Hi,
seems just one of these straightforward cases where we don't propagate
the complain argument thoroughly enough. Tested x86_64-linux.
Ok for mainline?
Thanks,
Paolo.
/////////////
/cp
2011-09-08 Paolo Carlini <paolo.carl...@oracle.com>
PR c++/50324
* typeck2.c (digest_init_r): Call complete_type_or_maybe_complain
instead of complete_type_or_else.
/testsuite
2011-09-08 Paolo Carlini <paolo.carl...@oracle.com>
PR c++/50324
* g++.dg/cpp0x/sfinae28.C: New.
Index: testsuite/g++.dg/cpp0x/sfinae28.C
===================================================================
--- testsuite/g++.dg/cpp0x/sfinae28.C (revision 0)
+++ testsuite/g++.dg/cpp0x/sfinae28.C (revision 0)
@@ -0,0 +1,16 @@
+// PR c++/50324
+// { dg-options -std=c++0x }
+
+struct complete { };
+struct incomplete;
+
+template<class T> auto f(T *) -> decltype(T{}) *;
+template<class T> char f(T);
+
+int main()
+{
+ complete *p = 0;
+ static_assert(sizeof(f(p)) == sizeof(void*), "");
+ incomplete *q = 0;
+ static_assert(sizeof(f(q)) == 1u, "");
+}
Index: cp/typeck2.c
===================================================================
--- cp/typeck2.c (revision 178685)
+++ cp/typeck2.c (working copy)
@@ -1,7 +1,7 @@
/* Report error messages, build initializers, and perform
some front-end optimizations for C++ compiler.
Copyright (C) 1987, 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
- 1999, 2000, 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2009, 2010
+ 1999, 2000, 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
Free Software Foundation, Inc.
Hacked by Michael Tiemann (tiem...@cygnus.com)
@@ -812,8 +812,9 @@ digest_init_r (tree type, tree init, bool nested,
/* We must strip the outermost array type when completing the type,
because the its bounds might be incomplete at the moment. */
- if (!complete_type_or_else (TREE_CODE (type) == ARRAY_TYPE
- ? TREE_TYPE (type) : type, NULL_TREE))
+ if (!complete_type_or_maybe_complain (TREE_CODE (type) == ARRAY_TYPE
+ ? TREE_TYPE (type) : type, NULL_TREE,
+ complain))
return error_mark_node;
/* Strip NON_LVALUE_EXPRs since we aren't using as an lvalue