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

Re: [custom] Custom Debian Distros need the help from debian developers



[Thomas Viehmann]
> Yeah, but low is not the priority never asked.
> If you absolutely want to go with preseeded debconf values, maybe
> you could ask Joey to allocate a priority "invisible" (or whatever
> you want to call it) that is exclusively used to do CDD
> configuration where, as in the case of NTP, prompting *a user* is
> completely unneccessary.

This is possible at the moment.  All you need to do is to have
"hidden" debconf questions with no code to present the question, and
only code to set and get the value.  Something like this in the config
script:

  if [ -f /etc/configfile ]
     value=`get_value /etc/configfile`
     db_set hidden/debconf/value "$value"
  fi

and in the postinst script, one read the current value:

  db_get hidden/debconf/value
  modify_value_unless_equal /etc/configfile "$RET"

This way the value in the debconf database is used unless the file in
/etc/ exist already, and the value in the current configuration file
is used if the file exist.  We had added two such invisible debconf
questions to the ntp package, to enable a local NTP server.  Debian
Edu are trying to get more such questions into the packages we use,
and continue to supply patches to make this happen.  All I am asking
for is that we don't have to do this work more then once for each
package. :)

> The thing that really disturbed me about the discussion is that the
> subject and introduction of Petter's mail suggested that package
> maintainers are unwilling to cooperate when they're only trying to
> reach policy goals (and minimization of user interaction is an
> eplicit goal in policy).

I did not intend this.  I believe and hope we will be able to work out
a solution that give us the best of both a good standard configuration
and the flexibility to preconfigure the packages differently at
install time.

> If you CDD guys (presently) need debconf for purposes other than
> user interaction (and again, I would not have any problem with
> that), you need to propose a way to do this and not whine about
> people cutting out debconf where user interaction is not needed.

I hope my pseudo-code above is a step in the right direction.



Reply to: