This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[C++ PATCH] Don't ICE on vla = { }; (PR c++/42059)


Hi!

We reject
void foo (int i) { int a[i] = {}; }
so IMHO we should also reject
void foo (int i) { int a[i]; a = {}; }
instead of ICEing on it.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk/4.4?

2009-11-16  Jakub Jelinek  <jakub@redhat.com>

	PR c++/42059
	* typeck.c (cp_build_modify_expr): For initializer list call
	check_array_initializer to make sure lhs isn't a VLA.

	* g++.dg/cpp0x/initlist26.C: New test.

--- gcc/cp/typeck.c.jj	2009-11-13 11:13:20.000000000 +0100
+++ gcc/cp/typeck.c	2009-11-16 14:38:56.000000000 +0100
@@ -6239,7 +6239,11 @@ cp_build_modify_expr (tree lhs, enum tre
       int from_array;
 
       if (BRACE_ENCLOSED_INITIALIZER_P (rhs))
-	rhs = digest_init (lhstype, rhs);
+	{
+	  if (check_array_initializer (lhs, lhstype, rhs))
+	    return error_mark_node;
+	  rhs = digest_init (lhstype, rhs);
+	}
 
       else if (!same_or_base_type_p (TYPE_MAIN_VARIANT (lhstype),
 				     TYPE_MAIN_VARIANT (TREE_TYPE (rhs))))
--- gcc/testsuite/g++.dg/cpp0x/initlist26.C.jj	2009-11-16 14:43:55.000000000 +0100
+++ gcc/testsuite/g++.dg/cpp0x/initlist26.C	2009-11-16 14:44:05.000000000 +0100
@@ -0,0 +1,10 @@
+// PR c++/42059
+// { dg-do compile }
+// { dg-options "-std=gnu++0x" }
+
+void
+foo (int i)
+{
+  int a[i];
+  a = { }; // { dg-error "may not be initialized" }
+}

	Jakub


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]