On 13 March 2026 at 09:33, Charles Plessy wrote: | Dirk, the list at the end includes your architecture-dependent packages that | directly or transitively depend, recommend or suggest r-cran-utf8. Can | you architecture-restrict them?
Le Fri, Mar 13, 2026 at 07:39:08AM -0500, Dirk Eddelbuettel a écrit :
That list seems, at glance, _way_ too long. It contains packages for which I am upstream and hence fairly certain I do not use package 'utf8'. It also contains a few well-established old packages for which I do not see how 'utf8' could creep in. Can you detail what you condition was here? Full recursive descent, and did you include suggested packages? If I run `apt-cache rdepends r-cran-utf8` I get one hit: `r-cran-pillar`.
Hi Dirk, Here is the code that I used. It is low quality and ends more like a scratch space than a real script. Please feel free to ask me to improve it before you take more time reading it. But if you spot bugs I would be very grateful to hear about them! https://salsa.debian.org/r-pkg-team/maintenance-utilities/-/blob/be533de817771558160865103d4aef7e3158acf6/rdeps.R The parsing of the dependency graph is fully recursive including Suggests, which in the Debian package are generated from the DESCRIPTION file. This is necessary because a package in Suggests may be necessary to run the package's test suite, which is what we run in the Debian package's autopkgtests. But obviously the impact on the dependency graph is strong. For instance, arch-restricting anything that r-cran-testthat transitively depends or build-depends on triggers the need to arch-restrict every package using r-cran-testthat in its test suite, because we run autopkgtests for every package that provide a test suite. (Because it allows us to quality control loong64, ppc64el and riscv64, which we support but on which CRAN/Bioconductor and many of their upstreams do not run test). I am not sure if that is the case for r-cran-utf8, but I hope that it makes the point that the dependency hairball becomes large very quickly… Have a nice day, Charles -- Charles Plessy Nagahama, Yomitan, Okinawa, Japan Debian Med packaging team http://www.debian.org/devel/debian-med Tooting from work, https://fediscience.org/@charles_plessy Tooting from home, https://framapiaf.org/@charles_plessy
