Re: Octave 6 transition
* Sébastien Villemot <sebastien@debian.org> [2020-12-01 17:47]:
I have tried to recompile all reverse-dependencies of octave.
Thanks for this work, Sébatien. Much appreciated.
It turns out that 11 packages maintained by the DOG fail to compile
against octave 6. I’ve already opened bug reports against those. The
full list is: strings, image², msh, nurbs, quaternion, struct,
sparsersb, stk, secs2d, interval, ltfat.
Have you already tried to fix one of those packages? Should we share the
porting work among us? For instance, some of them, like octave-strings,
seem trivial to fix and I can take care of it.
There are also 2 packages not maintained by the DOG that also failed to
compile: plplot and mathgl.
Those two packages actually use swig for generating their octave
bindings, and it turns out that swig currently does not work with
octave 6, as reported upstream by Rafael.³
@Rafael: I see that you decided to remove the octave binding from
plplot (in the latest experimental upload), which is indeed a way of
fixing the problem. I therefore guess that you would recommend the same
approach for mathgl?
The mathgl source package does build a file mathgl.oct using SWIG.
However, this file is not installed anywhere. I think we should simply
drop the build-dependency on liboctave-dev.
Once we have decided our strategy for dealing with these two packages,
I will request a transition slot from the Release Team.
I would say: go ahead! I hope that the large number of collisions will
not compromise our transition.
Of course, they might be reluctant to start the transition before the
freeze because of the high number of packages that are not yet ready.
On the other hand, most of those are under our control, so I guess we
can reasonably argue that we will be able to manage this swiftly (with
hopefully some help from upstream).
I think it is feasible. If some of the packages are lost they can be
added later, even via backports. It will not be a big deal.
Best,
Rafael
Reply to: