[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: selfmade package installation hangs



On Sat, Jul 02, 2005 at 02:33:02PM +0200, Markus Schulz wrote:
> Am Mittwoch, 29. Juni 2005 19:28 schrieb Tilman Koschnick:
> > On Wed, 2005-06-29 at 14:36 +0200, Markus Schulz wrote:
> > > Unpacking replacement antzsystem-dispatcher ...
> > > Setting up antzsystem-dispatcher (2.1) ...
> > > Starting dispatcher: ++: 2005-06-29 14:24:36 (19338:B7D1CA40)
> > > dispatcher.cpp:296: Daemonize now...
> > > Daemonize with logfile=/var/log/antzsystem/dispatcher.log and
> > > pidfile=/var/run/antzsystem/dispatcher.pid
> > > Now running all modules from /etc/antzsystem/init.d
> > > <And now it hangs>
> > > ****************************
> > > the child process of /usr/share/debconf/frontend then becomes a
> > > zombie. i don't understand why the install hangs after running the
> > > postinst script (which runs fine and exit with 0) from the package
> > > and use invoke-rc.d script which successfully starts the binary
> > > (which daemonize himself). The program works successfully after
> > > this, but the install process hangs.
> >
> > This snippet is taken from the end of another postinst script that
> > starts a daemon, maybe it applies to your case as well:
> >
> > ------8<------
> > # tell debconf we are done. otherwise, it hangs waiting for the
> > daemon. db_stop;
> >
> > exit 0
> > ------8<------
> >
> 
> thanks, this works fine.
> perhaps this could be applied automaticly if debhelper(dh_installinit) 
> postinst script generates invoke-rc.d code snippet?
This might be a reasonable idea; submit a wishlist bug on debhelper.
One problem might be that db_stop could cause problems if called
multiple times.  Then, people's custom-written postinst scripts, which
already call db_stop (somewhere in the middle of a conditional, or a
function up top, which is called conditionally).  Of course, we can't
go breaking everybody's packages (well, maybe debhelper 5 can:)

Other flow control in the postinst further confuses this issue; if
someone has an "exit 0", then the db_stop at the end won't help.

Its at least worth an automated comment at the bottom.

Justin



Reply to: