Re: Draft spec for new dpkg "triggers" feature (v2, repost)

[Florent Rougon <f.rougon@free.fr>]

Hi,

Frank Küster <frank@debian.org> wrote:

> I don't think so. Usually, the lines "quoted" above are caused by
> apt(itude) invoking dpkg like this:
>
> dpkg --configure lmodern other-font-package texlive-base-bin bar baz wrmft-foo
>
> And then the intended effect will be achieved.

OK. In this case, that's acceptable. Not perfect nevertheless, because I
think that in non-trivial upgrades, there are often several dpkg runs
(e.g., unpacking, configuring, unpacking again, configuring again), and
this would lead to mktexlsr & Co being run several times before
texlive-base-bin is configured in the end, but it's still much better
than the current situation.

> The reason can be extended by the scenario that I is installed after the
> Ts (and no T depends on I, maybe Suggests or Enhances).
>
>     A package I may be installed and configured after and independently
>     of packages T which would have activated one of its triggers, had it
>     been installed earlier.  Therefore, I's postinst must do any work
>     needed for trigger processing even when called with "configure".

I agree, although it's not strictly speaking "trigger processing" in the
last line, since there won't be any trigger in this case. It's rather
something like "the work that needs to be done, and would have been done
by trigger processing, if package I had not been in a state where it
ignores triggers". But it's just a matter a formulation, and we agree on
the meaning (« nous sommes d'accord sur le fond », as we would say in
french).

Regards,

-- 
Florent