[ZODB-Dev] restart log weirdness

[Guido van Rossum]

> > I was aware of this and passed an absolute filename.  Besides, I was
> > working with the ZEO storage server, which doesn't chdir.  (Why does
> > Zope chdir to var?)
> 
> It is polite for daemon to chdir, rather than relying on whatever
> directory its parent was using.  It avoid problems where a system
> administrator cant unmount a filesystem, or delete a directory,
> simply because it is 'in use' as the daemon's current directory.
> 
> It also gives a predictable location for core dumps.
> 
> IMO the zeo server should do the same.

Fair enough.  Should it also chdir when not using zdaemon?

(I've got to say, I'm getting the hang of zdaemon now, and it's
actually a joy to use now that it works as intended.  I still want to
add a feature that it won't restart a frequently dying subprocess, but
I need to think of the use case where the subprocess dying can fix
itself.)

--Guido van Rossum (home page: http://www.python.org/~guido/)