On Thu, Jan 22, 2004 at 01:12:57AM +0000, Colin Watson wrote:
> On Wed, Jan 21, 2004 at 08:03:30PM -0500, Branden Robinson wrote:
> > On Wed, Jan 21, 2004 at 11:47:02AM +0000, Colin Watson wrote:
> > > I think that we should recommend against using (= ${Source-Version})
> > > dependencies from Architecture: all to Architecture: any packages or
> > > vice versa. They're OK between architecture-independent packages and
> > > between architecture-dependent packages, but mixing those causes
> > > trouble.
[xlibs-static-dev and xlibs]
> If the tight binding is necessary, an alternative would be:
>
> Package: xlibs-static-dev
> Depends: xlibs (>= ${Source-Version}), xlibs (<< ${Next-Source-Version})
Ugh, ugh, ugh. That's so gross. So horribly gross.
Not saying I prefer the problem you're trying to fix. Just saying ugh.
> ... where Next-Source-Version (or whatever) is a substvar set to the
> hypothetical version of the next possible sourceful upload (so 4.3.0-1
> => 4.3.0-1.1; 4.3.0-1.1 => 4.3.0-1.2). I haven't tried writing the code
> to see how hard such a substvar would be to generate automatically.
"Next hypothetical version" is a big leap. It runs against the grain of
my instincts to put speculative "knowledge" into our package dependency
graph.
> > I'm tempted to support your proposal, however.
>
> Thanks, I appreciate knowing that I'm not off the rails here. :) I'd
> like there to be some example implementations of automatically
> generating dependencies like the above before turning this into a formal
> policy proposal.
It turns out I don't think I needed that dependency anyway, and I got
rid of it. The svn commit log and changelog entry make the case for
doing so.
Thanks for prompting me to check for it. :)
--
G. Branden Robinson | Religion is regarded by the common
Debian GNU/Linux | people as true, by the wise as
branden@debian.org | false, and by the rulers as useful.
http://people.debian.org/~branden/ | -- Lucius Annaeus Seneca
Attachment:
signature.asc
Description: Digital signature