Bug#535645: Wrongfull removal of ia32-libs-tools
Russ Allbery <rra@debian.org> writes:
> Goswin von Brederlow <goswin-v-b@web.de> writes:
>
>> So far I have only seen one reason stated clearly: Might hinder
>> transition to multiarch. I think I have shown how I intend to handle
>> that transition in ia32-libs-tools whenever and however Debian will do
>> it. So far nobody has pointed out flaws in that plan.
>
> Personally, the main thing that makes me uncomfortable with the
> ia32-libs-tools approach from a design perspective is that it's
> manipulating data that previously in Debian we've assumed is invarient.
> In this thread some of the ideas discussed have involved rewriting the
> Packages file so that uses via a different tool see different package
Yes. Only way to do it without having the packages duplicated in the
archive. Which as I might remind you ftp-master has REJECTED.
> metadata, and creating Debian packages via a somewhat-official tool from
> the main archive that do not appear in any of our package tracking systems
> but which users may perceive as being part of Debian.
The BTS situation can somewhat be remedied by providing
/usr/share/reportbug/ia32-<package>/control. Bugreports can be send to
either the original package or ia32-libs-tools with
that. If bugs are send to the original package then
/usr/share/reportbug/ia32-<package>/script can also add a note that
this package was converted by ia32-libs-tools and so on.
Everything reported via reportbug would be covered that way.
> Even apart from the question of what specific problems people can
> anticipate in advance and what solutions we can find for those problems,
> both of those strike me as bad architectural approaches because they break
> assumptions that users and software make about the data consistency of
> Debian as a whole. They're the kind of decisions that have a tendency to
> cause unexpected and unanticipated problems in the future.
In the future we have multiarch. Aparently even for squeeze, or so is
the plan.
In the 1 1/2 years ia32-libs-tools has existed in Debian already there
has not been a flood of mysterious bugs and I think that track record
would hold. With multiarch to be in squeeze, we are talking only
about unstable users too. If the multiarch cabal holds their promises
then we are only talking some month to a year here.
> --
> Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
MfG
Goswin
Reply to: