[PATCH] Fix shorten_compare (PR c/48197)
Jakub Jelinek
jakub@redhat.com
Mon Mar 21 17:59:00 GMT 2011
On Mon, Mar 21, 2011 at 04:31:31PM +0000, Joseph S. Myers wrote:
> On Sun, 20 Mar 2011, Jakub Jelinek wrote:
>
> > 2011-03-20 Jakub Jelinek <jakub@redhat.com>
> >
> > PR c/48197
> > * c-common.c (shorten_compare): If primopN is first sign-extended
> > to opN and then zero-extended to result type, set primopN to opN.
> >
> > * gcc.c-torture/execute/pr48197.c: New test.
>
> This is OK, in that I believe it is (a) correct and (b) a sensible place
> in the current context to fix this to avoid problems with shorten_compare
> not working with mixed extensions. It's worth checking if this fixes PR
> 42544 as well, since that also looks like it involves mixed extensions.
It does, so here is what I've actually committed:
2011-03-20 Jakub Jelinek <jakub@redhat.com>
PR c/42544
PR c/48197
* c-common.c (shorten_compare): If primopN is first sign-extended
to opN and then zero-extended to result type, set primopN to opN.
* gcc.c-torture/execute/pr42544.c: New test.
* gcc.c-torture/execute/pr48197.c: New test.
--- gcc/c-family/c-common.c.jj 2011-03-21 13:00:03.000000000 +0100
+++ gcc/c-family/c-common.c 2011-03-21 18:29:23.000000000 +0100
@@ -3301,6 +3301,20 @@ shorten_compare (tree *op0_ptr, tree *op
primop0 = get_narrower (op0, &unsignedp0);
primop1 = get_narrower (op1, &unsignedp1);
+ /* If primopN is first sign-extended from primopN's precision to opN's
+ precision, then zero-extended from opN's precision to
+ *restype_ptr precision, shortenings might be invalid. */
+ if (TYPE_PRECISION (TREE_TYPE (primop0)) < TYPE_PRECISION (TREE_TYPE (op0))
+ && TYPE_PRECISION (TREE_TYPE (op0)) < TYPE_PRECISION (*restype_ptr)
+ && !unsignedp0
+ && TYPE_UNSIGNED (TREE_TYPE (op0)))
+ primop0 = op0;
+ if (TYPE_PRECISION (TREE_TYPE (primop1)) < TYPE_PRECISION (TREE_TYPE (op1))
+ && TYPE_PRECISION (TREE_TYPE (op1)) < TYPE_PRECISION (*restype_ptr)
+ && !unsignedp1
+ && TYPE_UNSIGNED (TREE_TYPE (op1)))
+ primop1 = op1;
+
/* Handle the case that OP0 does not *contain* a conversion
but it *requires* conversion to FINAL_TYPE. */
--- gcc/testsuite/gcc.c-torture/execute/pr42544.c.jj 2011-03-21 18:25:01.000000000 +0100
+++ gcc/testsuite/gcc.c-torture/execute/pr42544.c 2011-03-21 18:24:29.000000000 +0100
@@ -0,0 +1,14 @@
+/* PR c/42544 */
+
+extern void abort (void);
+
+int
+main ()
+{
+ signed short s = -1;
+ if (sizeof (long long) == sizeof (unsigned int))
+ return 0;
+ if ((unsigned int) s >= 0x100000000ULL)
+ abort ();
+ return 0;
+}
--- gcc/testsuite/gcc.c-torture/execute/pr48197.c.jj 2011-03-21 18:29:23.000000000 +0100
+++ gcc/testsuite/gcc.c-torture/execute/pr48197.c 2011-03-21 18:29:23.000000000 +0100
@@ -0,0 +1,25 @@
+/* PR c/48197 */
+
+extern void abort (void);
+static int y = 0x8000;
+
+int
+main ()
+{
+ unsigned int x = (short)y;
+ if (sizeof (0LL) == sizeof (0U))
+ return 0;
+ if (0LL > (0U ^ (short)-0x8000))
+ abort ();
+ if (0LL > (0U ^ x))
+ abort ();
+ if (0LL > (0U ^ (short)y))
+ abort ();
+ if ((0U ^ (short)-0x8000) < 0LL)
+ abort ();
+ if ((0U ^ x) < 0LL)
+ abort ();
+ if ((0U ^ (short)y) < 0LL)
+ abort ();
+ return 0;
+}
Jakub
More information about the Gcc-patches
mailing list