Re: BinNMU changelog handling for Multi-Arch: same packages
On Wed, 2012-07-18 at 01:51:46 +0200, Philipp Kern wrote:
> On Tue, Jul 17, 2012 at 02:53:19PM +0200, Guillem Jover wrote:
> > > (I'd currently tend to accept it, however I'm not yet sure of all the
> > > consequences of this change.)
> > This change in itself should not have any consequence as it should not
> > affect the normal build if the value is not being used, or maybe you
> > are referring to *when* the value is being used?
> I don't want to special-case a single suite in buildd/sbuild to not put
> binary-only=something in there. And yes, those are only warnings, I know,
> but it would be annoying to have the build logs polluted with them.
I can understand not wanting to special-case this in buildd/sbuild,
but I don't really see the point in using this at least until wheezy
is released at which point squeeze will be oldstable, and then how
many binNMUs are usually performed for such a suite? (And polluting
seems to me a bit too much for at most few possible warning lines,
when lots of packages produce tons of compiler warnings for example.)
> (And it's one reason why we always expect stable to support the features we use
> in the archive.)
Well, just quiescing the warning on unknown changelog binary-only
keyword is not really supporting the feature. And I think doing that
would be a bad idea, because it might give the wrong impression such
feature is supported there, when it's not. At the same time I don't
think this feature should be backported either.