[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

How to find out why britney is trying to migrate unexpected (and unmigratable) packages?



We recently added some Breaks: to r-base, and britney is now trying to migrate a larger set of packages than I intended - including many packages that *can't* migrate now because they are not in testing at all.

What are good ways to investigate further (that don't involve waiting 10 days, because if anything needs fixing then we don't have that long)?

Which version of https://salsa.debian.org/release-team/britney2 is the currently running britney? head-of-master since there aren't any tags??

Packages I think we need:
src:r-base,src:r-bioc-bioccheck,src:r-bioc-pwalign,src:r-cran-argparser,src:r-cran-hdf5r,src:r-cran-plotly,src:r-cran-rcmdcheck,src:r-cran-rlang,src:r-cran-s2,src:r-cran-sf,src:r-cran-survminer,src:r-cran-svglite,src:r-cran-testthat,src:r-cran-timechange,src:r-cran-units

Packages britney actually passed to debci (see e.g. https://ci.debian.net/packages/r/r-cran-testthat/testing/amd64/60313000/ ):
--pin-packages=unstable=src:r-base,src:r-bioc-bioccheck,src:r-bioc-pwalign,src:r-cran-argparser,src:r-cran-bigmemory,src:r-cran-broom.helpers,src:r-cran-cards,src:r-cran-dendextend,src:r-cran-effectsize,src:r-cran-fitcoach,src:r-cran-hdf5r,src:r-cran-heatmaply,src:r-cran-mockr,src:r-cran-performance,src:r-cran-plotly,src:r-cran-rcmdcheck,src:r-cran-rlang,src:r-cran-rpact,src:r-cran-s2,src:r-cran-sf,src:r-cran-sjplot,src:r-cran-sjstats,src:r-cran-survminer,src:r-cran-svglite,src:r-cran-testthat,src:r-cran-tikzdevice,src:r-cran-timechange,src:r-cran-units,src:r-cran-zelig

All the packages I wasn't expecting are not in testing at all, except src:r-cran-broom.helpers, which can't migrate because the new version depends on such packages. (This is fine in itself, but I wasn't expecting it to block other packages; there appear to be no (Build-)Depends on it at all.)

There are test-Depends (debian/tests/control) on some of the not-in-testing packages, which might technically be an RC bug since it fails migration-reference autopkgtests, but I don't think Britney looks at those.

(Non-migration-reference autopkgtests can pass "in testing" with test-Depends that are only available in unstable, because of a known issue in autopkgtest that allows tests to take more packages than requested from unstable:
https://sources.debian.org/src/autopkgtest/5.49/runner/autopkgtest.1/#L292
and #1084880. This is probably how packages that test-depend on not-in-testing packages were able to migrate to testing. However, when they do that, these extra packages do not appear on the --pin-packages line, e.g. r-cran-dendextend in https://ci.debian.net/packages/r/r-cran-plotly/testing/amd64/60145266/ . Hence, I don't think that's what happened here.)


Reply to: