Hi, On Mon, 05 Oct 2026 at 22:10, Ricardo Wurmus <[email protected]> wrote:
> Yes, I use > > ./pre-inst-env guix refresh -t cran -u > > and later > > ./pre-inst-env guix refresh -t bioconductor -u > > A lot of time is spent on cleaning up the diff, e.g. to remove > excessive native inputs and to cut dependency cycles. Sometimes > this requires disabling tests. I also record properties to > hopefully simplify the next update. A very small number of > package updates I revert right away, because I know that I can't > build bundled artifacts from source. I can confirm the time-consuming task is to walk through the diff and clean up. Although the R ecosystem is reliable, sometimes the importer introduces “weird” packages. And it needs a couple of trial-errors to spot the usual suspects. > I often run "make" to see what packages I need to add or which > references I need to remove. IMHO, it also requires to try to build some packages for detecting the cycles and thus add/remove the correct ones. > More info in Message-ID: > <[email protected]>. For updates to R itself (= r-minimal and r-with-tests) I build the hidden r-with-tests once and r-minimal with "--rounds=3 -K" to see if there are reproducibility problems. […] […] When this works I push to r-team and let the build farm build out all R packages. Before I did not feel confident, before starting, I did: for p in $(cat gnu/packages/cran.scm | grep define-public | grep -E r\- | sort); do guix build $p -q done to have 1. the list of broken packages and 2. download beforehand all the dependencies. And I also looped with: guix build etc/teams/r/r-manifest.scm -q then ’guix build r-foo` where r-foo is the one which failed. This helped me to resolve the cycles. Cheers, simon
