Re: debian: blands: mathematics-dev: libmpfi libtestu01: git-format-patch
Hi Jerome,
On Fri, May 22, 2015 at 02:51:21PM +0200, Jerome BENOIT wrote:
> >> On the other hand, libnauty2-dev Provides libnauty-dev, which may
> >> be also provided by libnauty3-dev and so forth: my guess is that it
> >> would be more appropriate that the latest concrete package that
> >> Provides a virtual package.
> >
> > I agree that this sounds catchy but as you state yourself more than
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> > one package might provide a virtual package. If this is the case
> > what exact metainformation should be presented on the tasks page?
> > It's not always that simple to say that with the highest version
> > number.
>
> Let say that there are two kinds of virtual packages:
> 1] the ones that can be provided by different source packages, a good example might be pdf-viewer;
> 2] the ones that are provided by the same source package along time, good examples are library
> packages for which the concrete packages only reflect the increasing of the SO name.
>
> Your assertion is correct for the first category, but obviously not for the second one.
I'm perfectly aware of this but there is no really safe way (besides
wild guessing from the package name but that's not safe) to
differentiate.
> So it has some sense to only list real packages.
At least it is the simplest way to implement as long as we have no
better means to provide some robust metadata about virtual packages.
This idea is somewhere deep down on my todo list - so deep down that it
will probably never implemented (at least by me). This does not mean
that it is not interesting - there are just other way more challenging
things to be done. :-)
> > Ther is no such package and I also have not seen it in the list of
> > prospective packages in VCS. So what exatly do you expect to be
> > displayed? The log file at
> >
> > http://blends.debian.org/_logs/debian-science.log
> >
> > says:
> >
> > ERROR - blendstasktools.py (1847): Package libtestu01-dev in task
> > mathematics-dev neither in pool nor new and is lacking homepage and
> > description - ignored
>
> Indeed: it is an ITP present at Alioth:
ITP's are not parsed since they are not really parseable (since ITPers
sometimes use freetext - if you don't believe, feel free to provide an
ITP parser for UDD. It would be really welcome.)
> http://anonscm.debian.org/cgit/debian-science/packages/testu01.git
The commit logs claim "Wed May 20 11:19:31 2015 +0200" and depending
when you pushed this to Git. From this point in time on you need to
wait about 24h for cron jobs.
Ahh, and the d/control says:
Package: libtestu01-0-dev
Provides: libtestu01-dev
So guess what will be displayed on the web page. I commited
$ git diff HEAD^
diff --git a/tasks/mathematics-dev b/tasks/mathematics-dev
index 1d25936..048ade6 100644
--- a/tasks/mathematics-dev
+++ b/tasks/mathematics-dev
@@ -56,7 +56,7 @@ Suggests: libitpp-dev
Depends: libnauty2-dev
-Suggests: libtestu01-dev
+Suggests: libtestu01-0-dev
Suggests: liblrs-dev
Hope this clarifies the issue
Andreas.
--
http://fam-tille.de
Reply to: