Hi!

We can't call maybe_apply_pragma_weak for structured binding decls
at start_decl time, as that is too early, we don't have DECL_ASSEMBLER_NAME
computed yet for them and because we don't have the corresponding decls
around yet, we can't even compute it.

This patch defers the maybe_apply_pragma_weak call from start_decl to
cp_maybe_mangle_decomp where we just added the mangling the call needs.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2018-04-05  Jakub Jelinek  <ja...@redhat.com>

        PR c++/85208
        * decl.c (start_decl): For DECL_DECOMPOSITION_P decls, don't call
        maybe_apply_pragma_weak here...
        (cp_maybe_mangle_decomp): ... but call it here instead.

        * g++.dg/cpp1z/decomp42.C: New test.

--- gcc/cp/decl.c.jj    2018-04-04 21:33:20.532639396 +0200
+++ gcc/cp/decl.c       2018-04-05 13:29:39.128215719 +0200
@@ -5090,7 +5090,7 @@ start_decl (const cp_declarator *declara
     }
 
   /* If #pragma weak was used, mark the decl weak now.  */
-  if (!processing_template_decl)
+  if (!processing_template_decl && !DECL_DECOMPOSITION_P (decl))
     maybe_apply_pragma_weak (decl);
 
   if (TREE_CODE (decl) == FUNCTION_DECL
@@ -7483,6 +7483,7 @@ cp_maybe_mangle_decomp (tree decl, tree
       for (unsigned int i = 0; i < count; i++, d = DECL_CHAIN (d))
        v[count - i - 1] = d;
       SET_DECL_ASSEMBLER_NAME (decl, mangle_decomp (decl, v));
+      maybe_apply_pragma_weak (decl);
     }
 }
 
--- gcc/testsuite/g++.dg/cpp1z/decomp42.C.jj    2018-04-05 13:45:39.521220008 
+0200
+++ gcc/testsuite/g++.dg/cpp1z/decomp42.C       2018-04-05 13:45:22.507220268 
+0200
@@ -0,0 +1,9 @@
+// PR c++/85208
+// { dg-do compile { target c++11 } }
+// { dg-require-weak "" }
+// { dg-options "" }
+
+#pragma weak _ZDC1d1e1fE
+struct A { int i, j, k; };
+auto [a, b, c] = A (); // { dg-warning "structured bindings only available 
with" "" { target c++14_down } }
+auto [d, e, f] = A (); // { dg-warning "structured bindings only available 
with" "" { target c++14_down } }

        Jakub

Reply via email to