Re: Veusz in Debian Science?
Hi Jeremy,
On Sun, Jun 24, 2018 at 01:04:57PM +0200, Jeremy Sanders wrote:
> Thanks very much for looking at this. I've put my changes in
> https://salsa.debian.org/jeremysanders-guest/veusz as I can't modify the
> original in debian-science.
I've added you to Debian Science team. Please push there. I'd prefer
to avoid having too many clones hanging around.
> On 24/06/18 00:05, Andreas Tille wrote:
> > What exactly is the tricky part here?
>
> I think it's just lack of time for people to review the package. I tried a
> couple of times.
OK, I hope we can solve this. :-)
> > You might like to commit your debian/tests dir and let others have a
> > look. Possibly you just need xvfb as Depends in the
> > debian/tests/control file. Its hard to guess how to fix the test if
> > you do not publish your attempt.
>
> The rules has a test section in "override_dh_auto_test". This overrides the
> pybuild tests, which would normally run a python module test. These work in
> a pbuilder environment, just not in the salsa continuous integration. Xvfb
> is a requirement for the build.
>
> It fails in CI with
> ----
> This application failed to start because it could not find or load the Qt
> platform plugin "xcb"
> in "".
> ----
> But seems to work in pbuilder for me.
>
> Should I move the tests to debian/tests instead?
The build time test should remain where these are. I assumed you were
talking about autopkgtest which is something else than the salsa CI test
(I have not yet dealt with this so I can not comment on this).
> > One issue is worth discussing: You are adding a lot of -dbg packages to
> > Build-Depends. That's something I've never seen and which should be
> > removed if you have no good explanation why this is done. I'd also
> > recommend to drop
> >
> > $ grep "Package.*-dbg" debian/control
> > Package: python-veusz.helpers-dbg
> > Package: python3-veusz.helpers-dbg
>
> > If you do not add these manually the build process adds *-dbgsym
> > packages automatically. So there is no real point to add this manually.
> > Since I do not know your intention about this I did not patched this
> > and leave this as open question for the sponsering process.
>
> The old package predated the automated debug symbols and I didn't think
> about removing this. I've now removed the dbg packages and requirements in
> the build (see link above).
Good. Please push to the Debian Science repository and I'll have a
look.
> > Another item to think about: Do we *really* need the packages
> > containing Python 2 modules? If not it seems sensible to simply
> > drop python2-* packages in general since Python2 is end of life.
>
> I agree it would be nice to drop this as it simplifies things. I've made a
> py3-only branch above which does this. It drops the python-veusz and
> veusz-data subpackages, merging the data into python3-veusz. Perhaps the
> python3-veusz and veusz subpackages could also be merged (as is done in the
> current Python 2 only version in Debian).
Its your decision since you have the proper insight. If there might be
use case that the python3-veusz module would be used independently from
veusz it makes perfectly sense to modularise the packaging.
> However, I'm not sure whether some people might be using the python2 module
> provided in the current package. A lot of people in science seem to be stuck
> on python2.
That's true and I do not want to create pressure. However, I wanted to
make sure that you *really* intend to create a *new* package that
supports python2. At some point in time that will definitely be
removed. So please make soure you are really sure you want to do extra
work, which is not future proof for the only reason that you assume
scientist would be extremely conservative.
> I admit (as upstream), that there probably aren't too many
> people using the program for plotting without the GUI.
>
> The old veusz package provided a python-veusz. Would something need to be
> added to say this no longer exists?
I can not find such a package in the package pool. Am I missing
something?
> Thanks very much for you really helpful feedback
Kind regards
Andreas.
--
http://fam-tille.de
Reply to: