This seems to be caused by some little boogers in the core code. We
decide whether to use CORE_REAL_RLIM based on whether it is >
min_dumpsize, which seems to be pretty small, and has nothing to do with
the actual minimum that it will take to dump the core for this
particular binary.
It seems to just be a set size for the particular binary format (elf in
this case). So what I've done is to always for rlim to (1024*1024*1024)
when piped, and pass it via CORE_REAL_RLIM.
So you'll need to check this to decide if the core gets written out. If
it's 0, obviously you need not write anything. Remember this value is in
bytes (where as ulimit -c is set in 1k increments, so a ulimit of 1 ==
rlim of 1024 bytes).
Martin, pushing this code to git. You can either compile it from there,
or wait for the upload.
** Changed in: linux-source-2.6.20 (Ubuntu)
Importance: Undecided => Medium
Assignee: (unassigned) => Ben Collins
Status: Unconfirmed => Fix Committed
Target: None => herd-5
--
does not set CORE_REAL_RLIM correctly
https://launchpad.net/bugs/87065
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
