aaron.ballman added a comment. In D158472#4605228 <https://reviews.llvm.org/D158472#4605228>, @hazohelet wrote:
> This breaks the one-note-for-one-overload-candidate rule of overload > resolution failure diagnostics > (https://github.com/llvm/llvm-project/blob/ff08c8e57e39d7970b65637595cdc221901f4ed1/clang/lib/Sema/SemaOverload.cpp#L11517-L11526), > but in most cases this change would produce less lines of diagnostics. > If we don't like this additional note, we could instead suppress the fix-it > diagnostics when the fix-it location and the diagnosed candidate location are > not close (for example, more than 2 lines away). I'm uncomfortable with starting down the slippery slope of multiple notes per overload resolution failure; that leads to an explosion of extra diagnostics which makes it harder to spot the issue in the first place. My inclination is to try to resolve this by fixing the way we emit fix-it diagnostics so that we remove the extra whitespace from there because this seems like it's a general problem with fix-its. Or am I wrong about this being a general issue? CC @cjdb for opinions as well ================ Comment at: clang/test/Sema/overloadable.c:31 + accept_funcptr(f2); // expected-error{{no matching function for call to 'accept_funcptr'}} \ + // expected-note 2{{take the address of the argument with &}} } ---------------- Something odd is going on -- why is the note emitted twice? CHANGES SINCE LAST ACTION https://reviews.llvm.org/D158472/new/ https://reviews.llvm.org/D158472 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
