[pkg-bioc] i386 builds of CRAN

[Dirk Eddelbuettel]

On 5 June 2005 at 21:47, Steffen Moeller wrote:
| Hi all,
| 
| many thanks to Dirk for forwarding my initial work on i386 CRAN and its 
| positive reception. I did not do much, I must admit, I just ran the 
| existing code, so all the applause should reflect to the ones 
| responsible for that. Actually I found Egon's Wiki pages very helpful 
| and did some editing to these, please check it out at 
| http://wiki.debian.net/?AliothPkgBioc. If somebody knew how to tell 
| alioth to make the Wiki pages our project's home, then I would very much 

Matt or Rafael may know ...

| appreciate it. You also find something like a summary of your past 
| emails there with the troublemaking packages.
| 
| If I perceive this right, then we are now running into problems because 
| of novel-built packages being in conflict with existing R packages. I 
| think we should refrain from changing the script but rather change the 
| Debian packages whenever possible. E.g., should not r-cran-gregmisc 
| depend on the gtool, gdata and all the other greg packages instread? 

Ah, so you are volunteering to take over r-cran-gregmisc from you, to create
four new packages and to get them into Debian?  Good.  I'm not volunteering
to be your sponsor though :)

Kidding aside, Greg and I have agreed to keep r-cran-gregmisc the package
bundle for now.  So I would disagree with Steffen and suggest that yes,
cran2deb will need to adapt when the package sets change.

| There are dependencies on gdata which otherwise could not be resolved. I 

Not, we need to add the meta-knowledge that gdata is in the bundle gregmisc.
See how we're dealing with components of the VR bundle. Not that different,
really.

| think we make ourselves unhappier in the long end for any articificial 
| barrier between the upstream repository and the one of Debian.
| 
| Something else I thought of is
| 
| * mailing all the upstream authors to improve on the description of 
| their packages to get titles and descriptions nicer
| * to come up with a semiautomatic schema to get this done within the 
| script, analogous to the changelogs.

That is probably more feasible as it does not require cooperation from all
the package authors.

| As I said before, I rather make upstream better than the Debian 
| packages. Though, for my BioConductor packages I already invested quite 
| some efforts into a separation of dependencies and build-dependencies, 
| recommends, retrieval of email addresses etc that I would not like 
| losing too much. How would you think about implementing an optional 
| control and copyright file for every package, falling back to the 
| template if not existing? Should not the changelogs become part of our 
| repository on Alioth?

Sure.

Anybody have any clever ideas about the "databasing" of changelogs, control
files etc pp?

Dirk

-- 
Statistics: The (futile) attempt to offer certainty about uncertainty.
         -- Roger Koenker, 'Dictionary of Received Ideas of Statistics'