Re: modules

[dann frazier <dannf@dannf.org>]

On Wed, 2005-09-07 at 18:25 +0200, Sven Luther wrote:
> There is currently no way to solve this short of a rebuild, and only a
> reupload can do this rebuild.

Well sure - what I'm suggesting (and sounds we all are) is that we make
the upload a direct result of the thing that triggered it, or a single
automated action after the fact.

This could be done by:
  1) Have linux-2.6 build-dep on all well-behaving kernel-module-source 
     packages, and have it spit out the module binary debs.
  2) A separate source package that does the above, but avoiding 
     polution of the already heavyweight linux-2.6 package
  3) The automated build/sign/upload of each individual package.

  And I'm sure others exist.

Option #1 extends our existing approach of removing interpackage
dependencies by consolidating their build into a single source package.
It also means that everything moves into testing (or doesn't) at once,
and avoids the problem of having to manually prevent unbuildable or
incompatible modules from hitting a stable release.  

Though - I'm just throwing it out there.  I think #3 is my current
favorite.  I've been historically opposed to consolidating packages
because I've feared it would take some responsibility away from
otherwise active maintainers and pile it onto an already overextended
group.  However, linux-2.6 seems to be working out quite well, and I
think #3 is a good compromise - we don't take over maintainership, we
just take over maintenance uploads.  If the module stops building,
that's the maintainer's problem, not ours - and it won't stop us from
moving into testing.