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

Re: ARMv4-support in armel/squeeze?

+++ Andreas Barth [2010-12-21 21:22 +0100]:
> * Loïc Minier (lool@dooz.org) [101221 21:14]:
> > On Tue, Dec 21, 2010, Andreas Barth wrote:
> > > If it breaks, then yes. But what if the rdeps work, but perhaps with
> > > limited functionality for the user interface only? (Which means that
> > > any packages build-depending on it won't have any limitation.)
> > > 
> > > I think it would be ok to accept such limitations - but YMMV.
> > > 
> > > (Changing package names makes it harder for maintainers of such
> > > packages then just to say to configure "if there is no readline,
> > > build without")
> > 
> >  I'm not sure it's a good idea to allow building intermediate
> >  stages/packages with incomplete functionality with the same name as the
> >  fully functional one: they become undistinguishable for e.g. archive
> >  management software or end-users.

This is a really important question. Do we change the names and have
the real-package depend on 'foo-bootstrap', or do we accept that there
will be temporary 'reduced functionality builds' with something in the
control data to make them easily identifiable so they don't get
uploaded anywhere important?

> That's why I suggested to always put Bootstrap: yes within the
> packages control file. It still might not be a good idea.

I see no reason why this shouldn't work well. Personally I am leaning
towards reduced functionality builds with the same names, but am happy
to be persuaded the other way if we can see good reasons.

We are currently generating some build-dep graphs to define the
problem a little better. I'll post them when they look right. It at
least helps define the size of the problem, and may help clarify the
sorts of solutions that will work.

Principal hats:  Linaro, Emdebian, Wookware, Balloonboard, ARM

Reply to: