[Bug fortran/60341] ICE compiling Nonmem 6.2.0

2014-02-27 Thread schapel at umich dot edu 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60341

--- Comment #8 from Steve Chapel  ---
Yes, it's a regression from earlier versions of gfortran. I have verified that
with the workaround, Nonmem 6.2.0 compiles and runs properly with optimization
using gfortran 4.8.1 and 4.8.2. Turning optimization off also works.

[Bug fortran/60341] ICE compiling Nonmem 6.2.0

2014-02-26 Thread sgk at troutmask dot apl.washington.edu 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60341

--- Comment #7 from Steve Kargl  ---
On Wed, Feb 26, 2014 at 08:38:57PM +, mikael at gcc dot gnu.org wrote:
>
> Looks like an unguarded union access.
> This is a regression from the time there was now front-end optimization I
> guess?

Unfortunately, I've deleted versions of gfortran older than
4.6.x.  The code fails with all version above 4.6.x.  If
one adds -fno-frontend-optimize, then the code compiles.
So, yes, it is a regression.

[Bug fortran/60341] ICE compiling Nonmem 6.2.0

2014-02-26 Thread mikael at gcc dot gnu.org 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60341

Mikael Morin  changed:

   What|Removed |Added

 CC||mikael at gcc dot gnu.org

--- Comment #6 from Mikael Morin  ---
Looks like an unguarded union access.
This is a regression from the time there was now front-end optimization I
guess?

diff --git a/gcc/fortran/frontend-passes.c b/gcc/fortran/frontend-passes.c
index 52bd700..e663868 100644
--- a/gcc/fortran/frontend-passes.c
+++ b/gcc/fortran/frontend-passes.c
@@ -1391,7 +1391,9 @@ optimize_comparison (gfc_expr *e, gfc_intrinsic_op op)
   /* Replace A // B < A // C with B < C, and A // B < C // B
  with A < C.  */
   if (op1->ts.type == BT_CHARACTER && op2->ts.type == BT_CHARACTER
+  && op1->expr_type == EXPR_OP
   && op1->value.op.op == INTRINSIC_CONCAT
+  && op2->expr_type == EXPR_OP
   && op2->value.op.op == INTRINSIC_CONCAT)
 {
   gfc_expr *op1_left = op1->value.op.op1;

[Bug fortran/60341] ICE compiling Nonmem 6.2.0

2014-02-26 Thread sgk at troutmask dot apl.washington.edu 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60341

--- Comment #5 from Steve Kargl  ---
On Wed, Feb 26, 2014 at 07:15:49PM +, kargl at gcc dot gnu.org wrote:

The workaround should be obvious, but just encase

> Reduced testcase 
> 
>   subroutine modelg(ncm)
>   implicit none
>   integer, parameter :: pc = 30, pm = pc - 1
>   integer i
>   character*4 catt(pm,2)

character(len=7) stmp

>   integer ncm,iatt(pm,pc)
>   do i=1,ncm

   stmp = catt(i,1)//catt(i,2)
   if (stmp .eq. 'central') exit

>  if (catt(i,1)//catt(i,2).eq.'central') exit
>   end do
>   iatt(i,4)=1
>   end

A similar temporary should be applicable to the original code.

[Bug fortran/60341] ICE compiling Nonmem 6.2.0

2014-02-26 Thread kargl at gcc dot gnu.org 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60341

kargl at gcc dot gnu.org changed:

   What|Removed |Added

 CC||kargl at gcc dot gnu.org

--- Comment #4 from kargl at gcc dot gnu.org ---
Reduced testcase 

  subroutine modelg(ncm)
  implicit none
  integer, parameter :: pc = 30, pm = pc - 1
  integer i
  character*4 catt(pm,2)
  integer ncm,iatt(pm,pc)
  do i=1,ncm
 if (catt(i,1)//catt(i,2).eq.'central') exit
  end do
  iatt(i,4)=1
  end

[Bug fortran/60341] ICE compiling Nonmem 6.2.0

2014-02-26 Thread rguenth at gcc dot gnu.org 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60341

Richard Biener  changed:

   What|Removed |Added

 Status|UNCONFIRMED |NEW
   Last reconfirmed||2014-02-26
 Ever confirmed|0   |1
  Known to fail||4.7.3, 4.8.2, 4.9.0

--- Comment #3 from Richard Biener  ---
Confirmed.

f951: internal compiler error: Segmentation fault
0xc5fa7a crash_signal
/space/rguenther/src/svn/trunk/gcc/toplev.c:337
0x6ec628 gfc_dep_compare_expr(gfc_expr*, gfc_expr*)
/space/rguenther/src/svn/trunk/gcc/fortran/dependency.c:345
0x79503e optimize_comparison
/space/rguenther/src/svn/trunk/gcc/fortran/frontend-passes.c:1402
0x794af2 optimize_op
/space/rguenther/src/svn/trunk/gcc/fortran/frontend-passes.c:1222
0x7929fd optimize_expr
/space/rguenther/src/svn/trunk/gcc/fortran/frontend-passes.c:170
0x795b8c gfc_expr_walker(gfc_expr**, int (*)(gfc_expr**, int*, void*), void*)
/space/rguenther/src/svn/trunk/gcc/fortran/frontend-passes.c:1787
0x797866 gfc_code_walker(gfc_code**, int (*)(gfc_code**, int*, void*), int
(*)(gfc_expr**, int*, void*), void*)
/space/rguenther/src/svn/trunk/gcc/fortran/frontend-passes.c:2134

[Bug fortran/60341] ICE compiling Nonmem 6.2.0

2014-02-25 Thread kargl at gcc dot gnu.org 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60341

kargl at gcc dot gnu.org changed:

   What|Removed |Added

   Severity|critical|normal

[Bug fortran/60341] ICE compiling Nonmem 6.2.0

2014-02-25 Thread steve.chapel at a2pg dot com 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60341

--- Comment #1 from Steve Chapel  ---
Created attachment 32215
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32215&action=edit
source code file

[Bug fortran/60341] ICE compiling Nonmem 6.2.0

2014-02-25 Thread steve.chapel at a2pg dot com 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60341

--- Comment #2 from Steve Chapel  ---
Created attachment 32216
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32216&action=edit
included source code file