FreeFem++ crashes with exit status 132 about freefem-sources HOT 10 CLOSED

Here is a backtrace obtained with debian official package (3.61.1+dfsg1-4).

$ gdb FreeFem++
(gdb) set pagination 0
(gdb) run
Starting program: /usr/bin/FreeFem++
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".

Program received signal SIGILL, Illegal instruction.
__static_initialization_and_destruction_0 (__initialize_p=1, __priority=65535) at lg.ypp:105
105 lg.ypp: Aucun fichier ou dossier de ce type.
(gdb) bt
#0 __static_initialization_and_destruction_0 (__initialize_p=1, __priority=65535) at lg.ypp:105
#1 _GLOBAL__sub_I_lg.tab.cpp(void) () at lg.ypp:989
#2 0x0000555555f16dd5 in __libc_csu_init ()
#3 0x00007ffff728202a in __libc_start_main (main=0x555555a6c4c0 <main(int, char**)>, argc=1, argv=0x7fffffffe098, init=0x555555f16d90 <__libc_csu_init>, fini=, rtld_fini=, stack_end=0x7fffffffe088) at ../csu/libc-start.c:264
#4 0x0000555555a70cda in _start () at ../Graphics/rgraph.hpp:145

from freefem-sources.

We have never seen this bug before.

Could you try to compile FreeFEM (master or develop branch) to see if the bug persists ?

from freefem-sources.

Here is the full list of the different ways I have been trying to run FreeFem++.

1/ debian/unstable package (3.61.1+dfsg1-4), machine 1: exit status 132
2/ freefem package from github (3.62), machine 1: exit status 132
3/ debian/unstable package (3.61.1+dfsg1-4), machine 2: success
4/ rebuilding the same debian/unstable package on machine 1 + running on machine 1: success
5/ building current master on machine 1 + running on machine 1: success

Machine 1 and 2 are different but run the same debian/unstable/amd64. The crashes are reproducible. But they happen only on machine 1.

The fact that the bug disappears when I rebuild the debian package make me think that it is a real bug. However, I agree that it is a very subtle one.

Thank you for your time and for FreeFEM!

from freefem-sources.

Please check the Debian BTS.
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=924009#30

Bernhard Übelacker found out that the issue might result from running freefem++ with a processor lacking AVX support when the binary was compiled by a processor with AVX. I reproduced the crash by running freefem++ on another PC without AVX support.

from freefem-sources.

from freefem-sources.

You should configure with --enable-generic if you are cross compiling. Could you confirm that this fixes your issue, please?

from freefem-sources.

from freefem-sources.

As far as I could test, --enable-generic works theoretically and experimentally. ;-)

I rebuilt the debian package (version 3.61.1+dfsg1-4) with and without adding --enable-generic to configure using a PC with AVX support. I tested the resulting packages using 2 PCs without AVX support. The package built with --enable-generic doesn't crash while the other one always does.

Thanks a lot for your help!

from freefem-sources.

@es132 , are you the offical maintainer of the freefem debian package ?
We have some issues with our jenkins build script (https://github.com/FreeFem/FreeFem-sources/blob/master/etc/jenkins/deployDEB.sh), we only managed to create a package for ubuntu 18.04. Would you be able to help us configure it by submitting a PR? We are not sure if using checkinstall is the best way to do it.
Thank you in advance !

from freefem-sources.

@es132 , are you the offical maintainer of the freefem debian package ?

No. I can't help you with jenkins either. Sorry ...

from freefem-sources.