[Bug fortran/35820] internal compiler error with forall

["dick dot hendrickson at gmail dot com" <gcc-bugzilla@gcc.gnu.org>]

------- Comment #2 from dick dot hendrickson at gmail dot com  2008-04-03 22:12 -------
Subject: Re:  internal compiler error with forall

On Thu, Apr 3, 2008 at 3:30 PM, dominiq at lps dot ens dot fr
<gcc-bugzilla@gcc.gnu.org> wrote:
>
>
>  ------- Comment #1 from dominiq at lps dot ens dot fr  2008-04-03 20:30 -------
>  The code compiles fine on (powerpc|i686)-apple-darwin9. Would it be possible to
>  check how the executable works?

Not easily, from my point of view.  This is an isolated part of a
large test suite.  I
spent a couple of half days getting it down to this level.  The
problem is that the
ICE moves around if I do anything.  I've deleted about 160 of the original set
of tests for this isolation.  If I leave the 160 in and delete only 2 of the 3
subroutines I sent in the results are

With SA0136 "not removed"

gfortran%try_gfortranr s_sa0
working\s_sa0_mods.f
working\s_sa0_tests.f
working\s_sa0_exts.f
        1 file(s) copied.
rem.f: In function 'sa0160':
rem.f:5327: internal compiler error: Segmentation fault
[snip]

With SA0137 "not removed"

gfortran%try_gfortranr s_sa0
working\s_sa0_mods.f
working\s_sa0_tests.f
working\s_sa0_exts.f
        1 file(s) copied.
rem.f: In function 'sa0131':
rem.f:4412: internal compiler error: Segmentation fault
[snip]

With SA0138 "not removed"

gfortran%try_gfortranr s_sa0
working\s_sa0_mods.f
working\s_sa0_tests.f
working\s_sa0_exts.f
        1 file(s) copied.
rem.f: In function 'sa0128':
rem.f:4325: internal compiler error: Segmentation fault
[snip]


That is, the ICE appears in 3 different places, depending on which "bad"
subroutine I leave in the test set.  If I delete the 3 bad routines,
the other 160
all work correctly.

I'm reluctant to send in the whole set for two reasons.  It's large, the 160
subroutines are 190000 characters or so and they need a substantial
amount of structure code to work.  Also, the suite is my bread-and-butter
product and I'm reluctant to put part of it in the public domain.

If no one can reproduce the problem on a Windows machine, I can try to
make an executable that fails.

Dick Hendrickson
>
>
>  --
>
>
>  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35820
>
>  ------- You are receiving this mail because: -------
>  You reported the bug, or are watching the reporter.
>


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35820