Hi Dylan, For the record, you sent this to the wrong list, as https://release.debian.org/britney is not the domain of debian-ci but of the release team (of which I am also member). On 10-02-2020 15:52, Dylan Aïssi wrote: > Currently, the situation of CI for R packages is not optimal. > Almost all our R packages have individual scripts in d/tests/ to perform CI. > In the best cases, there is only the name of the package that change > between these files. > In the worst cases, the tests are wrongly implemented and required a > lot of manual work to fix them. > > In order to simplify the process and the maintenance, I try to > factorize all of this and move the code for testing R packages into > the package pkg-r-autopkgtest (src: dh-r) [1]. > I have uploaded this change into experimental to see which > bioconductor packages (r-bioc-*) it will break (not activated for > r-cran-* packages for now, only to test). > I had the hope to see broken packages here [2], but it seems none > tests was done. > > Did I miss something? Should I do something manually? For one thing, that page doesn't show anything if tests pass, so in that sense it can be confusing. However, if you check the corresponding .yaml file, you'll also find that britney (the migration software) didn't schedule any tests for pkg-r-autopkgtest. That means that there are no packages that depend on binaries from dh-r or have a declared test dependency on binaries from dh-r. The reason I think is that tests run via pkg-r-autopkgtest are coming from autodep8 and britney doesn't know that autodep8 depends on pkg-r-autopkgtest to run r tests. britney only schedules tests for direct (test) reverse dependencies, not for transitive (test) reverse dependencies. I hope this helps. Paul
Attachment:
signature.asc
Description: OpenPGP digital signature