Hi,
ok, it is obviously now: if i start batmand without the -d option and if i do not start 'batmand -c' everything runs just smoothly!
i will be experimenting with this in the next days (probably monday) and will report my experience.
thanks for your bug report. I just issued 2 patches. Number one reenables core dumps on segmentation fault. Batman listens for that signal and tries to clean up the routing table, interfaces, etc. Therefore batman exited gracefully and the core dump was not created.
But if batman had received SIGSEGV it would print the message "Error - SIGSEGV received !". It seems that did not happen ?! I got the idea that you may be disconnect with your client in the very moment when batman tries to write a message into the connection. The kernel would recognize the disconnect and sends SIGPIPE to batman which exits as this is the default behaviour. The second patch changes that. It is a wild guess ... :-)
Please, try it out.
Regards, Marek