[llvm] [DebugInfo] Don't set prologue_end behind line-zero call insts (PR #156850)
David Blaikie via llvm-commits
llvm-commits at lists.llvm.org
Mon Nov 17 09:08:52 PST 2025
dwblaikie wrote:
> > What about if we forced the location after the prologue, if it's line zero, to instead have the function's scope start location.
>
> I think he was saying the call should get scope-line instruction? I could be misinterpreting this conversation though, it would be good to get confirmation.
Oh, I think I'd misunderstood some more fundamental difference here - yeah, whether it's the start of the function line, or the start of the scope line, if the call happens to be within some narrower scope (like if the call was inlined from another function, that'd be especially relevant - whether it's part of the inline or not).
Yeah, I guess scope sounds marginally better on that basis.
Though we do produce file/line on DILexicalScopes and the subprgoram scopes of inlined instructions have that information to work with - we don't actually emit the file/line in a DW_TAG_lexical_scope (so we could consider dropping them from the metadata too). So maybe this distinction would only matter for inlined/subprogram scopes.
https://github.com/llvm/llvm-project/pull/156850
More information about the llvm-commits
mailing list