[pkg-bioc] debian/ simplifications

[David Vernazobres]

On Sun, Apr 22, 2007 at 03:16:52PM +0200, Steffen Moeller wrote :
> On Saturday 21 April 2007 20:07:53 David Vernazobres wrote:
> > On Sat, Apr 21, 2007 at 09:47:38AM -0500, Dirk Eddelbuettel wrote :
> > > I made another change to r-cran.mk, and so far all packages 'just work'
> > > by only source r-cran.mk.   The newest one was to have r-cran.mk grep for
> > > 'Priority: recommended' in the DESCRIPTION file and then switch to
> > > installing into /usr/lib/R/library/ (where reommeneded packages go)
> > > rather than /usr/lib/R/site-library/
> > >
> > > Are there other modifications we could or should address from r-cran.mk?
> > > The 'repositories' idea is also implemeneted.  A few month ago I added
> > > inclusin of debian/overrides for Lintian, if found.
> >
> > cool,..
> Yip!
> 
> Something I cannot stop thinking about since  had to remove some cran/bioc 
> because of some disk space problems emerging: Could we get the LaTeX bits 
> separated and possibly some other bits implemented in order to reduce the 
> minimally required resources a bit?

Hum, reducing, yes definitely. But I did not get the LaTeX part ? What
are you referring to ?
 
> > > Also, I no longer include the basically empty postinst/postrm.
> > >
> > > Lastly, one thing I find confusing when building the few dozen r-cran-*
> > > packages I manually maintain:  Sometimes an R  package is NOT listed in
> > > Depends: yet required under pbuilder (probably due to LazyLoad or
> > > something).
> >
> > well, in cran2deb I am still adding dependencies from time to time
> > manually. It's not a long term solution, but a solution.
> 
> We should strive for a close link to the upstream developers / bioc organisers 
> to feed our manually determined dependencies back into their descriptions. We 
> could e.g. publish our list ... hm ... we have it in a database now, don't 
> we? I could come up with a script to do that.

Yep, a kind of csv format file, that we import in the db at each run.
The DB need to be improved, with a web interface (in rails),... but well
I have a really busy month in perspective, so I won't have some real
coding free time....

 
> > > How reliable is the Depends/Suggests information in DESCRIPTION
> > > for us?
> >
> > Well, very low I will say. But some packages' descriptions are very good.
> > So, it's a starting point.
> 
> I presume that upstream developers just never built their bits from scratch 
> and are even plain unaware of their package's dependencies when writing the 
> descriptions. I think that our work and the fixing of the dependencies is 
> some kind of contribution to upstream, too. We just need to redefine our work 
> a bit :-)

Yes,... but it take time to do it manually, need to come with a
solution, or a redefinition of the packaging task.

david



-- 
David VERNAZOBRES, PhD student                     | dv at uni-muenster.de
Division of Bioinformatics, University of Muenster | Schlossplatz 4
(+49)(251)8321635                                  | D48149 Muenster
http://www.uni-muenster.de/Evolution.ebb/          | Germany