Patchwork Fix ICE with C compound literals (PR c/48517)

login
register
mail settings
Submitter Jakub Jelinek
Date April 11, 2011, 7:52 p.m.
Message ID <20110411195238.GA17079@tyan-ft48-01.lab.bos.redhat.com>
Download mbox | patch
Permalink /patch/90641/
State New
Headers show

Comments

Jakub Jelinek - April 11, 2011, 7:52 p.m.
Hi!

On the following testcase build_unary_op ICEs, because
the element type of the array variable (which is TREE_READONLY)
unexpectedly is not TYPE_READONLY.
The problem seems to come from store_init_value, which replaces
the type of the variable with build_distinct_type_copy (TYPE_MAIN_VARIANT (type));
on which it sets TYPE_DOMAIN.  The following patch restores the
quals back.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk/4.6?

2011-04-11  Jakub Jelinek  <jakub@redhat.com>

	PR c/48517
	* c-typeck.c (store_init_value): Set TREE_TYPE (decl) to
	qualified type.

	* gcc.dg/pr48517.c: New test.


	Jakub
Joseph S. Myers - April 11, 2011, 9:29 p.m.
On Mon, 11 Apr 2011, Jakub Jelinek wrote:

> 2011-04-11  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR c/48517
> 	* c-typeck.c (store_init_value): Set TREE_TYPE (decl) to
> 	qualified type.
> 
> 	* gcc.dg/pr48517.c: New test.

OK, though I think the test would go better in gcc.c-torture/compile since 
it requires no special options and does not test for any diagnostics.
Jeff Law - April 12, 2011, 3:47 p.m.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 04/11/11 13:52, Jakub Jelinek wrote:
> Hi!
> 
> On the following testcase build_unary_op ICEs, because
> the element type of the array variable (which is TREE_READONLY)
> unexpectedly is not TYPE_READONLY.
> The problem seems to come from store_init_value, which replaces
> the type of the variable with build_distinct_type_copy (TYPE_MAIN_VARIANT (type));
> on which it sets TYPE_DOMAIN.  The following patch restores the
> quals back.
> 
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk/4.6?
> 
> 2011-04-11  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR c/48517
> 	* c-typeck.c (store_init_value): Set TREE_TYPE (decl) to
> 	qualified type.
> 
> 	* gcc.dg/pr48517.c: New test.
OK.
jeff
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJNpHQcAAoJEBRtltQi2kC7G00H/0YVn3lAw5tf3wDfTwBK8G1U
Jpy7iroBBgkBb/X+DSG1PRbWmxy7hsbsQvqsNI5iYVq/nbSxl7BFPxP1RxOqo07G
AbFtnLlgOuAG2XfhuN/oUc4SwiU23JyX8zBhYXZEHqho6oFDTIxUOMh4+LYM92ll
vPZoHprQZkS3/HsQu6BzZoEyf8mmUh5D2DMU6X3kW4Lb3d1+URdcq6xIzTXRMjga
79ZRyL/WozaTjdtw9dr1mIqhlLs2V6bOydItRl41xQTALkq52iqIYT/3+YvYuiLR
z4MGT4gGO4JmE6a9K+jVVxjjtMrgiJP20l9C2/UjWiqSyuaJAHcxMrakX46gB0Q=
=4Uv+
-----END PGP SIGNATURE-----

Patch

--- gcc/c-typeck.c.jj	2011-03-31 08:51:03.000000000 +0200
+++ gcc/c-typeck.c	2011-04-11 15:19:25.000000000 +0200
@@ -5773,11 +5773,13 @@  store_init_value (location_t init_loc, t
 	      /* For int foo[] = (int [3]){1}; we need to set array size
 		 now since later on array initializer will be just the
 		 brace enclosed list of the compound literal.  */
+	      tree etype = strip_array_types (TREE_TYPE (decl));
 	      type = build_distinct_type_copy (TYPE_MAIN_VARIANT (type));
-	      TREE_TYPE (decl) = type;
 	      TYPE_DOMAIN (type) = TYPE_DOMAIN (TREE_TYPE (cldecl));
 	      layout_type (type);
 	      layout_decl (cldecl, 0);
+	      TREE_TYPE (decl)
+		= c_build_qualified_type (type, TYPE_QUALS (etype));
 	    }
 	}
     }
--- gcc/testsuite/gcc.dg/pr48517.c.jj	2011-04-11 15:21:59.000000000 +0200
+++ gcc/testsuite/gcc.dg/pr48517.c	2011-04-11 15:21:16.000000000 +0200
@@ -0,0 +1,13 @@ 
+/* PR c/48517 */
+/* { dg-do compile } */
+/* { dg-options "" } */
+
+void bar (const unsigned short *);
+
+void
+foo (void)
+{
+  static const unsigned short array[] = (const unsigned short []) { 0x0D2B };
+  const unsigned short *ptr = array;
+  bar (ptr);
+}