llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT--> @llvm/pr-subscribers-clang Author: Matheus Izvekov (mizvekov) <details> <summary>Changes</summary> Issue #<!-- -->178324 was actually fixed by #<!-- -->187755 We lost the "declaration does not declare anything" warning since the regression was introduced, but that was because: 1) Since #<!-- -->78436 we treat __builtin_FUNCSIG in a dependent context effectivelly as if it contained a template parameter. 2) Our decltype implementation treats eexpressions containing template parameters as if they were completely opaque (but alas this goes against the spec, which says in [temp.type]p4 this should be looking only at type dependence). 3) Since the decltype is opaque, we don't know what lookup will find, so we can't issue the warning because we don't know if we are going to end up with a type or an expression. Fixes #<!-- -->178324 --- Full diff: https://github.com/llvm/llvm-project/pull/190490.diff 1 Files Affected: - (modified) clang/test/SemaCXX/source_location.cpp (+10) ``````````diff diff --git a/clang/test/SemaCXX/source_location.cpp b/clang/test/SemaCXX/source_location.cpp index c8b65d6eab50d..eaa6cb04c5d1c 100644 --- a/clang/test/SemaCXX/source_location.cpp +++ b/clang/test/SemaCXX/source_location.cpp @@ -1081,3 +1081,13 @@ static_assert(X{}. static_assert(X{}. foo() == 10001); } + +#ifdef MS +namespace GH178324 { + struct a { + using e = int; + }; + void current(const char * = __builtin_FUNCSIG()); + template <class> void c() { decltype(a(current()))::e; } +} // namespace GH178324 +#endif `````````` </details> https://github.com/llvm/llvm-project/pull/190490 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
