AaronBallman wrote: Thank you for your patience!
In general, I think adding a new qualifier to Clang needs to come with significant justification for the implementation and maintenance costs. In this specific case, those costs are somewhat amortized because we already have the `__ptrauth` qualifier and this piggybacks off that work. But I don't think the benefits justify the costs: if the user writes `__ptrauth` on an integer type, we can determine whether that type is "good enough" or not and issue a diagnostic for invalid or suspicious uses, same as if the user spelled it `__ptrauth_restricted_intptr`. The implementation already has to update typechecking in various places regardless of which way we spell the qualifiers, so there should not be too much extra implementation burden from only using one qualifier. So on balance, I think we should only upstream the single qualifier and use a header file + feature tests for compatibility with older compilers. Is that something you can get behind? (You're still the expert in this space, so if I'm trivializing concerns, please speak up!) https://github.com/llvm/llvm-project/pull/137580 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
