[clang] [llvm] Fix Windows EH IP2State tables (remove +1 bias) (PR #144745)

[via llvm-commits]

================
@@ -37,9 +37,11 @@ cond.end:                                         ; preds = %entry, %cond.true
 ; CHECK: testq
 ; CHECK: je
 ; CHECK: callq alloc
+; CHECK-NEXT: nop
----------------
sivadeilra wrote:

I implemented the more-accurate optimizations for NOP padding (only inserting when we're also building IP2State tables).  It had the benefit of removing NOPs from a lot of codegen situations that were not even using EH.  I've pushed an update to the PR as a separate commit, which should make it a bit easier to review.

https://github.com/llvm/llvm-project/pull/144745