* x86_64 argument passing fix
@ 2002-08-11 5:43 Jan Hubicka
2002-08-14 2:46 ` Gwenole Beauchesne
0 siblings, 1 reply; 2+ messages in thread
From: Jan Hubicka @ 2002-08-11 5:43 UTC (permalink / raw)
To: gcc-patches, rth
Hi,
the offset of nested structure fields is computed wrong. I am
installing the patch to mainline and queueing for 3.2.1. It affects only
the x86_64 specific code but has been bootstrapped on i386.
Sat Aug 10 19:56:31 CEST 2002 Jan Hubicka <jh@suse.cz>
* i386.c (classify_argument): Fix computing of field's offsets.
*** i386.c.old Sat Aug 10 18:58:35 2002
--- i386.c Sat Aug 10 18:58:38 2002
*************** classify_argument (mode, type, classes,
*** 1634,1640 ****
return 0;
for (i = 0; i < num; i++)
{
! int pos = (offset + bit_offset) / 8 / 8;
classes[i + pos] =
merge_classes (subclasses[i], classes[i + pos]);
}
--- 1634,1640 ----
return 0;
for (i = 0; i < num; i++)
{
! int pos = (offset + (bit_offset % 64)) / 8 / 8;
classes[i + pos] =
merge_classes (subclasses[i], classes[i + pos]);
}
*************** classify_argument (mode, type, classes,
*** 1671,1677 ****
for (i = 0; i < num; i++)
{
int pos =
! (int_bit_position (field) + bit_offset) / 8 / 8;
classes[i + pos] =
merge_classes (subclasses[i], classes[i + pos]);
}
--- 1671,1677 ----
for (i = 0; i < num; i++)
{
int pos =
! (int_bit_position (field) + (bit_offset % 64)) / 8 / 8;
classes[i + pos] =
merge_classes (subclasses[i], classes[i + pos]);
}
*************** classify_argument (mode, type, classes,
*** 1717,1723 ****
num = classify_argument (TYPE_MODE (type),
type, subclasses,
! (offset + bit_offset) % 256);
if (!num)
return 0;
for (i = 0; i < num; i++)
--- 1717,1723 ----
num = classify_argument (TYPE_MODE (type),
type, subclasses,
! (offset + (bit_offset % 64)) % 256);
if (!num)
return 0;
for (i = 0; i < num; i++)
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: x86_64 argument passing fix
2002-08-11 5:43 x86_64 argument passing fix Jan Hubicka
@ 2002-08-14 2:46 ` Gwenole Beauchesne
0 siblings, 0 replies; 2+ messages in thread
From: Gwenole Beauchesne @ 2002-08-14 2:46 UTC (permalink / raw)
To: Jan Hubicka; +Cc: gcc-patches, rth
On Sun, 11 Aug 2002, Jan Hubicka wrote:
> the offset of nested structure fields is computed wrong. I am
> installing the patch to mainline and queueing for 3.2.1. It affects only
> the x86_64 specific code but has been bootstrapped on i386.
As a reminder, both cases of PR target/7559 are hereunder:
2002-08-10 Gwenole Beauchesne <gbeauchesne@mandrakesoft.com>
PR target/7559
* testsuite/gcc.c-torture/execute/20020810-1.c: New test.
--- gcc-3.2/gcc/testsuite/gcc.c-torture/execute/20020810-1.c.x86_64-struct-args 2002-08-13 10:59:50.000000000 +0200
+++ gcc-3.2/gcc/testsuite/gcc.c-torture/execute/20020810-1.c 2002-08-13 11:03:23.000000000 +0200
@@ -0,0 +1,38 @@
+/* PR target/7559
+ This testcase was miscompiled on x86-64, because classify_argument
+ wrongly computed the offset of nested structure fields. */
+
+extern void abort (void);
+
+struct A
+{
+ long x;
+};
+
+struct R
+{
+ struct A a, b;
+};
+
+struct R R = { 100, 200 };
+
+void f (struct R r)
+{
+ if (r.a.x != R.a.x || r.b.x != R.b.x)
+ abort ();
+}
+
+struct R g (void)
+{
+ return R;
+}
+
+int main (void)
+{
+ struct R r;
+ f(R);
+ r = g();
+ if (r.a.x != R.a.x || r.b.x != R.b.x)
+ abort ();
+ return 0;
+}
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2002-08-14 9:46 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-08-11 5:43 x86_64 argument passing fix Jan Hubicka
2002-08-14 2:46 ` Gwenole Beauchesne
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).