[R-sig-Debian] packages in cran2deb and Debian

Ross Boylan ross at biostat.ucsf.edu
Fri Feb 12 03:28:47 CET 2010


Thanks for some good tips.  More detailed responses below:
On Fri, 2010-02-12 at 00:23 +0000, Liviu Andronic wrote:
> On Fri, Feb 12, 2010 at 12:07 AM, Ross Boylan <ross at biostat.ucsf.edu> wrote:
> > Some packages are available in cran2deb and Debian.  After adding
> > cran2deb to my sources.list it seems the cran2deb ones are favored
> > because X.Y-ZcranN is "more recent" than X.Y-Z.
> >
> Hmm, my experience has been different: Debian packages were preferred
> over cran2deb. Can you give a specific example that bugs you?
I gave an example in the next paragraph of the original: r-cran-rmpi.

The example in your email [1] is a situation in which the base version
is higher in Debian, roughly X.Y-(Z+1) vs cran2deb at X.Y-Z.
> 
> 
> > In the case of r-cran-rmpi the effect of this was to cause aptitude to
> > want to uninstall all my mpi stuff.  Apparently the automatically
> > generated package doesn't have mpi as a dependency.
> >
> Wouldn't it be sufficient to set mpi as "manually installed", as
> opposed to automatically, as a dependency?
That should work; I may need to do that with a couple of mpi packages.
The debian 0.5.8-1 r-cran-rmpi has
Depends: libc6 (>= 2.2.5), libopenmpi1.3, r-base-core (>= 2.10.1),
mpi-default-bin

> 
> 
> > Do people have any general advice about how to handle this situation?
> >
> You might want to try an approach similar to this [1].
> [1] https://stat.ethz.ch/pipermail/r-sig-debian/2009-December/000970.html
> 
Thanks; that's very handy.  I didn't realize that cran2deb had a
distinct name.
> 
> > There probably is no one rule for what to do.  The official packages are
> > more carefully done, but the automatic packages may be more recent.  At
> >
> I am curious, what makes you think so?
I made several assertions and don't know to which you refer.

Ross



More information about the R-SIG-Debian mailing list