Hello Tibor,

> It would not work "as is" on FreeBSD where the "f" and "u" options
> behave differently.  The output information also differs there.  See
> <http://www.freebsd.org/cgi/man.cgi?query=ps>.

Uh, how different indeed!

> Ideally, we should not rely on "guessing" task PIDs via get_my_pid()
> way of doing things, but rather the tasks should store their PIDs into
> dot-pid files when they are started, so bibsched can retrieve them
> cleanly later if needed.  (Kind of like it manages its own PID via
> /opt/cds-invenio/var/run/bibsched.pid.)  This would constitute a
> proper solution to the problem.  We have not got to implement this
> yet.  Until done, I think we can safely stick to the current
> technique, unless it makes troubles on some of your systems?

No, no trouble at all.  I initially wrote it trying to help a Solaris
user a couple of years ago
(http://cdsware.cern.ch/lists/project-cdsware-users/archive/msg00374.shtml
and
http://cdsware.cern.ch/lists/project-cdsware-users/archive/msg00378.shtml).
I've been happily using on my systems since then, with some updates.  I
still think that calling a single external command is cleaner, but I
won't insist on that.  I can wait for the proper pid file solution.

Thanks,

Ferran

Reply via email to