Fznamznon wrote: Thank you for the digging @zmodem !
> The linked bug does say "reverting it locally fixed the testing failures" > which is a pretty strong indication. I wanted to make sure there is no bugs in the test as well. > Now, in some other translation unit (not sure which yet), we decide that we > need the actual vector deleting destructor, and emit the definition for it. There are two cases when vector deleting destructor body is emitted: - It is exported via __declspec(dllexport) - new[] is called for the type. If the test was using __declspec(dllexport) on the destructor or class, we should never see the alias to scalar deleting destructor, so I think this is not the case. Some translation unit used new[] where destructor is only declared, but before this patch we were only emitting scalar deleting destructor for the class overall, which is also a bug, since a pointer produced by new[] should be deleted via delete[] which in absence of vector deleting destructor body would be problematic. So I would really like to see the reproducer still. Otherwise it sounds like there is no support for weak symbols in llvm/lib/CodeGen/AsmPrinter/WinCFGuard.cpp ? https://github.com/llvm/llvm-project/pull/185653 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
