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

Re: ntp 4.2.0 in experimental



On Sun, Feb 15, 2004 at 05:17:35PM +0000, Mark Brown wrote:
> On Sun, Feb 15, 2004 at 10:48:31AM -0600, Steve Langasek wrote:
> > On Sun, Feb 15, 2004 at 10:30:18AM +0100, Andreas Barth wrote:

> > > What's again using debconf with two low-priority questions:
> > > (1) Use debconf to configure this package [yes]

> > The first of these questions is *always* *always* *always* a bug.  There
> > is no reason to ask if debconf should be used, instead of just asking
> > the questions about the config settings debconf will change, except when
> > the debconf handling does not correctly preserve local modifications,
> > and asking with debconf does not excuse this behavior.

> The first question is often asked in order to allow users who don't want
> to use the Debconf configuration to avoid having to sit through reams of
> configuration which will only be ignored anyway.  That strikes me as
> being useful.

If there are configuration settings that warrant a high-priority debconf
question (meaning there is no reasonable default value), why would you
want to skip it when configuring the package?

If you don't want to see questions about settings that have reasonable
defaults, why would you ever have configured debconf to show questions
that are at lower than high priority?

The real reason for this question appearing in most packages where it
does is that it paints over a real bug in the package's re-parsing of
the config, resulting in a loss of local config changes (the "debconf is
not a registry" bug), which is not acceptable; and without this
underlying bug, the added question is not useful.

-- 
Steve Langasek
postmodern programmer

Attachment: signature.asc
Description: Digital signature


Reply to: