Proposal: Drop support for R 3.3.2 about r-base-feedstock

[isuruf]

I don't have an opinion about dropping 3.3.2.
To drop it with conda-smithy=3, you have to change the conda_build_config.yaml file in conda-forge-pinning package

[ocefpaf]

Thanks @jdblischak for doing this!

[jakirkham]

Yes, this can be done. Please see PR ( #38 ).

It may not have worked, but it is still nice to have the history. Maybe there was some choice made that comes up again.

All I did was checkout Eric's copy of the branch merge with master and then overwrite everything in the merge commit with the contents of master. So it's a no-op commit that takes all the other changes along for the ride. :)

[jdblischak]

I know everyone is busy with the migration to conda-build 3, but I'd appreciate some feedback on this proposal. I need to know 2 things to be able to proceed:

1. Do you support dropping R 3.3.2? If yes, please +1 this Issue.
2. How can I update conda-smithy to drop R 3.3.2? Would it be possible to do a minor release, e.g. 2.4.5 -> 2.4.6, to make this small change prior to the cb3 migration? If it'd be preferred to update conda-smithy 3.*, I'd need to be shown where to make the change.

Thanks!

cc: @conda-forge/core @bgruening @isuruf @jakirkham

[ocefpaf]

2. How can I update conda-smithy to drop R 3.3.2? Would it be possible to do a minor release, e.g. 2.4.5 -> 2.4.6, to make this small change prior to the cb3 migration? If it'd be preferred to update conda-smithy 3.*, I'd need to be shown where to make the change.

I'd say let's do that with cb3. It will be simpler.

[jdblischak]

OK. I've submitted PRs to staged-recipes and the pinning feedstock:

conda-forge/conda-forge-pinning-feedstock#28
conda-forge/staged-recipes#5325

[jdblischak]

Pinging a bunch of conda-forge R users (I skimmed the last month of PRs to staged-recipes). Do any of you still rely on support for R 3.3.2? We have been unable to build it with the latest conda-forge pinned dependencies. Please +1 this Issue if you are ok with this.

@ArneKr @bgruening @bsennblad @cbrueffer @daler @FrodePedersen @jenzopr @MathiasHaudgaard

[cbrueffer]

I support dropping R 3.3.2; thanks for the initiative!

[bsennblad]

This should be fine! You got my support.

[jdblischak]

Great. Looks like everyone agrees on this change. The plan will be for it to go into effect once conda-smithy switches to conda-build 3.

[jdblischak]

@conda-forge/r-base Could one of the maintainers please delete the 3.3.2 branch? I don't have permission, and I couldn't find a way to submit a PR that deletes a branch.

[ericdill]

@jdblischak done. I've got the 3.3.2 branch on my fork so we can easily restore it if necessary

[jakirkham]

Would it be better to merge that branch into master so the history is not lost?

[ericdill]

Would it be better to merge that branch into master so the history is not lost?

Can this be done without rewriting the history of the master branch so as to not actually change the tip of master and force a rebuild?

[isuruf]

That branch never worked. (All the packages built from that branch were broken)

[jdblischak]

Can this be done without rewriting the history of the master branch so as to not actually change the tip of master and force a rebuild?

I did a quick test merge locally, and this creates lots of merge conflicts.

There were only a few commits made on the 3.3.2 branch:

conda-forge:master...ericdill:3.3.2

Most are just updating the pins and other fiddling. The only one of interest (to me at least) was my patch to backport zlib support:

5c6666a

But as @isuruf noted, despite successfully building, these packages still ended up being broken, so it was a short-lived success.

[jdblischak]

Now that conda-build 3 support has been merged into staged-recipes (conda-forge/staged-recipes#5274), R 3.3.2 is no longer supported on conda-forge. Thanks everyone for your feedback and assistance!

And related, R 3.5.0 was released today, so R 3.5.1 is just a few months away.