[systemd-devel] Wtrlt: Antw: [EXT] Re: Why are core dumps named vgcore.*?

[Ulrich Windl]

(Once again forgot the list)

>>> Ulrich Windl <Ulrich.Windl at rz.uni-regensburg.de> schrieb am 15.06.2021 um
14:25
in Nachricht <60C8B857.ED38.00A1.0 at rz.uni-regensburg.de>:
>>>> Mantas Mikulenas <grawity at gmail.com> schrieb am 15.06.2021 um 10:45 in
> Nachricht
> <CAPWNY8WdFU3d_of5JAyi=zODAjbs0N+MvJD4RJVgFfsh9U9H5Q at mail.gmail.com>:
> > On Tue, Jun 15, 2021 at 9:04 AM Ulrich Windl <
> > Ulrich.Windl at rz.uni-regensburg.de> wrote:
> > 
> >> Hi!
> >>
> >> I'm developing a program that dumps core on some failed assertions. I
> >> noticed that the core dumps are created in the user directory as
> >> vgcore.<PID>.
> > 
> > 
> > This doesn't sound like a coredumpctl-managed core dump at all. In fact
it
> > sounds like the dump was created in userspace by Valgrind.
(Systemd-managed
> > core dumps would always go to /var/lib/systemd/coredump, not to the cwd.)
> 
> Thanks, I wasn't aware  that valgrind affects the name of the core dump.
> 
> 1)Actually I was using all three: Run the program normally to let it dmp 
> core.
> 2) Run the program in gdb to avoid core dumps
> 3) Run the program in valgrind occasionally.
> 
> I was assuming the core dumps are controlled by systemd as cordumpctl gdb 
> seemed to open the correct core dump.
> 
> Regards,
> Ulrich
> 
> 
> > 
> > 
> >> Where does the name vgcore come from?
> >>
> > 
> > I used codesearch.debian.net to track it down to valgrind's m_coredump
> > component.
> > 
> > 
> >> And is it OK to remove just those files, or does coredumpctl store
> >> additional infos?
> >>
> > 
> > Most likely coredumpctl doesn't have *any* information about this dump,
> > since it didn't go through the kernel or systemd in the first place.
> 
> Untrue:
> Tue 2021-06-15 12:54:29 CEST  22788  1025  1009   6   
> /usr/lib64/valgrind/memcheck-amd64-linux
> 
> > 
> > But in general, coredumpctl is fine with missing/removed coredump files
--
> > that's part of the normal operation; actual dumps are cleaned out much
> > faster than the corresponding journal entries. You'll probably already
see
> > some of them marked "missing" in the list.
> 
> Regards,
> Ulrich
> 
> > 
> > -- 
> > Mantas Mikulėnas
> 
> 
> 
>