This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH] don't complain too small an array component size if overflow
- From: Olivier Hainque <hainque at adacore dot com>
- To: gcc-patches at gcc dot gnu dot org
- Cc: hainque at adacore dot com
- Date: Sat, 22 Oct 2005 15:14:09 +0200
- Subject: [PATCH] don't complain too small an array component size if overflow
A recent check has been introduced in layout_type for array types, possibly
triggering
error ("alignment of array elements is greater than element size");
This error shows up spuriously on the following Ada testcase, expected
to compile silently:
procedure P is
package Matrixes is
type Matrix_T is array (Integer, Integer) of Integer;
end Matrixes;
begin
null;
end;
This test builds an array[integer] of (array[integer] of integer)
type, without allocating any object of this type (which would raise
storage_error at run-time).
The inner size overflows and the error triggers because the code
doesn't check if TREE_CONSTANT_OVERFLOW is set on the component size
before comparing the value to the alignment.
The patch below addresses that by adding such a check.
Bootstrapped and reg tested on i686-pc-linux-gnu, configured with
--enable-languages=all,ada
Thanks in advance,
Olivier
2005-10-22 Olivier Hainque <hainque@adacore.com>
* stor-layout.c (layout_type): Don't complain about a too small
an array element type size compared to its alignement when the
size overflows.
--- stor-layout.c.ori 2005-10-18 09:57:15.000000000 +0200
+++ stor-layout.c 2005-10-21 16:10:52.000000000 +0200
@@ -1827,10 +1827,11 @@ layout_type (tree type)
TYPE_MODE (type) = BLKmode;
}
}
if (TYPE_SIZE_UNIT (element)
&& TREE_CODE (TYPE_SIZE_UNIT (element)) == INTEGER_CST
+ && !TREE_CONSTANT_OVERFLOW (TYPE_SIZE_UNIT (element))
&& !integer_zerop (TYPE_SIZE_UNIT (element))
&& compare_tree_int (TYPE_SIZE_UNIT (element),
TYPE_ALIGN_UNIT (element)) < 0)
error ("alignment of array elements is greater than element size");
break;