[R-sig-Debian] Installing R 3.5 on Ubuntu bionic stopped(?) working

[Dirk Eddelbuettel]

On 22 November 2019 at 16:51, Tim Head wrote:
| On Fri, Nov 22, 2019 at 2:15 PM Dirk Eddelbuettel <edd using debian.org> wrote:
| > | With repo2docker we are in a slightly weird position where we want to
| > | enable container image builds that worked 6months or 12months ago to also
| > | work today. This means if a user built a container with R 3.5 in the past
| >
| > You could maybe base that on the versioned Rocker container (the "r-ver"
| > stack) which is also tagged to corresponding dates at MRAN.  No support for
| > binary packages though.
| >
| 
| The drawback of using rocker or other base images is that we currently have
| the ability to compose several different ways of installing software (You
| want R and Python? No problem!) which gets lost if we start using different

All Rocker containers use standard Debian (or Ubuntu, it depends) so you can
just say 'apt-get install python3-scipy' (to pick one example).  That is the
whole point of using a rich distro with over 20,000 packages.

Rocker itself has hybrid containers as we also ship tensorflow etc pp.

| base images. Or at least we have to be careful about what base images we
| use so that they all share Ubuntu bionic as the base.

Doing it consistently through time is a hard task.  As I said earlier, we (as
in: these add-on repos giving you R) generally do not attempt it.
| 
| As we are only after the R executables themselves maybe using conda to
| install them or compiling from source is a way to explore.

I have very little experience with conda. All is see, usually, is people
mixing and matching and ending up in tears. But you can always try.

Cheers, Dirk

-- 
http://dirk.eddelbuettel.com | @eddelbuettel | edd using debian.org