On 2010-04-19, at 10:51 AM, Timothy J Giese wrote:
> On 04/19/2010 12:56 AM, Andreas Svrcek-Seiler wrote:
>>> http://ambermd.org/downloads/AmberTools.18apr10.tar.bz2
>>> http://ambermd.org/downloads/Amber11.18apr10.tar.bz2
>> ...I (not very surprisingly) still get a 100% failure rate when running
>> "make test.nab" on fedora core 12 x86_64 with a vanilla gcc-4.4.3, since
>> eeach and every test case stumbles over some garbled string (wrecking
>> e.g. the name of a file to open or a force field option string).
>> This even happens when replacing all optimization by "-O0 -g3", which
>> will hopefully help tracking this down.
>> It just might be a hideous compiler bug in gcc 4.4.3 only striking on
>> FC12 x86_64, but I doubt that and valgrind corroborates my doubts.
>>
>
> I've noticed the same problem. I can get it to work with opensuse (version ?)
> with a recent gcc version, but not fedora 12... so it might be a fedora
> issue vs a gcc issue, but I can't be absolutely sure about it.
>
> Here are some notes that I sent to myself in the past regarding this issue:
>
> ===================================================================================
> opensuse (version?) x86_64; Xeon X5355; gcc/gfortran 4.1.2 MSI Calhoun Cluster
>
> Fails to compile rism/_safemem.f
> gfortran -c -O3 -ffree-form -o safemem.o _safemem.f
> _safemem.f: In function ‘reallocate_pointer_1d_character’:
> _safemem.f:914: internal compiler error: Segmentation fault
> Line 914 reads:
> IF(ASSOCIATED(p)) then
>
Is this a prerelease version of 4.1.2? There has been another report of this problem with 4.1.2 prerelease from 2006/11/15. I primarily use 4.1.2 on CentOS and have never come across this problem.
Tyler
>
> ===================================================================================
> opensuse (version?) x86_64; Xeon X5355; gcc/gfortran 4.3.3 MSI Calhoun Cluster
> opensuse (version?) x86_64; Xeon X5355; gcc/gfortran 4.4.0 MSI Calhoun Cluster
>
> Compiles and PASSES all NAB tests
>
>
> ===================================================================================
> Fedora 12 x86_64; Nehalem E5520; gcc/gfortran 4.4.2 Machine 1
> Fedora 12 x86_64; Nehalem E5520; gcc/gfortran 4.4.3 Machine 1
> Fedora 12 x86_64; Nehalem E5520; gcc/gfortran 4.4.3 Machine 2
> Fedora 12 x86_64; Nehalem E5520; gcc/gfortran 4.4.3 Machine 2 with -m64 flag My machine
>
> All nab tests fail with garbled strings.
> Machine 1 and machine 2 are identical. Tested on both to make sure
> that I hadn't done something weird to (my) machine 1.
>
> -Tim
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
________________________________________________________________
Tyler Luchko
Postdoctoral Researcher
BioMaPS Institute
Rutgers University
Piscataway, New Jersey
tluchko.rutgers.edu
732-445-0334
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Apr 19 2010 - 08:30:03 PDT
