On 2016-06-07 17:21, m.roth@xxxxxxxxx wrote:
> Got a user who claims he was running this program, then it broke recently.
> Almost no updates in a while, and none relevant. I'm guessing the
> program's compiled from fortran to c.... Anyway, the issue's on two
> servers. On one, I installed a couple of compat libs, and it runs. The
> other still fails (but it doesn't have some of the i686 libs. When it
> fails, it's immediate, and reports dying with a SIGKILL. ldd claims it's
> not a dynamic executable, but file tells me it's a 32 bit ELF.... Running
> it under strace gives me no more. So, as one last resort (Confess! You
> have three last resorts!), I'm trying to run it under gdb. On the system
> it runs on, I can b main, and it gets there when I tell it run. On the
> other... the address appears to be the same, it's in his NFS-mounted home
> directory - it never gets to main, it just reports the sigkill.
>
> So, what I'd *really* like to do is not tell it run, but next, or step,
> but I can't.
>
> Anyone heavy enough into gdb to be able to suggest where I can tell it to
> break during the program initialization phase, before it gets to main?
Possibly something still not right in the execution environment.
Try "export LD_DEBUG=all" then run the program. ("unset
LD_DEBUG" to reverse.) Did you edit ld.so.conf and run ldconfig
on the second machine? Also check "sudo ausearch --success no
--interpret -ts today"
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
https://lists.centos.org/mailman/listinfo/centos
