On 7/20/20 4:18 PM, Patrick Palka wrote:
In the below testcase, duplicate_decls wasn't merging the tsubsted
friend declaration for 'void add(auto)' with its definition, because
reduce_template_parm_level (during tsubst_friend_function) lost the
DECL_VIRTUAL_P flag on the invented 'auto' template parameter, which
made template_heads_equivalent_p deem the two template heads as not
equivalent in C++20 mode.

This patch makes reduce_template_parm_level carry over the
DECL_VIRTUAL_P from the original TEMPLATE_PARM_DECL.

Passes 'make check-c++' and the cmcstl2 testsuite.  Does this look OK
for trunk after a full bootstrap and regtest?

OK.

gcc/cp/ChangeLog:

        PR c++/96106
        * pt.c (reduce_template_parm_level): Carry over DECL_VIRTUAL_P
        from the original TEMPLATE_PARM_DECL.

gcc/testsuite/ChangeLog:

        PR c++/96106
        * g++.dg/concepts/abbrev7.C: New test.
---
  gcc/cp/pt.c                             |  1 +
  gcc/testsuite/g++.dg/concepts/abbrev7.C | 14 ++++++++++++++
  2 files changed, 15 insertions(+)
  create mode 100644 gcc/testsuite/g++.dg/concepts/abbrev7.C

diff --git a/gcc/cp/pt.c b/gcc/cp/pt.c
index defc2a9abd8..3f7b89141b6 100644
--- a/gcc/cp/pt.c
+++ b/gcc/cp/pt.c
@@ -4453,6 +4453,7 @@ reduce_template_parm_level (tree index, tree type, int 
levels, tree args,
                              type);
        TREE_CONSTANT (decl) = TREE_CONSTANT (orig_decl);
        TREE_READONLY (decl) = TREE_READONLY (orig_decl);
+      DECL_VIRTUAL_P (decl) = DECL_VIRTUAL_P (orig_decl);
        DECL_ARTIFICIAL (decl) = 1;
        SET_DECL_TEMPLATE_PARM_P (decl);
diff --git a/gcc/testsuite/g++.dg/concepts/abbrev7.C b/gcc/testsuite/g++.dg/concepts/abbrev7.C
new file mode 100644
index 00000000000..443c1b7871b
--- /dev/null
+++ b/gcc/testsuite/g++.dg/concepts/abbrev7.C
@@ -0,0 +1,14 @@
+// PR c++/96106
+// { dg-do compile { target concepts } }
+
+template<typename>
+struct number {
+  friend void add(auto);
+};
+
+void add(auto) { }
+
+void foo() {
+  number<int> n;
+  add(n); // { dg-bogus "ambiguous" }
+}


Reply via email to