diff mbox

[Ada] Fix ICE on aggregate of multidimensional array type

Message ID 201007232153.44788.ebotcazou@adacore.com
State New
Headers show

Commit Message

Eric Botcazou July 23, 2010, 7:53 p.m. UTC 
> 2010-07-22  Eric Botcazou  <ebotcazou@adacore.com>
>
> 	* gcc-interface/utils.c (gnat_types_compatible_p): Don't require strict
> 	equality for the component type of array types.

Unfortunately this is now too lax and leads to other problems.  Hence this new 
fix, tested on x86-64-suse-linux, applied on the mainline.


2010-07-23  Eric Botcazou  <ebotcazou@adacore.com>

	* gcc-interface/utils.c (gnat_types_compatible_p): Revert latest change
	and recurse only for multidimensional array types instead.


2010-07-23  Eric Botcazou  <ebotcazou@adacore.com>

        * gnat.dg/aggr16.ad[sb]: New test.
	* gnat.dg/aggr16_pkg.ads: New helper.
diff mbox

Patch

Index: gcc-interface/utils.c
===================================================================
--- gcc-interface/utils.c	(revision 162425)
+++ gcc-interface/utils.c	(working copy)
@@ -2081,7 +2081,7 @@  gnat_types_compatible_p (tree t1, tree t
     return 1;
 
   /* Array types are also compatible if they are constrained and have the same
-     domain and compatible component types.  */
+     domain(s) and the same component type.  */
   if (code == ARRAY_TYPE
       && (TYPE_DOMAIN (t1) == TYPE_DOMAIN (t2)
 	  || (TYPE_DOMAIN (t1)
@@ -2090,7 +2090,9 @@  gnat_types_compatible_p (tree t1, tree t
 				     TYPE_MIN_VALUE (TYPE_DOMAIN (t2)))
 	      && tree_int_cst_equal (TYPE_MAX_VALUE (TYPE_DOMAIN (t1)),
 				     TYPE_MAX_VALUE (TYPE_DOMAIN (t2)))))
-      && gnat_types_compatible_p (TREE_TYPE (t1), TREE_TYPE (t2)))
+      && (TREE_TYPE (t1) == TREE_TYPE (t2)
+	  || (TREE_CODE (TREE_TYPE (t1)) == ARRAY_TYPE
+	      && gnat_types_compatible_p (TREE_TYPE (t1), TREE_TYPE (t2)))))
     return 1;
 
   /* Padding record types are also compatible if they pad the same