jmorse wrote: Overall I feel OK about this patch as it stands, mostly because solving a problem in one place (the overall function-local-types project) is creating problems in a smaller / rarer place. I feel we should have plan for fixing GenerateVarArgsThunk though, to ensure we're not backing into a corner we can't get out of. From my reading of the previous tickets, particularly https://github.com/llvm/llvm-project/issues/33277, it sounds like this arrangement was implemented to avoid too much metadata cloning, and I wonder whether LLVM does that any better 8 years later. Or perhaps we can implement the originally-desired solution which was to defer when these functions are generated.
@wolfy1961 do you have any recollection of https://reviews.llvm.org/D39396 and whether the issue with cloning temporary debug-info might have changed in the meantime? https://github.com/llvm/llvm-project/pull/167758 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
