ahatanak marked 10 inline comments as done. ahatanak added inline comments.
================ Comment at: include/clang/AST/DeclCXX.h:443 + /// SMF_MoveConstructor, and SMF_Destructor are meaningful here. + unsigned HasTrivialSpecialMembersForCall : 6; + ---------------- rjmccall wrote: > I think you could probably get away with only three bits here (and below) if > you reorder those three values to be the first three. I don't know if that's > important in terms of packing DefinitionData at all. Should be okay to not > do for now. Yes, only three bits are needed here. I'll fix this later. ================ Comment at: include/clang/AST/DeclCXX.h:478 + /// ill-formed if it is annotated with "trivial_abi". + unsigned DropTrivialABI : 1; + ---------------- rjmccall wrote: > So is it actually interesting to drop the attribute from the AST? And why is > this bit necessary vs., say, just checking whether the attribute is present > but the class is not trivial-for-calls? I think we need to check whether a struct annotated with "trivial_abi" is ill-formed before ActOnFields and CheckCompletedCXXClass set the triviality flags in Sema::ActOnFinishCXXMemberSpecification and drop the attribute if it is ill-formed. If we don't want to drop the attribute, we will have to find out whether "trivial_abi" has no effect before calls to FunctionDecl::setTrivialForCall are made. I added this bit to CXXRecordDecl in the previous patch because I felt hesitant to iterate over the base classes and members of the class again, but I think that is better than adding a bit just to check whether the class is ill-formed. I've removed the bit in the new patch. https://reviews.llvm.org/D41039 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
