[Bug c/51562] New: Expression evaluation with commas seems incorrect in gcc 4.5.2, 4.4.4

[Bug c/51562] New: Expression evaluation with commas seems incorrect in gcc 4.5.2, 4.4.4

[willus0 at hotmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51562

             Bug #: 51562
           Summary: Expression evaluation with commas seems incorrect in
                    gcc 4.5.2, 4.4.4
    Classification: Unclassified
           Product: gcc
           Version: 4.5.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: willus0@hotmail.com


I apologize if this has been reported or isn't even a bug, but I wasn't even
sure what to search for.  The code below doesn't evaluate the way you think it
might in gcc 4.5.2 (and gcc 4.4.4).  I'm not even sure how it should
legally/correctly evaluate, but there is a powell() function in Numerical
Recipes in C (1st and 2nd editions) that uses code like below and expects the
answer to the second line to be 5 (i.e. for the x*x expression to be evaluated
with the two different values of x and summed correctly).  That's how I
discovered this issue.  gcc-compiled code in gcc 4.5.2/4.4.4 reports this:

1+2=4
1^2+2^2=8

I tried different levels of optimization, on both Linux and Windows (MinGW),
all with the same result.  Tiny CC-compiled code reports this:

1+2=4
1^2+2^2=5

Those are the only compilers I've tried.  Here's the program:

#include <stdio.h>
void main(void)
    {
    int x;
    printf("1+2=%d\n",(x=1,x)+(x=2,x));
    printf("1^2+2^2=%d\n",(x=1,x*x)+(x=2,x*x));
    }

Thoughts??

[Bug c/51562] Expression evaluation with commas seems incorrect in gcc 4.5.2, 4.4.4

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51562

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> 2011-12-15 09:57:32 UTC ---
(x=1,x)+(x=2,x)

is invoking undefined behavior as there is no sequence point inbetween
the two assignments to x.

[Bug c/51562] Expression evaluation with commas seems incorrect in gcc 4.5.2, 4.4.4

[willus0 at hotmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51562

--- Comment #2 from willus0 at hotmail dot com 2011-12-15 13:16:37 UTC ---
(In reply to comment #1)
> (x=1,x)+(x=2,x)
> 
> is invoking undefined behavior as there is no sequence point inbetween
> the two assignments to x.

Is the same also true of the second statement, in which case the Numerical
Recipes code is simply poorly written?

[Bug c/51562] missing -Wsequence-point warning for expression with commas.

[manu at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51562

Manuel López-Ibáñez <manu at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |NEW
            Version|4.5.2                       |4.7.0
           Keywords|                            |diagnostic
   Last reconfirmed|                            |2011-12-15
                 CC|                            |manu at gcc dot gnu.org
         Resolution|INVALID                     |
     Ever Confirmed|0                           |1
            Summary|Expression evaluation with  |missing -Wsequence-point
                   |commas seems incorrect in   |warning for expression with
                   |gcc 4.5.2, 4.4.4            |commas.

--- Comment #3 from Manuel López-Ibáñez <manu at gcc dot gnu.org> 2011-12-15 14:45:46 UTC ---
Not sure why -Wsequence-point is missing this, but it is a bug.

Clang 3.1 does not warn either.

[Bug c/51562] missing -Wsequence-point warning for expression with commas.

[willus0 at hotmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51562

--- Comment #4 from willus0 at hotmail dot com 2011-12-16 14:01:31 UTC ---
Thanks Richard and Manuel for prompting me to learn more about sequence points.
 After 25 years of C programming, I've learned something new.  A couple of
links I found helpful:

http://www.angelikalanger.com/Articles/VSJ/SequencePoints/SequencePoints.html

http://en.wikipedia.org/wiki/Sequence_point

[Bug c/51562] missing -Wsequence-point warning for expression with commas.

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51562

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |mpolacek at gcc dot gnu.org
         Resolution|---                         |DUPLICATE

--- Comment #5 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
PR65430 has a better testcase + explanation why that happens, so closing this
one as a dup.

*** This bug has been marked as a duplicate of bug 65430 ***