R 3.5 pinning

[dpryan79]

At the moment, we're adding conda_build_config.yaml files to the various bioconductor packages that can only exist for R 3.5. This is less than ideal and is leading to a lot of extra files. For the next bioconductor release we should clean this up and come up with a better way of handling things.

Xref: #11452

[dpryan79]

I'll use this as an issue tracker for updates we should make before the next bioconductor release hits bioconda:

• Update the bioconductor skeleton to pin to other bioconductor dependencies within the same release. In effect, this means >=X.Y.0,<X.Y+2.0.
• Ensure that the skeleton properly quotes or otherwise handles packages summaries with colons in them. The bioconductor-derfinder summary has one and it's always breaking things.
• Ensure that any existing extra section is kept.
• Make sure only one R release is built for new bioconductor packages.

[nsoranzo]

From @msarahan's comment in #11452:

The problem here is basically the same as people see with python. Specifying any kind of constraint for python or R breaks conda-build's handling of it. The core languages are very special cases. It is best to use

build:
 skip: True  # [some r version limit]

instead. PRs to improve this part of conda-build would be most welcome.

[bgruening]

@dpryan79 any change you can push your skeleton changes somewhere? Our bioconductor friends would like to build on top of this and improve it.

[dpryan79]

I don't have any changes to push at the moment, I just made ad hoc scripts to change things. When is the next release due? It's pretty soon as I recall. My goal was to finish with the last one and then work on the skeleton (assuming no one else updated it first).