On Wed, Dec 02, 2020 at 10:15:25PM -0500, Jason Merrill wrote:
> Jakub noticed that we weren't recognizing a default argument for a consteval
> member function as being in immediate function context because there was no
> function parameter scope to look at.
>
> Note that this patch doesn't actually push the parameters into the scope,
> that happens in a separate commit.
Shouldn't we also be testing how it behaves in templates?
The following testcase is an attempt to test both non-dependent and
dependent consteval calls in both function and class templates, and with
your committed patch it now passes.
Ok for trunk?
2020-12-03 Jakub Jelinek <ja...@redhat.com>
* g++.dg/cpp2a/consteval-defarg2.C: New test.
--- gcc/testsuite/g++.dg/cpp2a/consteval-defarg2.C.jj 2020-12-03
10:26:16.340256056 +0100
+++ gcc/testsuite/g++.dg/cpp2a/consteval-defarg2.C 2020-12-03
10:19:33.215853790 +0100
@@ -0,0 +1,29 @@
+// Test that late-parsed default args have the same consteval semantics.
+// { dg-do compile { target c++20 } }
+
+template <int N>
+consteval bool foo (bool x) { if (x) throw N; return false; }
+consteval bool qux (bool x) { if (x) throw 1; return false; }
+template <int N>
+consteval bool bar (bool x = foo<N> (true)) { return true; }
+template <int N>
+consteval bool corge (bool x = qux (true)) { return true; }
+template <int N>
+struct S
+{
+ consteval static bool baz (bool x = foo<N> (true)) { return true; }
+ consteval static bool garply (bool x = qux (true)) { return true; }
+};
+struct T
+{
+ template <int N>
+ consteval static bool baz (bool x = foo<N> (true)) { return true; }
+ template <int N>
+ consteval static bool garply (bool x = qux (true)) { return true; }
+};
+constexpr bool a = bar<0> (true);
+constexpr bool b = corge<0> (true);
+constexpr bool c = S<0>::baz (true);
+constexpr bool d = S<0>::garply (true);
+constexpr bool e = T::baz<0> (true);
+constexpr bool f = T::garply<0> (true);
Jakub
