[Bug fortran/51218] New: [4.7 Regression] Potential optimization bug

[anlauf at gmx dot de]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51218

             Bug #: 51218
           Summary: [4.7 Regression] Potential optimization bug
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: anlauf@gmx.de


Hi all,

the attached code crashes at runtime with gfortran 4.7
(svn rev. 181390) when compiled with optimization (-O1 and higher).

It works when compiled with -O0, with gfortran 4.6 and earlier
at any optimization level tested, and a series of other compilers.

Can anybody shed some light on this problem?

Depending on the active line in the main program,

!####################################
! Either expression leads to a crash:
  u  = sum(xi*wi) + sum(xi*wi)
! u  = sum(xi*wi  +     xi*wi)
!####################################

I get a backtrace or a

*** glibc detected *** ./a.out: double free or corruption (fasttop): 0x080743a0
*

Activating the following print

!#######################################
!print *, "vector_times_vector: barrier"
!#######################################

in module mo_dec_matrix makes the crash disappear.

Thanks,
Harald