Now, just as I was leaving my job, we had finished porting our code to
Unix.
I'm not sure how they solved the problem, but just writing 'core' to a
directory obviously won't work - we will overwrite them on the next run.
Having 'core.pid' or 'program_name.pid.core' or something like that would
have made life a lot easier.
Does anyone know of a better way to handle this situation? Should there
be a
way for a process to tell the kernel the name of its core file? That
seemed
like a good idea to me, and I was going to play with it in Linux, but
switched jobs and never got the chance to play with it.
I guess we could have used some sort of daemon which would look for the
'core' files and file them away some place too. What do major software
vendors do to handle this situation? Assume their code never crashes :-)?
-- Erik
talvola@gnu.ai.mit.edu
(or in case of last resort, talvola@aol.com)