Are there Debian/Ubuntu packages of the betaversions of R-2.9.0

Le dim. 05 avr. ? 21:53, Paul Leo a ?crit :

Thankyou Dirk,
That was a great summary I think I understand the philosophy now. I will
go move forward with compiling the source code with caution.

(Phew, shouldn't have been away today! Sorry folks, but I'll sort of reply
to many messages at once.)

First, Paul, if you want to compile R on Ubuntu from the Debian sources,
you'll want to use or, at least, have a look at my build scripts located at:

? ? ? ?https://vgoulet.act.ulaval.ca/svn/R/cran-ubuntu/

(This is the public svn repository Dirk was referring to.) The important one
is build-r-base.sh; the script makes all the modifications to the Debian
sources needed to compile on Ubuntu releases from Dapper to Intrepid and
then launches the build process.

Regarding the distribution of binary packages of beta versions of R for
Ubuntu, I fully agree with Michael.

I also think Doug's proposal to distribute patched sources is a nice
compromise. That said, I'm not too sure which files I should make available.
Dirk may confirm that the needed files would be, for example:

? ? ? ?r-base_2.9.0~20090327-1intrepid0.diff.gz
? ? ? ?r-base_2.9.0~20090327-1intrepid0.dsc
? ? ? ?r-base_2.9.0~20090327-1intrepid0_i386.changes
? ? ? ?r-base_2.9.0~20090327.orig.tar.gz

I never noticed at which stage of the build process the first three files
are created. Is it possible to create them without fully compiling the
package?

Cheers,

Vincent

Cheers
Paul

-----Original Message-----
From: Dirk Eddelbuettel <edd at debian.org>
To: Paul Leo <p.leo at uq.edu.au>
Cc: Dirk Eddelbuettel <edd at debian.org>, Michael Rutter <mar36 at psu.edu>,
Douglas Bates <bates at stat.wisc.edu>, r-sig-debian
<r-sig-debian at r-project.org>
Subject: Re: [R-sig-Debian] Are there Debian/Ubuntu packages of the
betaversions of R-2.9.0
Date: Sun, 05 Apr 2009 19:29:26 -0500


On 6 April 2009 at 10:04, Paul Leo wrote:
| I not sure what the problem is but I'm a new ubuntu user , so forgive my
| ignorance.. Why can I not have a stable and development version of R
| installed and go between them as I wish? The library path includes the R
| version number; can the binaries not go to /usr/local/R2.9 ? Guess I am
| not sure what the problem is.

/usr/local/ is yours, /usr is for the packaging system. ?That is a strict
boundary that is never crossed.

Under /usr, the current package lives in one place and does therefore not
allow multiple versions. ?In theory, what you suggest is perfectly
feasible. Emacs does it, and many other apps allow to have multiple
versions
installed. I once engineered the same rewrite to have several Octave
versions
installed. ?It is possible --- but it also requires a more-or-less
complete
rewrite of the package infrastructire, and then tests, of all associated
scripts and files. I will have to leave that for other volunteers as I
don;t
have the time to drive that change.

So in the short run, go with your own builds in /usr/local if you need
several versions at the same time.

| Yes all the Bioconductor development version is on R 2.9. There are many
| thousands of bioconductor users who stay on the development version of
| R . This is where I prefer to be, this was easy with windows, ?I could
| install as many versions of R as I like 2.7, 2.8.1, 2.9 .... The
| bioconductor libraries are smart enough to know the R version you are
| running ?and use the correct repositories for their libraries. But that
| is the dark ages I don't want to return to!
|
| The option of a virtual install seems unnecessary surely ? I am familiar
| with that but I don't want to chop up my memory of each instance as I
| need it all for some jobs.
|
| As you can tell this is my first foray into this topic and I certainly
| don't to put any more demands on you time. I guess I did not realise it
| was either one or the other 2.8.1 or 2.9.0 ? I did not anticipate that.
| Is the reason just trivial , same excutable names and default install
| path or is there more to it?
|
| I do not mind compiling from source etc; actually part of joining this
| email group was to get a sense of what was the "standard approach" that
| is used for ubuntu and development versions of R. I also wanted to get a
| feel for ess compatibility with the development version of R.
|
| So summary
| 1) I can't just install the stable and development versions ?from deb
| files and have them play nice ?

Sure. But not from .deb files as you can have exactly one R binary at the
same time.

[ But you can use that for the default R and then keep a bleeding edge one
in
/usr/local too. ]

| 2) I can't compile from source code without secret sauce?

Sure you can. It just won't be as integrated as the pre-packaged version.

| I don't agree there should be a test for who can use the ubuntu
| development version , it is development and unstable as so no one can
| reasonable expect it to be foolproof. General direction are all that
| anyone could reasonably expect.

Most of us do not perceive Ubuntu to be the 'developer distro'. ?It really
is
easy enough to split this over several different machines too. ?I.e. under
my
desk is an Ubuntu box driving the desktop and a headless Debian box
supporting additional development work. ?There are many possible
approaches.
What we attempt to provide for Ubuntu is a mainstream package to satisfy
users of the _currently released_ R versions.

Everybody is invited (and welcone) to start their own project for packages
of
a development version. ?But I think you just cannpt expect already-busy
volunteers like Vincent, Michael or myself to do it for you.

Cheers, Dirk

| Cheers
| Paul
|
| -----Original Message-----
| From: Dirk Eddelbuettel <edd at debian.org>
| To: Michael Rutter <mar36 at psu.edu>
| Cc: Paul Leo <p.leo at uq.edu.au>, Douglas Bates <bates at stat.wisc.edu>,
| r-sig-debian <r-sig-debian at r-project.org>
| Subject: Re: [R-sig-Debian] Are there Debian/Ubuntu packages of the
| betaversions of R-2.9.0
| Date: Sun, 05 Apr 2009 08:18:09 -0500
|
|
| On 5 April 2009 at 08:36, Michael Rutter wrote:
| | If we make the beta debs of R available for Ubuntu, I think they
| | should be in a separate repository, as not to be pushed to people
| | uncomfortable with beta software. ?If we went down this path, we could
| | also limit ourselves to the latest Ubuntu release as well as the
| | latest long term service release versions of Ubuntu.
| |
| | I do not use bioconductor, so I have to ask how many packages are
| | built in the approximately one month period the R beta is available
| | for public release? ? I visited the web site, and there is a devel
| | branch. ?Are all of those packages built against R 2.9?
| |
| | My personal opinion is that we should not follow the google definition
| | of beta, and reserve the tag for testing purposes. ?If you want to
| | test the beta version, there should be a minimum amount of "expertise"
| | required for entry. ?For Ubuntu, having to compile from source could
| | be that bar. ?We could supply directions on how to build from the deb
| | source packages. ?It would probably need a script that makes the
| | corrections needed to compile under Ubuntu.
|
| I agree. I think we should not push down 2.9.0 on all Ubuntu users.
|
| Paul, given that Debian does have the 2.9.0 deb files ready
| for the picking, you could also try a chroot or virtual machine running
| Debian unstable. ?By using a virtual machine (like virtualbox, very easy
to
| set up in stock Ubuntu) you get a way to have the bleeding edge software
| running on your system without affecting the general stability of your
| system.
|
| Otherwise, take the Debian .deb source and ask us for help with local
package
| builds based on those source (as e.g. right now where Debian's package
has a
| debian/rules which uses dh_prep which you need to turn back to dh_clean
etc
| pp). ?I sometimes do that at work (eg to get current Debian MPI packages
onto
| Ubuntu etc).
|
| Dirk
|


? ? ? ?[[alternative HTML version deleted]]