[PATCH] D83468: [Debuginfo] Fix for PR46653

[Adrian Prantl via Phabricator via llvm-commits]

aprantl added a comment.

I'm skeptical of this patch. Removing the debug location entirely causes the instruction to "inherit" the location of the instruction before it, but it also completely throws away inline information for the instruction, which can be anywhere from misleading to asserting in the LLVM verifier (if the instruction is a call).

Can you post the output of llvm-dwarfdump --debug-line before and after the patch?

I fear that this patch only "fixes" the observable behavior by accident and the real problem you're after is somewhere else.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D83468/new/

https://reviews.llvm.org/D83468