https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97407
Bug ID: 97407 Summary: Expanding alias template in concept satisfaction error is undesirable Product: gcc Version: 10.2.1 Status: UNCONFIRMED Keywords: diagnostic Severity: normal Priority: P3 Component: c++ Assignee: unassigned at gcc dot gnu.org Reporter: redi at gcc dot gnu.org Target Milestone: --- Compiling this with -std=c++20: #include <ranges> template<typename T> concept False = false; template<typename R> requires False<std::ranges::iterator_t<R>> void f(R&) { } int main() { int a[2]; f(a); } Produces the expected error: alias.C: In function ‘int main()’: alias.C:13:6: error: use of function ‘void f(R&) [with R = int [2]]’ with unsatisfied constraints alias.C:7:6: note: declared here alias.C:7:6: note: constraints not satisfied alias.C: In instantiation of ‘void f(R&) [with R = int [2]]’: alias.C:13:6: required from here alias.C:3:30: required for the satisfaction of ‘False<decltype (std::__detail::__ranges_begin(declval<_Container&>()))>’ [with _Container = int[2]] alias.C:3:38: note: the expression ‘false’ evaluated to ‘false’ However the second to last line refers to std::__detail::__ranges_begin(declval<_Container&>())) This is meaningless to users, and "_Container" is not used in the definitions of ranges::iterator_t or anything it uses. The compiler seems drunk and is remembering the template argument of a different, unrelated use of the same alias template. That's probably PR 95310. I think it would be better to leave that alias unexpanded, as std::ranges::iterator_t<R>.