Comment 4 for bug 718635

Revision history for this message
Kees Cook (kees) wrote :

I think there is too much speculation happening in this bug report. :)

- /etc/default/apport's maxsize value isn't used anywhere since commit 1195 (and should be removed from this config file -- this is a separate issue, bug 719564).
- /etc/security/limits.conf is only used by PAM. X running as root from /etc/init (gdm) will not end up with these rlimits applied (and, regardless, coredump limits are ignored for the /proc/sys/kernel/core_pattern handler except when _not_ catching the crash (i.e. coredump writing pass-through)).

The error message given here relates strictly to size of the coredump vs available physical memory. If the coredump is larger than 3/4 of available memory, it will not attach it to the crash report, so this error message is accurate.

So, the question remains: how large is the X coredump, and how much free memory do you have?

As an example, on my system, I could catch an X crash. I have about 5GB of memory available, and X's memory footprint is about 128MB:

$ echo $(cat /proc/meminfo | egrep '^(MemFree|Cached|Writeback):' | awk '{print $2}') - + p | dc
4993428
$ grep ^VmSize /proc/$(pidof X)/status
VmSize: 128796 kB