aaronpuchert added a comment. In D110216#3041117 <https://reviews.llvm.org/D110216#3041117>, @mizvekov wrote:
> In D110216#3040748 <https://reviews.llvm.org/D110216#3040748>, @craig.topper > wrote: > >> Looks like this fixes PR51282. > > I guess it does fix it, but the underlying implementation of alignment is > very fragile. The alignment is stored in `AttributedType` nodes, which are > always sugar, so they are not considered part of the type in a structural > sense... Unless I'm missing something, alignment on aliases is itself fragile. Let's say you have some template, then the instantiation with the alias is of course the same as the instantiation with the canonical type, so if that template has a member or local variable of that type parameter that won't get the alignment. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D110216/new/ https://reviews.llvm.org/D110216 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
