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

Bug#758234: transitive dependencies


Charles Plessy:
> on my side I agree that self-contained priority levels are not needed anymore
> and are even becoming harmful.  This said, there were objections to the removal
> of this rule in this thread and in #759260, and I do not remember if we had
> good answers to each of them.  Matthias, do you think that you could make a
> summary of the pros and cons that were discussed in these threads ?

If I read #759260 correctly, Gerrit Pape <pape@dbnbgs.smarden.org> objected
to allowing depending on lower-priority packages and said that the current
"file a bug and raise the priority" process is just fine. However, IMHO it
clearly is, not because

* there's no technical reason (any more) to require this fix-override-file
* it installs superfluous libraries when you exclude something when
* we forget to de-prioritize libraries when they're no longer needed in
  the high-priority set

so I'd like to specifically invite arguments against this change here,
instead of digging through archives and inferring any opponent's reasoning.

> Regarding your proposed change, I wonder what is the practical case for
> forbidding conflicts with higher-priority packages.  Could you give an example
> showing that it is strictly necessary ?

I do not know of a concrete example. I did infer, from previous discussion,
that debootstrap et al. can't handle such conflicts. It seems prudent to
proactively forbid such dependencies, rather than deal with a last-minute
nonfixable bug filed on debootstrap.

> Lastly, while we are at it, let's insert a clarification that in Debian, the
> priority of the packages are determined by the archive administrators, and that
> the source package control file is not the canonical source of information for
> a binary package's priority when this package is distributed in the Debian
> archive.  (This would close #616055).
I see these issues as mostly orthogonal, so let's discuss this one first.

Besides, I'd rather get rid of the Override file as "canonical source" and
reduce it to a role as a temporary fix for wrongly-prioritized binary
packages, with the understanding that any Override entry shall have an
accompanying bug on the package it overrides.

-- Matthias Urlichs

Attachment: signature.asc
Description: Digital signature

Reply to: