steakhal planned changes to this revision. steakhal added a comment. In D139195#3966675 <https://reviews.llvm.org/D139195#3966675>, @aaron.ballman wrote:
>> Previously, only void returning functions were considered for noreturn >> attribute candidates. This patch removes this artificial restriction. > > C2x 6.7.12.6p6: The implementation should produce a diagnostic message for a > function declared with a noreturn attribute that appears to be capable of > returning to its caller. > > p7 has an example showing `[[noreturn]] int h(void);` with the comment > "Implementations are similarly encouraged to diagnose the declaration of h() > because it appears capable of returning to its caller due to the non-void > return type." > > So is this really an artificial restriction? To me, putting the `noreturn` > attribute on a function with a return type makes no sense whatsoever. The > interface is saying "I promise that calling me will not return" and "when I > return, this is the type of the value I will give you." Hmm, I implicitly took the `__attribute__((noreturn))` suggestion and the non-void return type as a code-smell for the exact same reason. You are right, we should have a clean diagnostic message for these cases instead of this implied suggestion. IDK why I haven't thought about this 😅 Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D139195/new/ https://reviews.llvm.org/D139195 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
