Re: Seems we need a transition for r-base instead of lots of single bugs against single packages (Was: Bug#1039645: r-cran-epi: autopkgtest failure with r-base (4.3.1-1))
Séb,
On 28 June 2023 at 17:07, Sébastien Villemot wrote:
| The question to answer to determine whether a transition is needed is
| the following: are CRAN packages compiled against R 4.2 binary-
| compatible with R 4.3? Or do they need a rebuild to work with 4.3?
They are. There is no binary API transition, just as there wasn't from R 4.0
to 4.1, or from 4.1 to 4.2. (And as before they cheated _a little_ and
changes the graphics API so the likes of agg, ggplot2, ... need a rebuild.)
| If a rebuild is needed, then this means that we have to go through a
| proper transition (involving the Release Team).
|
| I note that the r-base-core package Provides the virtual package r-api-
| 4.0 in both versions 4.2 and 4.3. So in essence, the fact that this
| Provides field has not been updated indicates that a transition is not
| needed.
Correct, and fully agreed (now with my 'r-base maintainer' hat on).
If dply / tidyverse break other packages (as Nilesh implied) then maybe that
package needs a transition. But I don't see value in pointing fingers at
r-base when it demonstrably does not have an issue.
| If it turns out that a transition is needed, then the first step is to
| change the Provides field of r-base-core to something like r-api-4.3.
We do not need to. I should email Paul Murrell and ask him "to predict the
future" and see if the graphics API will regularly change and we could add an
r-graphics-api-* tag in the future.
It is pretty far from R 4.4.*. If I were a betting man I'd say chances for
an API break increase as times goes on (we had one maybe every five-ish
years) but then again there have not been any great changes inside R. I run
r-devel (the branch forming the next release) just as many other (R-) package
developers do and will know soon enough coming up to R 4.4.0 next April.
But there is no current fire to extenguish here.
Amicalement, Dirk
--
dirk.eddelbuettel.com | @eddelbuettel | edd@debian.org
Reply to: