]> git.ipfire.org Git - thirdparty/gcc.git/commit
c++: Fix up maybe_warn_for_constant_evaluated calls [PR114580]
authorJakub Jelinek <jakub@redhat.com>
Tue, 9 Apr 2024 07:31:42 +0000 (09:31 +0200)
committerJakub Jelinek <jakub@redhat.com>
Sun, 21 Apr 2024 04:08:05 +0000 (06:08 +0200)
commitae3b6dea0445f9650cf1a684527efac06497f1b4
tree710be539e87f36ff962a6011c915784eaa581d0e
parent38af0d59043da4cc07cd62c17da599e43668e3be
c++: Fix up maybe_warn_for_constant_evaluated calls [PR114580]

When looking at maybe_warn_for_constant_evaluated for the trivial
infinite loops patch, I've noticed that it can emit weird diagnostics
for if constexpr in templates, first warn that std::is_constant_evaluted()
always evaluates to false (because the function template is not constexpr)
and then during instantiation warn that std::is_constant_evaluted()
always evaluates to true (because it is used in if constexpr condition).
Now, only the latter is actually true, even when the if constexpr
is in a non-constexpr function, it will still always evaluate to true.

So, the following patch fixes it to call maybe_warn_for_constant_evaluated
always with IF_STMT_CONSTEXPR_P (if_stmt) as the second argument rather than
true if it is if constexpr with non-dependent condition etc.

2024-04-09  Jakub Jelinek  <jakub@redhat.com>

PR c++/114580
* semantics.cc (finish_if_stmt_cond): Call
maybe_warn_for_constant_evaluated with IF_STMT_CONSTEXPR_P (if_stmt)
as the second argument, rather than true/false depending on if
it is if constexpr with non-dependent constant expression with
bool type.

* g++.dg/cpp2a/is-constant-evaluated15.C: New test.

(cherry picked from commit cfed80b9e4f562c99679739548df9369117dd791)
gcc/cp/semantics.cc
gcc/testsuite/g++.dg/cpp2a/is-constant-evaluated15.C [new file with mode: 0644]