rjmccall wrote: It's certainly possible to see push/pop pragmas in late-parsed code, but Sema should immediately emit an error and ignore them because, like I said, they're only allowed at file context, and AFAIK late-parsed code is never at file context. If there's some sneaky way to late-parse file-context declarations, that seems like a serious problem, because it's very important to process pragmas in order as we parse the top level. And contrariwise, if there's some sneaky way to put push/pop pragmas in non-file contexts, that also seems like a serious problem, because the way we process them is not designed to understand local scopes, which means we're just doing random stuff instead of implementing any sort of intelligible language design.
Now, I don't really mind the late-parsing code being overly cautious about things possibly pushing and popping on the pragma stack as long as it's not costing us a significant amount of performance, but it would also be fine for it to assert that that hasn't happened and then just save and restore the active state, which is the only thing that late parsing should ever touch. https://github.com/llvm/llvm-project/pull/70646 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
