Re: Status of transition and fastcluster issue
Hi Graham,
On Thu, Jun 07, 2018 at 01:04:46PM +0200, Graham Inggs wrote:
> Thanks for all your work on this transition!
You are welcome. As I said most of it was scripted and I keep on
inviting others to test those scripts. As usual the social challenge
(culminating in the attempt to restrict my freedom of speach to a free
software developer) were higher than the technical ones.
> On 07/06/2018 12:21, Andreas Tille wrote:
> > The issue is caused due to the fact that only one debhelper buildsystem
> > (R or pybuild) can be used. The last one specified wins and thus
> > "--buildsystem=R" just does nothing. I need to investigate this. All
> > other package uploads where scripted and thus easy to do but for this
> > issue I need some chunk of spare time (hopefully before, if not at the
> > weekend). Any volunteer is welcome to beat me in finding a fix.
>
> How about the following as a temporary workaround?
No need for it any more. Problem was easier to solve than expected.
R-cran-fastcluster and r-bioc-cummerbund are uploaded.
> You've always made me feel very welcome to do team uploads, and I have done
> so when needed. I'm pretty sure the commits I have made and not uploaded,
> have been minor things which I didn't feel justified an upload on their own.
:-)
In r-bioc-biovizbase I needed to force r-cran-foreign and r-cran-nnet to
(Build-)Depends. If you look at its control file it contains:
# <dh-update-R>do not delete this</dh-update-R>
r-cran-foreign,
r-cran-nnet
Please note the comment has NO EFFECT on dh-update-R yet. It just rises
awareness when inspecting the diff of changes (which I always did before
uploading). I'm also not really sure whether we should implement a
mechanism in dh-update-R to stop it from removing dependencies it does
not expect or whether we should rather find the reasons why these manual
additions are really needed.
In this case I was able to find the missing depends by asking
codesearch.d.n for
https://codesearch.debian.net/search?q=foreign+path%3ADESCRIPTION
which quickly uncovered hmisc as the responsible package for the issue.
I'd say an autopkgtest would have been helpful here to spot the issue
earlier and more easily. I filed bug #900974 with a patch.
I was not as successful in seeking the reason why r-cran-lavaan needs
a manual injection of r-cran-codetools but I stopped seeking for it
at this time. I'm not sure whether the reason is always a forgotten
dependency specification or if there is a chance that the DESCRIPTION
file is just missing the information. I'll leave this for further
investigation - the comment in d/control will inform you about issues
of this kind.
Kind regards
Andreas.
--
http://fam-tille.de
Reply to: