[Bug c/50142] New: There is bug when swap elements of an array via chain expression.

[Bug c/50142] New: There is bug when swap elements of an array via chain expression.

[thomas.c.zhao at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50142

             Bug #: 50142
           Summary: There is bug when swap elements of an array via chain
                    expression.
    Classification: Unclassified
           Product: gcc
           Version: 4.0.3
            Status: UNCONFIRMED
          Severity: critical
          Priority: P3
         Component: c
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: thomas.c.zhao@gmail.com


[gcc version]: 4.0.3
[system type]: sparc-sun-solaris2.10
[build options]: none
[source code]:
main()
{
 int a[] = {1,2};
 a[0] ^= a[1] ^= a[0] ^= a[1];
 printf("a[0]=%d a[1]=%d\n", a[0], a[1]);
}

[expected result]: a[0]=2 a[1]=1
[actual result]: a[0]=0 a[1]=1

btw: gcc3 is ok.

[Bug c/50142] There is bug when swap elements of an array via chain expression.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50142

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-08-21 11:54:44 UTC ---
Evaluation order is undefined since there is no sequence point involved.

[Bug c/50142] There is bug when swap elements of an array via chain expression.

[thomas.c.zhao at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50142

--- Comment #2 from ZHAO Xiaogang <thomas.c.zhao at gmail dot com> 2011-08-22 00:36:34 UTC ---
(In reply to comment #1)
> Evaluation order is undefined since there is no sequence point involved.

I don't think so.
The associativity of operator ^= is Right to Left.

I have build the code with other compiler.
gcc 3.x is OK.
Sun cc is OK.
Microsoft VC is OK.

If there isn't any array, it's also OK when use gcc4 build.
main()
{
 int a = 1, b = 2;
 a ^= b ^= a ^= b;
 printf("a=%d b=%d\n", a, b);
}

[expected result]: a=2 b=1
[actual result]: a=2 b=1

So, I guess, the defect associated with the gcc4 new feature which optimized
array access.

[Bug c/50142] There is bug when swap elements of an array via chain expression.

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50142

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-08-22 06:06:29 UTC ---
You are wrong, please read the standard, in particular ISO C99 5.1.2.3 and
Annex C.

[Bug c/50142] There is bug when swap elements of an array via chain expression.

[thomas.c.zhao at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50142

ZHAO Xiaogang <thomas.c.zhao at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|critical                    |trivial

--- Comment #4 from ZHAO Xiaogang <thomas.c.zhao at gmail dot com> 2011-08-22 07:59:45 UTC ---
(In reply to comment #3)
> You are wrong, please read the standard, in particular ISO C99 5.1.2.3 and
> Annex C.
I understand.
Thanks.