On Mon, Dec 18, 2006 at 04:29:10AM +0100, Peter Palfrader wrote: >On Mon, 18 Dec 2006, Steve McIntyre wrote: > >> >Only enabling it using dpkg-reconfigure makes changes persistent accross >> >upgrades. >> >> Pardon me for being dense, but I don't see a way in the postinst of >> knowing whether I'm being called via dpkg-reconfigure. Any clues? > >I don't think your postint needs to care. As I understand the >debconf-devel(7) manual page (section 'Config file handling') you need >to find out the current status of the system in your .config file, then >use db_set to inform debconf about that status, so you can use it in >your postinst script.
Yeah, OK. That's the bit I was missing. I can see that the postinst is called again, though... >In fact you already appear to be doing at least some db_sets in your >.config maintainer script, only CVS_ENAB_PSERV somehow needs to get >properly set based on the current inetd.conf. > >Something like this fixes the enable/disable problem: > >--- cvs.config.orig 2006-12-18 04:13:03.766969143 +0100 >+++ /var/lib/dpkg/info/cvs.config 2006-12-18 04:13:15.000000000 +0100 >@@ -49,13 +49,10 @@ > if [ "$CVS_PSERV_RESPAWN" ]; then > db_set cvs/pserver_spawnlimit "$CVS_PSERV_RESPAWN" || true > fi >- if [ "$CVS_ENAB_PSERV" ]; then >- CVS_ENAB_PSERV=`echo "$ENAB_PSERV" | tr A-Z a-z` >- if [ "$CVS_ENAB_PSERV" = "yes" ]; then >- db_set cvs/pserver true || true >- else >- db_set cvs/pserver false || true >- fi >+ if grep -q '^cvspserver' /etc/inetd.conf; then >+ db_set cvs/pserver true || true >+ else >+ db_set cvs/pserver false || true > fi > if [ "$OTHER_ROTKEEP" ]; then > db_set cvs/rotatekeep "$OTHER_ROTKEEP" || true > >[It seems update-inetd has no method to query the status of a service. Eww.] Yeah, that does seem like quite an omission :-( >However, this does not fix all of the cvs package's problems. For >instance the spawn limit is not read back either, resulting again in >manual config changes being lost on upgrades: <snip> >Another, related problem is the following: <snip> >That is, enabling pserver, when it is commented out with just a '#' does >not work. The reason for this is that update-inetd only touches lines >starting with '#<off># ' or the service port. It ignored lines that are >commented out: <snip> Yup. >I realize that fixing all this in time for etch will be hard. It would >be nice if the enable/disable change on upgrades could be fixed. The >other items probably are better left for later. If you agree please >clone the bug and handle them sepeartly. To be honest, I'm beginning to think that trying to manage the inetd entries using debconf is actually a mistake and doomed to failure. Maybe the first time the package is installed, it can be correct and accurate, but later on it's not possible to do things correctly. -- Steve McIntyre, Cambridge, UK. [EMAIL PROTECTED] Armed with "Valor": "Centurion" represents quality of Discipline, Honor, Integrity and Loyalty. Now you don't have to be a Caesar to concord the digital world while feeling safe and proud.
