Bug#492626: lintian: If only internal debconf questions are used, do not report not-using-po-debconf errors
tags 492626 pending
thanks
On Sun, Jan 11, 2009 at 11:08:49AM -0800, Russ Allbery wrote:
> Julian Andres Klode <jak@debian.org> writes:
> > On Sat, Jan 10, 2009 at 08:27:05PM -0800, Russ Allbery wrote:
> >> Russ Allbery <rra@debian.org> writes:
> >>> This is somewhat intentional since I hadn't yet seen a case of this
> >>> where doing that makes sense, other than when using some debconf
> >>> frameworks that Lintian recognizes. Could you point me at an example?
>
> >> We didn't get a reply to the above question, so I'm going to go ahead
> >> and close this bug with the reasoning in the paragraph above. If you
> >> do have examples where Lintian gets this wrong, please feel free to
> >> re-open.
>
> > For example, readahead-list just uses the debconf questions for
> > preseeding, has no config script and does not need translated
> > descriptions (as we do not display them).
>
> Okay, thanks. I'll reopen.
>
> This is exactly the case where this doesn't make sense to me, though. Why
> would you support pre-seeding and not prompting? That to me seems
> contrary to the point of debconf and makes configuring the package
> unnecessarily obscure.
>
> If you asked the question with a priority of low, only those users who
> want to see all options would see it. It seems like if it's important
> enough to make controllable via preseeding, it's important enough to ask a
> low-priority question about.
>
> But maybe I'm missing something?
There are quite a few cases in the installer where this is exactly what
we want; we use debconf as a database and not every question we put
there corresponds to something we want to ask the user. Indeed, some of
these questions come before we are in a position to ask questions at
all: usb-discover is the example I have to hand just now, which runs
before the menu starts but can still retrieve preseeded answers from the
database.
It seems to me that putting "for internal use" in the description
constitutes an override, and while we might not agree in every case the
purpose of that string is to override Lintian. Besides, it's already
supported in checks/debconf as something where we don't check for
description best practices, and so it doesn't seem to make sense to
check for translations either.
I've gone ahead and committed a fix for this. If it's still
controversial after the explanation above, let's talk about it further;
perhaps a compromise would be to make it udeb-only, although I don't see
cause for that at the moment.
Cheers,
--
Colin Watson [cjwatson@debian.org]
Reply to: