Re: coredumpctl integration for F25

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]



On Sat, 2016-05-14 at 07:09 +0200, Jakub Filak wrote:
> Well, if I need to analyze core dump files during development I
> prefer 
> no intermediate tool - no ABRT, no systemd-coredump. I just set
> 'ulimit 
> -c unlimited' and run gdb on the file generated in CWD. Actually,
> that's 
> not true, I run my programs in gdb, so I don't need to analyze local 
> core dump files  most of the time. I usually need to analyze core
> dump 
> files generated on foreign systems.

I prefer coredumpctl, because:

 * It's always listening. You don't need to set ulimit, run a program
from a terminal, and then try reproducing a tough to hit crash. You can
just use your computer as normal, and wait for it to crash.
 * It works for all processes on your system, including daemons which
are otherwise tough to get coredumps for.
 * Simply 'coredumpctl gdb' to open the most recent crash in gdb.
 * No need to remember to delete the coredumps; core dumps are stored
in one place and removed automatically based on reasonable disk space
limits. (ABRT does this too when not changing ulimit.)
 * It's upstream. Users coming to Fedora from Arch and other distros
already know how to use it, and will not be frustrated when it doesn't
work as expected. (Currently we have it installed by default, but
totally broken out of the box.)

Grabbing coredumps from cwd feels like stone age to me. Having to set
ulimit is really inconvenient and frustrating when using an old system
that doesn't have systemd.

I used to have ABRT disabled in favor of coredumpctl, but now that ABRT
and coredumpctl work nicely together, I don't see any inherent conflict
anymore. So I still want to proceed with this for F25.


> I do prefer abrt-cli (or abrt - which can run gdb too) when I need
> to 
> work with errors/crashes/problems of any source - Python, Java,
> C/C++, 
> Kernel oops, Ruby, etc. I can analyze them and eventually report
> them.

No doubt ABRT is better for automatic bug reports and for handling non-
C/C++ crashes. Fortunately, we can have both!

> systedm-coredump makes ABRT's job harder. Currently, ABRT fails to
> get 
> core dump files from systemd-journal because systemd-coredump
> switched 
> from xz compression to lz4 compression. My bad, I should have been 
> watching systemd development.

OK, this is a problem. :( You're planning to fix it?

Michael
--
desktop mailing list
desktop@xxxxxxxxxxxxxxxxxxxxxxx
http://lists.fedoraproject.org/admin/lists/desktop@xxxxxxxxxxxxxxxxxxxxxxx




[Index of Archives]     [Fedora Users]     [Fedora KDE]     [Fedora Announce]     [Fedora Docs]     [Fedora Config]     [PAM]     [Red Hat Development]     [Red Hat 9]     [Gimp]     [Yosemite News]

  Powered by Linux