Re: another preselections issue
>>>>> "Stephane" == Stephane Bortzmeyer <bortzmeyer@pasteur.fr> writes:
Stephane> On Wednesday 3 March 1999, at 2 h 42, the keyboard of Adam
Stephane> Di Carlo <adam@onshore.com> wrote:
>> Select 'minimal X11' task run dselect, update, install get error
>> about unmet dependancies: mc: Depends:libpam0g Depends:libpwdb0g
Stephane> This is a neverending task! How can we say that we release
Stephane> slink for four architectures when perfectly portable
Stephane> packages, not arch-specific in any way, exist only for some
Stephane> architectures of slink? The "master" file has now more
Stephane> conditionals than real lines. We lie when we say that Debian
Stephane> runs on several processors.
Woah, there big boy. Slow down.
There *are* libpam0g packages, and libpwdb0g packages.
It's obvious that the preselection mechanisms we currently use suffer
from serious maintainability problems. I think it's crisis-level
crucial that any preselection mechanisms can check all depends (and
even optionally recommends) against the current archive for all
architectures.
I personally think we should move towards a metapackage approach.
This way the perselections, as pretty much empty big packages that
depend on lotsa little packages, can be contributed to by all, can be
continued to be maintained and used by users (not just at
installation), etc.
We'd still need to take steps to add the consistency checking.
Stephane> New version for Sparc committed. Many packages excluded for
Stephane> the Sparc.
Stephane> It is still not perfect because some packages in the Sparc
Stephane> still depend on hamm stuff (e.g. xbase).
Dude, you *totally* shouldn't be using xbase at all.
Let me reiterate that: don't use the xbase pacakge!
It's just a meta package.... It's evil. Its purpupse was to make
sure upgrading users don't lose packages. It's going to suck in xdm
and you really don't want newbies installing xdm.
--
.....Adam Di Carlo....adam@onShore.com.....<URL:http://www.onShore.com/>
Reply to: