Re: r-base once again blocked by i386 autopkg tests
Le Tue, Nov 18, 2025 at 04:35:57PM -0600, Dirk Eddelbuettel a écrit :
>
> Six of your packages are holding r-base from migrating to testing as can be
> seen on the 'excuses' page for r-base:
>
> https://qa.debian.org/excuses.php?package=r-base
>
> All failing tests are on i386, a platform which upstream (both R and CRAN,
> CC'ed here) stopped checking years ago as it is less relevant these days.
> Could we maybe arrive at a situation where you do not enable autpkg tests on
> that platform?
Hi Dirk,
Short answer: I will ask today to the release team to remove these packages
from Testing so that r-base can migrate.
Longer answer: We really want to remove all the r-cran-* and r-bioc-* packages
maintained by the "Debian R Packages Maintainers" team from the non-64-bit
non-little-endian architectures, but I am struggling to:
- find a reliable method that predicts what other packages will need to be
removed as a consequence, to show that there is no blocker.
- find the shortest list of packages on which I have to open RC bugs to remove
them transiently from Testing.
Once this is done, I will open the RC bugs to clear Testing, and then
mass-upload the packages with architecture limitations by depending on the
architecture-is-64-bit and architecture-is-little-endian pseudo-packages.
For this I will probably need to wait for the approval of the Release team
as it may cause delays in ongoing transitions. I hope to to the pending
Bioconductor transition at this time.
This approach does not change the Architecture field and therefore it empowers
users and porters to override our choice, and saves us from re-uploading all
the packages in order to support new 64-bit little-endian release architctures
when they are added. But the drawback is that Testing must be cleared
otherwise the new packages will never migrate.
It is very heavy work for little reward for our users, but I think that this
is the best compromise we can make in front of the pressure that Debian puts
on us to support architectures before or after they have a clear user base
for a specific subset of packages like the r-cran/bioconductor ones...
Have a nice day,
Charles
--
Charles Plessy Nagahama, Yomitan, Okinawa, Japan
Debian Med packaging team http://www.debian.org/devel/debian-med
Tooting from home https://framapiaf.org/@charles_plessy
- You do not have my permission to use this email to train an AI -
Reply to: