On Mon, Apr 28, 2025 at 03:36:54PM +0100, Rebecca N. Palmer wrote: > Thanks, and I agree with the above analysis: the buildd logs confirm that > r-cran-svglite was binNMUd recently on armhf only, and debci says > r-cran-svglite/armhf is currently broken in testing but working in testing + > R 4.5.0, > https://ci.debian.net/packages/r/r-cran-svglite/testing/armhf/ > > r-base 4.5 is expected to migrate along with the fixes for some other bugs: > https://lists.debian.org/debian-r/2025/04/msg00016.html
> Would it help to upload a new r-cran-svglite that explicitly depends on > r-base-core (>= 4.5.0~) ? That's what I'm doing with r-cran-rlang, the > other package known to break when built in R 4.5 and run in R 4.4, but > r-cran-rlang was caught (by debci) before it migrated, while r-cran-svglite > is already broken in testing. Maybe. If r-base 4.5 migrates soon, obviously the problem is gone in trixie. To avoid broken half-upgrades, r-cran-svglite should explicitly depend on r-base-core with a versioned dependency. Some solution also needs to be found going forward: generally Debian expects packages to be binNMU-safe, and binNMUs can generally migrate on their own. r-cran-svglite is currently not binNMU-safe, apparently. Probably none of the rdeps of r-can-cpp11 are binNMU-safe? Chris
