Re: binNMU for libthai/libdatrie transition
+ Loïc Minier (Fri, 17 Apr 2009 13:24:15 +0200):
> On Fri, Apr 17, 2009, Adeodato Simó wrote:
> > No big worries. It’s very likely I would’ve missed myself the fact that
> > pango1.0 would be rendered uninstallable for a while with that scheme.
> > And thanks for pushing for decreasing the spurious linkage in the archive!
> Not sure who you were saying that to, but I did raise that we should
> avoid the transition; perhaps I should have used stronger words
The message you linked does not talk about avoiding the libdatrie
transition at all, but about not propagating it to packages that do not
depend on it. That would’ve resulted (well, has resulted) in pango1.0
not being part of the transition, so that’s what I thanked you about,
but I can hardly see how we could’ve skipped the transition. But I’m
afraid I haven’t read all the thread.
> On Fri, Apr 17, 2009, Adeodato Simó wrote:
> > The thing is that it’s infinitely better to rebuild pango1.0 and make it
> > depend on libdatrie1, and then rebuild it again once libthai was fixed
> > in order to make the dependency go away completely, than to wait and
> > rebuild it only once when a fixed libthai is available. Because with the
> > former way you don’t render it uninstallable at any point, but with the
> > latter you do.
> I think you're mixing libdatrie and libthai above, not sure; I was
> proposing to:
> - change libthai/unstable to not cause linkage on libdatrie
> - rebuild pango against that to drop the datrie dep
> - upload new datrie
> - rebuild libthai against it
Yes, but that’s not the way it happened AFAIK. The order was more like
(enumerating the above four items): 1, 3, 4, 2 or so, which did render
pango1.0 uninstallable. Your proposed order is of course much better,
because you don’t do spurious rebuilds and pango1.0 doesn’t get
uninstallable; I was merely pointing out that, with the given order of
uploads as they happened, spurious rebuilds would have been better than
uninstallability of pango1.0. I hope I explained myself clearly.
> > Of course, you weren’t supposed to be aware of this, since it’s rather
> > specific knowledge, so no blame is involved. :-)
> Theppitak is upstream for datrie and libthai as well
Well, I meant knowledge of the Debian environment in order to know what
order of events is preferable. You, for example, proposed a very good
order, but for some reason it was not followed. With the order of
uploads as they happened and AIUI, Theppitak thought it was better to
rebuild pango1.0 only once a fixed libthai.{pc,la} was available, which
I think it’s a very reasonable guess of what’s best, but it is actually
not.
I hope this cleared up things a bit. It’s just too sad the delay this
imposes in several of our ongoing transitions, but as I said, no blame
is involved AFAIC.
Cheers,
--
- Are you sure we're good?
- Always.
-- Rory and Lorelai
Reply to: