On Tue, 2010-02-02 at 07:25 +0100, Bas Wijnen wrote: > On Mon, Feb 01, 2010 at 07:42:50PM +0100, Roland Clobus wrote: > > On Mon, 2010-02-01 at 19:24 +0100, Roland Clobus wrote: > > And I forgot to mention: > > > > I've also removed the directory part for /var/run, because it is not > > needed anymore, and I don't think a directory for only one file makes > > sense. > > This is true, but it is the only way to make things work: > - when using --background, start-stop-daemon does a double fork, to > orphan the (grand)child from the parent.
I remember working on the init script some time ago. This time I did run the meta-server as a regular program (no --daemon) and decided to let the start-stop-daemon program handle the PIDs. This turned out to be working fine. (So there is no double fork, just a single one, only by start-stop-daemon). > - pioneers-console must write its own pidfile; if start-stop-daemon does > that, it writes the dying child's pid. The running grandchild's pid > is not recorded. > > - The directory is needed because pioneers-meta-server isn't allowed to > write to /var/run. start-stop-daemon is invoked with 'make-pid-file' so it will create the pidfile by itself. That also means that the access rights are OK. I noticed that the pidfile needs to be removed by /etc/init.d/pioneers-meta-server because start-stop-daemon will only create it. (Not present in the patch) > If your patch works (I didn't test yet), then start-stop-daemon is > probably changed. Do you know if this is the case? With this new init-script I'm able to successfully start and stop the meta-server. I think the main change was in the invocation of the meta-server. With kind regards, Roland Clobus
signature.asc
Description: This is a digitally signed message part
