Re: Bug#338435: status of the ocaml 3.09.0 migration ...
On Mon, Dec 19, 2005 at 03:32:40AM -0800, Steve Langasek wrote:
> On Mon, Dec 19, 2005 at 08:46:14AM +0100, Sven Luther wrote:
>
> > 04:24 < vorlon> hey all. Where are we on being able to close bug #338435?
> > 04:25 < vorlon> it turns out KDE needs ocaml to be pushed first, because
> > abiword-plugins uses gtkmathview and abiword and koffice both depend on
> > libwpd.
> > 04:26 < vorlon> (so, in fact, they need to be pushed in at the same time,
> > hurrah.)
>
> > Well, i guess this is good, as it seems ocaml is being used in more outside
> > projects, there is also the other package of Jeffries which is linked to KDE.
> > For info, 338435 is the blocker bug i put to avoid ocaml 3.09.0 entering
> > testing without us noticing.
>
> From discussions on IRC, there are three outstanding issues for ocaml:
>
> - FTBFS on hppa. This is already a separate RC bug against ocaml (342704).
> - the ABI-breaking fix for cduce; it's my understanding that this patch has
> not yet been included in the ocaml package in unstable, and including it
> would require rebuilding all of the ocaml packages, but that this doesn't
> necessarily have to wait for ocaml 3.09.1.
> - a handful of packages that still build-depend on ocaml-3.08.3 or similar,
> and require sourceful uploads for the ocaml transition. These packages
> are advi, approx, camlrpc, cduce, cryptokit, lablgtk, missinglib, mlgtk,
Fixed approx has been uploaded, mlgtk and lablgtk should go. For the rest, i
would just quick them out of testing right now and be done with it, they can
reenter when they are ready.
> ocamldsort, pycaml, spamoracle, syslog-ocaml. Each of these packages
> should have a separate RC bug filed against it if there are no immediate
> plans to upload them. Of these packages, only ocamldsort has significant
> reverse-dependencies; so if the maintainers are not reactive, the other
> packages ought to be removed from testing to allow this transition to
> complete, and catch up again on their own time.
Indeed.
> None of the above issues should be particularly difficult to resolve, and
> none warrant keeping a dummy bug open against ocaml (except that the bugs
> from point 3 should probably be filed before closing this, I guess). Also,
> if cduce is the only package broken by issue #2, I'm not sure we should hold
> up the transition for it either at this point.
The dummy bug is there also for documentation purpose to lessen the guys
filling bugs because of the transition.
> What is needed to get an ocaml upload to fix #1, and possibly #2 as well?
#1 will be solved by dropping native support for hppa, #2, i guess we kick cduce
out of testing and wait for ocaml 3.09.1, easiest that way.
Both can happen fairly quickly.
Friendly,
Sven Luther
Reply to: