I've tried everything (I think) and I cannot get a core dump!

I did ulimit in bash, limit in csh. It says core size is unlimited.

I've edited config files. I've scoured around the net for hours.

Why can't I get a core dump in Zenwalk? I have version 6.

I can gdb the program, no problem. I'm not a noob.

But it's important for my project to get core dumps when crashes occur.

Any ideas?

For me, the code:



gcc -ggdb -o test test.c

generates a core dump... I am using Zenwalk 6 too.

Well, I tried that. I get the seg fault, but no dump.

I did a fresh install of Zenwalk 6 Standard. I set the core limit to unlimited, I made sure permissions were set on the current directory for rwx, I even tried it in the home directory, and then as root. Nada.

I'm at a loss.

And the plot thickens...

I installed 6.2 RC1. Did the ulimit thing. Checked permissions. Logged in a root even. Ran "test" - Seg fault yes, core dump, no.

Went to another machine with 6.0 installed. Same deal. No dump. How frustrating!

Sorry for asking but how are you setting the core limit? As far as I remember, I didn't need to set the limit. This is very odd... do you get cores in other distributions?

In bash, "ulimit -c unlimited"
In csh, "limit core unlimited"

I read that to prevent core dumps, set the limit to zero; which is by default. It also said that to enable them, make sure the permissions are set for the current user to write files, and that the core limit is above the size of the core dump; unlimited is prefered.

Well, it doesn't dump a core. I haven't tried to put "test" on other distros. It probably will dump a core. I could try other distros but then I run into compatibility issues with old hardware or trying to set up libraries. (Out of frustration of trying to get the BFD - binary file descriptor - library to work in other distros, I switched to Zenwalk because it's installed and working by default).

Is there something I'm doing wrong in setting up Linux? I basically do a default setup. 8GB linux file system with a 1GB swap. Does the file system format matter? I chose the default.

In the kernel did you check CONFIG_ELF_CORE option?

I don't know how or where to find that information out about CONFIG_ELF_CORE.

To see your current kernel configuration you can try to unzip /proc/config.gz. I can only unzip it after copying it to another directory. Something like:

cd ~
cp /proc/config.gz .
unzip config.gz
grep CONFIG_ELF_CORE config (or .config, don't remember right now)

Hope that helps.

Neither CONFIG_ELF_CORE nor CONFIG_IKCONFIG_PROC (/proc/config.gz) are activated by default in the kernel. At least not in ZW5.2. If you did not rebuild or upgrade your kernel you can check in /boot/config:
grep -E "CONFIG_ELF_CORE|CONFIG_IKCONFIG_PROC" /boot/config

If this is the case you should rebuild your kernel and enable CONFIG_ELF_CORE. If you also enable /proc/config.gz support you can check after:

zcat /proc/config.gz | grep CONFIG_ELF_CORE
CONFIG_ELF_CORE=y

ulimit -c unlimited
./test
Segmentation fault (core dumped)
ls -l core*
-rw------- 1 ivan users 139264 2009-08-26 06:40 core

I can confirm that with this flag and ulimit -c unlimited a core dump is created.

It works! I simply (or not so simply) had to rebuild the core with the CONFIG_ELF_CORE option turned on. Which brings up an interesting point: Why wasn't this on in the first place?! Sheesh.

Thanks for your help!