[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Packaging and sponsoring of ompl (Was: r3898 - in /projects/science/trunk/debian-science/tasks: robotics robotics-dev)



A Dimecres, 30 d'octubre de 2013, Andreas Tille va escriure:
> Hi Leopold,
> 
> I noticed that you tried to follow my Sponsering of Blends procedure[1]
> and injected some data to the robotics tasks.  That's great and it is
> really welcome.  I always like to teach at these examples who to do this
> properly and efficiently.  So let me comment your changes first:

Ok,


[...]
> >  
> > +Package: ompl
> > +Homepage: http://ompl.kavrakilab.org
> > +License: BSD-3-clause
> > +Responsible: Leopold Palomo-Avellaneda <leo@alaxarxa.net>
> > +WNPP: 706133
> > +Pkg-Description: Sampling-based motion planning library
> > + Consists of a set of sampling-based motion planning
> > + algorithms. The content of the library is limited to these algorithms,
> > + which means there is no environment specification, no collision 
> > + detection or visualization. The library is designed so it can be easily
> > + integrated into systems that provide the additional needed components. 
> > +
> 
> Comments:
> 
>  1. You never should use the "Package" field when specifying package
>     dependencies in a task file.  It is always Depends / Recommends /
>     Suggests (see [2]).
>
>  2. The specified dependency should be a binary package name which
>     correlates to a binary package name in your debian/control file.
>     You specified the source package name 'ompl' - but you can not
>     depend from a source package.  I changed this to ompl-demos and
>     used "Suggests" (since this sounded appropriate to me - feel free
>     to change this to "Depends").

Yes, I'm sorry. I did a mistake here and do a cut and paste from control file 
:-(  and not pay attention to that. I'm sorry to waste your time in this 
stupid mistake. 

>  3. You have specified several fields in addition to the Suggests
>     field.  This is not necessary any more if you have injected the
>     packaging into the team VCS.  It is probably not properly
>     documented in [2] (feel free to send a patch) but all these data
>     are obtained automatically to safe your time in maintaining the
>     tasks file.  To demonstrate this I removed those fields that were
>     overriden anyway by the packaging content.  The only field I left
>     untouched was WNPP.  The reason is a problem in your
>     debian/changelog file (see below for the sponsoring comments).

Ok, good to know it.


> > +
> > +Package: libompl-dev
> > +Homepage: http://ompl.kavrakilab.org
> > +License: BSD-3-clause
> > +Responsible: Leopold Palomo-Avellaneda <leo@alaxarxa.net>
> > +WNPP: 706133
> > +Pkg-Description: Sampling-based motion planning library development files
> > + Consists of a set of sampling-based motion planning
> > + algorithms. The content of the library is limited to these algorithms,
> > + which means there is no environment specification, no collision 
> > + detection or visualization. The library is designed so it can be easily
> > + integrated into systems that provide the additional needed components.
>  
> Comments:
> 
>  1. Same as above - I did a s/Package/Depends/.
>  2. Here you correctly used the binary package name libompl-dev that exists
>     in your packaging - that's fine.
>  3. Same as above - I just left the WNPP bug.

Ok.

 
> Now my comments to your packaging[3]
> 
> debian/changelog:
>   - You are trying to close the ITP bug in an "historic" changelog
>     paragraph (0.12.2-Source-1).  This does not work.  Bugs can only
>     be closed in recent changelog entries (here is also the answer why
>     this information is not parsed by the Blends tools - they are only
>     regarding the latest changelog entry).
>   - Moreover you are mentioning unstable as target distribution of
>     those past entries.  However, there was never any upload to unstable
>     at all but these are rather all "UNRELEASED".  Generally speaking:
>     It does not make much sense to maintain a debian/changelog before
>     the first version is really inside Debian.  The history of the
>     packaging can be obtained from the Git log.  => Please drop old
>     entries and leave only
>       * Initial release (Closes: #706133)
>     in the current entry.

Ok, but from the packager point of view, it helps you to see how many packages 
you have worked before. I can delete whatever you think convenient, but I 
began to package it with one version of the library and changed it. So, I 
tried to document it. Can I leave the entries but change it from unstable to 
UNRELEASED?


> debian/README.Debian:
>   - Please delete this file.  The information is not relevant and just
>     contained in the metadata of debian/copyright.
> 
> debian/README.source:
>   - Please delete this empty file as well.
> 

ok, done.


> debian/control:
>   - Please choose "Priority: extra".  I guess I did mention it several
>     times here on this list that extra is for *-dbg packages or things
>     like this.

ok, done. Changed for optional.

>   - Pre-Depends: dpkg (>= 1.15.6~)
>     What is the rationale for this?  Usually it should not be needed.
>   - Please give
>        cme fix dpkg-control
>     a try.  I guess you will like the result.

First of all, I must admit that I don't understand what are you asking me with 
the dpkg-control. 

I put this field because I checked the generated packages with "lintian -EviIL 
+pedantic". And then, I got a lintian warning like:

N:    The package contains a preinst maintainer script that uses
N:    dpkg-maintscript-helper but does not declare a versoned pre-dependency
N:    on dpkg (>= 1.15.7.2~) that provides that script.


so, I do that.


> pristine-tar branch is missing:
>   - If you imoprt the orig.tar.gz please use
>       git import-orig --pristine-tar /path/to/package_version.orig.tar.gz
>     to make sure your sponsor / team member can easily recreate a byte
>     identical tarball.

Ok, the problem I found is that the official version got some problems that I 
report to upstream and changed it in mercurial. Ti me is ok the version with 
the commit hash, so I imported that one. But it seems I forget some step. I 
solve it soon.


> lintian warnings:
>   - please check `lintian -I -i` and consider to fix some lintian
>     nitpicking issues.
> 
> So far for the sponsoring request.

Thanks for your time.

Leo




-- 
--
Linux User 152692
Catalonia

Attachment: signature.asc
Description: This is a digitally signed message part.


Reply to: