclr/jit/fgopt.cpp:4523

• Consider adding or updating JIT unit tests to cover branch reversal scenarios near EH boundaries (try/handler entries and call-finally pairs) to ensure this new logic prevents invalid block moves.

                            bDest->isBBCallFinallyPair())

[dotnet-policy-service]

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

[amanasifkhalid]

No diffs.

[amanasifkhalid]

/ba-g android-x64 CoreCLR build stuck