Difference between revisions of "Stack Trace"

Revision as of 15:01, 8 February 2011

What is a Stack Trace?

If GnuCash terminates abnormally (i. e. it crashes), then a stack trace is a very helpful information for the developers to figure out the program error that caused this crash.

Additionally, some interesting logging messages are written into the Tracefile at each run of GnuCash.

Also on Windows?

However, on Windows it is currently not so easy to obtain such a stack trace. Some explanation on how to do this for experts is on Windows#Debugging_with_gdb, but we are aware of the fact that this is rather difficult for the average user. If you have been asked to provide a stack trace, but you are using GnuCash on Windows only and feel too unfamiliar with gdb debugging, please respond along the lines of "I'm sorry, but I encountered this problem only on Windows and I am not familiar enough with gdb on Windows to obtain stack traces there". Thanks a lot. Nevertheless, in many cases your bugreport can only be worked on if you or someone else can actually produce such a stack trace.

How to obtain a Stack Trace

There are two possibilities to obtain a stack trace. You can choose either one, whichever seems easier or works better for you.

Attach gdb to running process

The program "gnucash" itself is not an executable but a script. Therefore it cannot be called directly by the gdb debugger. We propose to "attach" the gdb to the running gnucash process instead.

To attach to the running gnucash process:

1. Start gnucash normally
2. Use "ps afx" (some people prefere "ps aux") to find out the process number (PID) of the running gnucash process; it might be called "gnucash" or "gnucash-bin". Say the PID is 12345.
3. Start gdb by typing "gdb"
4. Attach to the running process by typing "attach 12345"; gdb will temporarily stop the program
5. Continue program execution by typing "continue" at the "(gdb)" prompt
6. Provoke the crash; type "backtrace" or shorthand "bt" at the gdb prompt to obtain the backtrace ("bt full" might provide additional info about local variables).

Please submit the backtrace together with the instructions on how to reproduce this crash as a new bug report into Bugzilla. Thank you for your contribution.

Starting GnuCash under gdb

Sometimes GnuCash might crash too early in the startup sequence such that you cannot attach gdb to the running process. In that case you can run gnucash under gdb. This is much easier as of GnuCash 2.0. To run GnuCash under gdb from the beginning:

1. run "gnucash-env gdb gnucash-bin"
2. at the gdb prompt, type: "run". Also add your parameters here, like "run --nofile"
3. Provoke the crash; type "backtrace" or shorthand "bt" at the gdb prompt to obtain the backtrace ("bt full" might provide additional info about local variables).

See also

Other explanations on how to obtain stack traces:

• Windows#gdb for Windows-specific comments
• http://live.gnome.org/GettingTraces
• http://www.dirac.org/linux/gdb/06-Debugging_A_Running_Process.php
• (in German) http://www.linuxwiki.de/gdb

Debug symbol information

Gnucash needs to be compiled in a way that has the debug symbols still included in the resulting binary libraries. When compiling gnucash from the tarball yourself, this is achieved by using the configure option ./configure --enable-debug.

Gentoo

When emerging gnucash on gentoo, you need to make sure that

1. the "debug" USE flag is enabled
2. and "FEATURES=nostrip" is set to disable symbol stripping.

Ubuntu

On Ubuntu, the debug symbols are available in a separate package called gnucash-dbgsym. See https://wiki.ubuntu.com/DebuggingProgramCrash

This issue is also explained here https://wiki.ubuntu.com/Backtrace.

Debian

FIXME: Can Debian use the same dbgsym packages as Ubuntu? Maybe this helps: http://packages.ubuntu.com/feisty/devel/pkg-create-dbgsym which is explained here http://lwn.net/Articles/201302/