[R-sig-Debian] R 4.1.0 and Graphics Packages

[Johannes Ranke]

Am Freitag, 21. Mai 2021, 19:03:59 CEST schrieb Dirk Eddelbuettel:
> On 21 May 2021 at 10:56, Johannes Ranke wrote:
> | Hi all,
> | 
> | The NEWS for R 4.1.0 contain the note:
> | 
> | - The graphics engine version, R_GE_version, has been bumped to 14 and so
> | packages that provide graphics devices should be reinstalled
> | 
> | And indeed, I just ran into this and got a
> | 
> |   Graphics API version mismatch
> | 
> | error when using the tikzDevice package with my fresh CRAN backport of R
> | 4.1.0 that Dirk uploaded to experimental. The error went away after
> | reinstalling tikzDevice.
> 
> Eeek. Didn't think of that.
> 
> | For CRAN backports users, I just added a note on the Debian page
> | 
> |   https://cran.r-project.org/bin/linux/debian/#debian-bullseye-testing
> | 
> | (it will take a while for the mirrors to sync).
> | 
> | Before the r-api system was introduced, I used to set up fresh
> | repositories
> | when R introduced breaking changes, in order to avoid that an apt-get
> | upgrade breaks installed R package functionality. This one slipped by my
> | attention.
> | 
> | For the Debian R packages, I think we should find out which of the R
> | packages in the Debian archive are affected by this (r-cran-rgl,
> | r-cran-svglite, r-cran- vdiffr which embeds svglite, ggplot2, ...) and
> | add versioned Breaks.
> | 
> | Or should the r-api Version be bumped from r-api-4.0 to r-api-4.1?
> 
> I would prefer not, and don't think it is called for. But then I often
> argued for a more 'laissez-faire' approach that others (on the other list,
> i.e. debian-r).
> 
> Once the release is made, I will put R 4.1.0-* into unstable and rebuild at
> least all the packages from experimental.  Me thinks we can handle this via
> the normal bug track mechanism.

A more systematic way would be to have R 4.1.0-2 provide r-graphics-api_14 and 
only upload packages providing graphics devices that have a respective 
dependency from now on. But I don't know if it's worth the trouble.

> But the backport may have extra issue. But
> maybe your list of 'has graphics' packages is good enough?

At this point I don't really see what we can do other than spreading the word 
so the backports users can quickly address the problem by reinstalling the 
affected packages.

Cheers,

Johannes
> 
> Dirk


-- 
Johannes Ranke
Wissenschaftlicher Berater
07624 8099027
https://jrwb.de