llvm-beanz wrote: > any chance you could A/B test this on a bootstrap of clang, for instance? To > help validate that this really is NFC/dead code?
I did a little more archeology, and I'm actually struggling to find any point in the history where this change was required. The difference between the two sides of the if is really that if the declaration is a specialization declaration, it creates the pointed-to object then explicitly queries for the size and alignment before creating the pointer type. On the other side of the branch it calls `getOrCreateType` with the pointer type, which eventually calls into `CGDebugInfo::CreatePointerLikeType`, which as far back as I've looked (https://github.com/llvm/llvm-project/commit/443f6773029a45a705ae2873fb620904289513dd), has always included alignment and size. Here's a simplified diff of the two sides of the branch with comments and whitespace differences removed to make it more clear: ```diff --- /Users/cbieneman/dev/scratch/a 2024-12-12 09:29:27 +++ /Users/cbieneman/dev/scratch/b 2024-12-12 09:29:32 @@ -1,10 +1,4 @@ -const PointerType *ThisPtrTy = cast<PointerType>(ThisPtr); -uint64_t Size = CGM.getContext().getTypeSize(ThisPtrTy); -auto Align = getTypeAlignIfRequired(ThisPtrTy, CGM.getContext()); -llvm::DIType *PointeeType = - getOrCreateType(ThisPtrTy->getPointeeType(), Unit); -llvm::DIType *ThisPtrType = -DBuilder.createPointerType(PointeeType, Size, Align); +llvm::DIType *ThisPtrType = getOrCreateType(ThisPtr, Unit); TypeCache[ThisPtr.getAsOpaquePtr()].reset(ThisPtrType); ThisPtrType = DBuilder.createObjectPointerType(ThisPtrType); Elts.push_back(ThisPtrType); ``` Is there something I'm missing that explicitly setting the alignment handles? I feel like if `getOrCreateType` didn't properly handle the type's alignment we would be in a world of different hurt. https://github.com/llvm/llvm-project/pull/119445 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
