rsmith added a comment. Are you planning on extending this to also cover templated functions?
================ Comment at: include/clang/Basic/DiagnosticGroups.td:78 @@ -77,2 +77,3 @@ DiagGroup<"gnu-string-literal-operator-template">; +def UnavailableTemplate : DiagGroup<"unavailable-template">; ---------------- undefined rather than unavailable? "unavailable" means something different in Clang. ================ Comment at: include/clang/Basic/DiagnosticSemaKinds.td:3826 @@ +3825,3 @@ +def warn_var_template_missing : Warning<"instantiation of %q0 required here, " + "but corresponding template is not found">, InGroup<UnavailableTemplate>; +def note_inst_declaration_hint : Note<"add an explicit instantiation declaration " ---------------- not found -> not defined? The template is found, or the code would be ill-formed. Also, 'corresponding template' isn't quite right; in `S<int>::m`, the template is `S`, which must be defined. Maybe you could map to the pattern declaration for the instantiation, and warn that it isn't defined, so we'd say something like "instantiation of 'S<int>::m' required here, but 'S<T>::m' is not defined" ================ Comment at: lib/Sema/SemaTemplateInstantiateDecl.cpp:3909 @@ +3908,3 @@ + << Var; + Diag(PointOfInstantiation, diag::note_inst_declaration_hint) << Var; + } ---------------- Explicit instantiation declarations are a C++11 feature. Should we suppress this warning, or at least the note, in C++98? http://reviews.llvm.org/D16396 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
