aaron.ballman added a comment. In D150226#4404808 <https://reviews.llvm.org/D150226#4404808>, @rupprecht wrote:
> I suppose including this warning in `ShowInSystemHeader` with your diff above > would be a good first step anyway, right? If we're about to make it a hard > error anyway, then making it a `ShowInSystemHeader` error first would ease > that transition. Yes and no. If we're going to turn something into a hard error, letting folks implementing system headers know about it is important, so from that perspective, it makes sense to enable the diagnostic in system headers. However, *users* have no choice in their system headers (oftentimes) which makes the diagnostic unactionable for them as they're not going to (and shouldn't have to) modify system headers, so from that perspective, enabling the diagnostic in a system header introduces friction. If this diagnostic is showing up in system headers, I think we would likely need to consider adding a compatibility hack to allow Clang to still consume that system header while erroring when outside of that system header (we've done this before to keep STL implementations working, for example). Between this need and the friction it causes users to have an unactionable diagnostic, I think we probably should not enable `ShowInSystemHeader`. CHANGES SINCE LAST ACTION https://reviews.llvm.org/D150226/new/ https://reviews.llvm.org/D150226 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
