Re: [binNMU] facile
On 11-09-2007, Steve Langasek <firstname.lastname@example.org> wrote:
> On Mon, Sep 10, 2007 at 04:59:28PM +0000, Sylvain Le Gall wrote:
>> On 10-09-2007, Julien Cristau <email@example.com> wrote:
>> > On Mon, Sep 10, 2007 at 13:40:27 +0200, Stefano Zacchiroli wrote:
>> >> You mean, I guess, that all that other 50 packages are ocaml-related and
>> >> can't be installed for the same reason above, right? If you have the
>> >> list at hand can you please let me (or firstname.lastname@example.org) nows
>> >> which packages you have scheduled?
>> > The list is at
>> > http://ftp-master.debian.org/~vorlon/transition-binnmus.txt
>> Taking a look at the list, i think that some binNMU are not needed (ulex,
>> pcre-ocaml, mtasc, extlib...) and some will certainly fails if not
>> taking care of dependency (e.g. ocamldap which need to be built before
>> ocamlnet and to reflect this dependency in the build-dep).
> Every package that's in that list is one that has a binary dependency on
> ocaml 3.09.2. Perhaps some of these packages aren't binNMU-safe, but every
> one of the packages is in need of a rebuild for ocaml 3.10.
The list is out of date, (because it has been build before a mass upload
of packages). Take a look at ulex:
It has been built with ocaml 3.10.0...
> Yes, there will be some packages that won't be buildable yet, but that will
> be addressed in due course. As for ocamldap and ocamlnet, you seem to have
> the dependency backwards...
OCaml dependencies is still an open issue among debian ocaml package
maintainers. We don't know how to show that there is no backward
dependency among certain version of libraries (i.e. package dependency
should reflect modules signatures changes... which is not that simple).
Anyway, concerning ocamlnet and ocamldap, i was simply stating that you
must build and install ocamlnet before trying to build ocamldap.
Otherwise, you won't be able to satisfy build dependency (because when
rebuilding ocamlnet, you will upgrade ocaml ABI dependency to 3.10.0).
In fact, i was pointing on the fact that rather than asking for a binNMU
(which can be useful for certain package), i think we should take a
closer look. This set of packages are being changed... I think it is
useless to fall into a "race condition"....
Sylvain Le Gall