11-11-2010
Stack only core dumps
I'm working on a program in Linux with a group of people scattered around the country. When we have a crash, I like to send a core dump to the appropriate person so that they can understand the problem better. The problem is that our application uses several gigabytes worth of data and these core dumps are very large. Sending these core dumps is getting difficult.
From my experience, all I really need from a core dump is the stack. I'll look at the back trace in gdb, and maybe inspect some local variables. Is there a way to tell Linux only to dump the stack rather than all memory?
I've tried using "ulimit -c" to change the upper limit on the size of the core dump but there doesn't seem to be any guarantee that this will include the stack. In my case, these truncated core dumps don't produce a backtrace in gdb.
Another thought that I had is that there may be a utility out there that will extract the stack from the core dump and produce a second, smaller core dump. I've been searching for a while and I can't find anything. Does anyone have any suggestions?
Thanks!
10 More Discussions You Might Find Interesting
1. UNIX for Advanced & Expert Users
Does anyone know how to read core dumps. Is gdb the only tool for it ? The OS is Solaris.
Thanks (2 Replies)
Discussion started by: suntan
2 Replies
2. Solaris
I have the following set up on a Sun server running solaris 5.8 for core dump generation
coreadm
global core file pattern: /var/core
init core file pattern: /var/core
global core dumps: enabled
per-process core dumps: enabled
global setid core dumps:... (4 Replies)
Discussion started by: handak9
4 Replies
3. Programming
hi everyone,
Right now when I do:
strcmp(s1, s2);
i get a core dump because at times s1 or s2 can be nothing so that makes strcmp() core dump.
What is the solution, if at times I expect one of them (or both) to be NULL?
I want to be able to compare that s1 is NULL and s2 is "blah" or... (6 Replies)
Discussion started by: annie
6 Replies
4. SCO
Anyone know how you go about interrogating a panic / core dump with crash for SCO Unix (5 Replies)
Discussion started by: ccarcher
5 Replies
5. UNIX for Advanced & Expert Users
Hi,
Can we modify the GDB source code so as to analyze core dumps from different targets? From my analysis, I think we need to build our section table statically for each target. i.e., including the various address boundaries in build_section_table() function. If this is the case, then the GDB... (2 Replies)
Discussion started by: nsdeeps
2 Replies
6. Programming
My program is not dumping core when hitting a segmentation violation inside a thread.
However, it dumps core when the segv occurs within main.
Any ideas on how to diagnose this?
AIX 5.3 (4 Replies)
Discussion started by: bean66
4 Replies
7. Solaris
i had a situation where a process was defunct. preap would not reap the process and gcore would not work properly (not sure why). therefore, the suggestion was to force a panic and collect the core dump. obviously you could do a savecore -L and capture the dump without bringing down the system.... (3 Replies)
Discussion started by: pupp
3 Replies
8. Ubuntu
Hello,
I am debugging a program which works as daemon. It sigfaults, unfortunately I'm unable to generate core dump file. Here is what I am doing:
tsurko@bastila:~$ ulimit -c unlimited
tsurko@bastila:~$ ulimit -c
unlimited
tsurko@bastila:~$ cat /etc/sysctl.conf | grep 'core_pattern'... (1 Reply)
Discussion started by: tsurko
1 Replies
9. Red Hat
Hi
I have a Fedora installed and I try to generate my application's core dump file.
My system has no coredump limit:
$ ulimit
core file size (blocks, -c) unlimited
But when my application crashes no core dumps generated. I can generate dump file using gcore but it is not appropraite... (1 Reply)
Discussion started by: xyzt
1 Replies
10. HP-UX
kill -SEGV <pid> gives me the core file for that process but also terminates the process.
Can I not get the core dump without terminating the process ? (2 Replies)
Discussion started by: mohtashims
2 Replies
LEARN ABOUT CENTOS
abrt-action-generate-core-backtrace
ABRT-ACTION-GENERA(1) ABRT Manual ABRT-ACTION-GENERA(1)
NAME
abrt-action-generate-core-backtrace - Generates coredump-level backtrace
SYNOPSIS
abrt-action-generate-core-backtrace [-v] [-r] [-d DIR]
DESCRIPTION
This tool uses coredump from the file coredump and binary at the path stored in file executable in the problem directory to generate
coredump-level backtrace.
Coredump-level backtrace resembles ordinary backtrace in that it contains information about call frames present on the stack at the time of
the crash. However, it only contains information that can be obtained from the coredump without debugging symbols available - mainly
relative addresses of the stored instruction pointers. Such backtrace can still be useful for reporting and reproducing the bug and does
not require debugging information files to be installed.
The result is saved in the problem directory in a file named core_backtrace.
Integration with libreport events
abrt-action-generate-core-backtrace can be used as an analyzer for application crashes which dump core.
Example usage in report_event.conf:
EVENT=analyze analyzer=CCpp
abrt-action-generate-core-backtrace
OPTIONS
-d DIR
Path to problem directory.
-r
Do not hash function fingerprints. Useful for debugging.
-v
Be more verbose. Can be given multiple times.
AUTHORS
o ABRT team
abrt 2.1.11 06/18/2014 ABRT-ACTION-GENERA(1)