How does bioconductor_full compare to build system?

[seandavi]

Has anyone done any formal checking to see if bioconductor_full is equivalent to the build system?

[mtmorgan]

Nitesh has tried to build and check all Bioconductor packages; with high success; a challenge is that some packages fail anyway (because they are broken in their current state) and some packages fail for transient reasons (e.g., because a web resource is not available). I believe that >1700 of teh ~1750 packages in the current release built and checked; Nitesh is not available for a couple of week so a definitive answer will have to wait...

[nturaga]

Hi @seandavi and @mtmorgan,

The bioconductor_full image will not be a complete replica of the linux build system although it's very very close.

There are a few reasons for this,

1. We purposefully avoid installing a few dependencies which make the image very large. Eg: xps, Rmpi, ccfindR. Any packages which depend on these will also not install.

2. All packages which fail on the build system because of flaws on the RELEASE_3_X or devel will also fail on the docker image.

We've had a very high success rate on the number of packages. I haven't tested the bioconductor_full image lately, but the terra_bioconductor_jupyter image which closely replicates the system has been tested recently. It successfully installs 1693 packages out of 1735 total software packages, when I last tested it (excluding the ones which fail because of issues inherent to the packages themselves).

We can have a more comprehensive test by simply installing the packages on the latest bioconductor_full image.