bogner wrote: This doesn't look right to me. Changing the type printer so it doesn't differentiate between named and anonymous structs inside a target extension type like this looks like it will break round-tripping through LLVM IR, since the fact that this was associated with a named struct and not an anonymous one is lost.
I suspect we're actually creating the wrong types (in DirectXTargetCodeGenInfo::getHLSLType) in the first place, since target extension types don't seem to handle named structs at all. Additionally, we should probably add checks when creating a target extension type to prevent creating them with named struct types. https://github.com/llvm/llvm-project/pull/115971 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
