C++ PATCH for c++/87080, ICE with -Wpessimizing-move

[Marek Polacek]

The problem in this testcase was that we were calling is_std_move_p from
template context, which breaks in cp_get_fndecl_from_callee.  This warning
is not meant to warn while parsing a template, so I think we should apply this.

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2018-08-24  Marek Polacek  <pola...@redhat.com>

        PR c++/87080
        * typeck.c (maybe_warn_pessimizing_move): Do nothing in a template.

        * g++.dg/cpp0x/Wpessimizing-move5.C: New test.

diff --git gcc/cp/typeck.c gcc/cp/typeck.c
index 122d9dcd4b3..24647e29a55 100644
--- gcc/cp/typeck.c
+++ gcc/cp/typeck.c
@@ -9192,6 +9192,11 @@ maybe_warn_pessimizing_move (tree retval, tree functype)
   if (cxx_dialect < cxx11)
     return;
 
+  /* Wait until instantiation time, since we can't gauge if we should do
+     the NRVO until then.  */
+  if (processing_template_decl)
+    return;
+
   /* This is only interesting for class types.  */
   if (!CLASS_TYPE_P (functype))
     return;
diff --git gcc/testsuite/g++.dg/cpp0x/Wpessimizing-move5.C 
gcc/testsuite/g++.dg/cpp0x/Wpessimizing-move5.C
index e69de29bb2d..02ad2113505 100644
--- gcc/testsuite/g++.dg/cpp0x/Wpessimizing-move5.C
+++ gcc/testsuite/g++.dg/cpp0x/Wpessimizing-move5.C
@@ -0,0 +1,14 @@
+// PR c++/87080
+// { dg-do compile { target c++11 } }
+// { dg-options "-Wpessimizing-move" }
+
+struct a {
+  template<typename b> a &operator<<(b);
+};
+a c();
+template<typename>
+a fn2()
+{
+  int d = 42;
+  return c() << d;
+}