================ @@ -179,6 +179,8 @@ def err_opencl_unknown_type_specifier : Error< def warn_unknown_attribute_ignored : Warning< "unknown attribute %0 ignored">, InGroup<UnknownAttributes>; +def ext_unknown_attribute_ignored : Extension< + "unknown attribute %0 ignored">, InGroup<UnknownAttributes>; ---------------- AaronBallman wrote:
> Curiously, and after I sent that, I realized why WE of all compilers SHOULD > warn here, even if others don't. Clang supports/recognizes the prefixes of > all of the major compilers, which drastically increases the chance that an > unrecognized one is a typo rather than an intentional difference. +1; the current behavior of warning even when we don't know the namespace is entirely intentional for exactly that reason. I think it's reasonable for us to have a new warning group so if there is an attribute namespace heavily used in a project, the developer can disable just the unknown namespace ones while still getting other diagnostics about unknown attributes. However, what should the user do when they want to silence diagnostics about `[[frobble::gobble]]` but not silence diagnostics about `[[edg::whatever]]`? Do we want to support a grouping like `-Wno-unknown-attribute-namespace=frobble` where leaving off the `=` means we silence all unknown attribute namespaces? https://github.com/llvm/llvm-project/pull/120925 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
