[Bug c/58540] New: Incorrect warning message for '*=' statement and different results based on optimization

[bernardwidynski at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58540

            Bug ID: 58540
           Summary: Incorrect warning message for '*=' statement and
                    different results based on optimization
           Product: gcc
           Version: 4.7.3
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: bernardwidynski at gmail dot com

Created attachment 30900
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30900&action=edit
Example program which generates incorrect message

The code shown below generates a warning message

c_ex.c: In function ‘main’:
c_ex.c:15:8: warning: operation on ‘u.y’ may be undefined [-Wsequence-point]
c_ex.c:15:8: warning: ‘u.y’ is used uninitialized in this function
[-Wuninitialized]

Also, when compiled with optimization -O3, it prints the correct result.

When compiled without optimization it prints a zero.

Obviously, u.y is initialized.  It is set to x + r.

The compiler should not generate a warning message and should
produce the same results regardless of the optimization level.

-------------------------------------------------------------------------

#include <stdio.h>

static double r = 0.;
static double x = 2.7182818284590452353602;

int main (void) {

   union {
      double y;
      unsigned int w[2];
   } u;

   /* Set y = (x + r) squared */

   u.y *= u.y = x + r;

   printf("y = %lf\n",u.y);

   return 0;

}