Description
Bioconductor package 'sojourner' has a dependency on the CRAN package 'sampSurf' that has been archived.
The package 'sojourner' is supposed to fail to build/check/install on the build system, but shows as passing because the dependency packages are saved in the library folder of the R installation.
The packages are flushed each time R is updated, which essentially wipes out the library folder (hosting all the installed libraries), which then makes packages such as sojourner fail correctly.
The issue here is there is a time gap between reality which is a failure and the build system showing the false positive success. This needs to be corrected, as the author/maintainer of such packages won't know until the package fails (approx) 2-3 weeks after the incident of archiving dependencies. Meanwhile, a user thinks that a build passing package on the Bioconductor BBS is not installing on their local machine even though they have the correct version of R and Bioconductor. (they won't know this unless they do a deep dive into the cause of failure).
One way to solve this is to remove all the packages in the library path, except the base packages as mentioned by @hpages. This will not break the R installation. This also avoids the re-installation of R each time packages need to be "flushed".
@jennifer Wokaty Maybe we can take a look at this together? The reason I brought this up is the failure “rate” in the binary packages is hard to gauge especially when our build system (which is thought to be a canonical representation of success) is not reflective of the truth.