CRAN release looming

[jarioksa]

Due to changes in R 4.2.0 we must have a new release of vegan (see resolved issues #507, #447). The deadline CRAN gives us is April 23. Let's keep to it.

This would be release 2.6.0. The 2.5-branch is diverged too much, and we may just close the gap by having 2.6.0 (or 2.6.1 to keep 2.6.0 as the pre-release github version).

If there are any other things you want to get in the vegan, now you know the timeline.

[jarioksa]

Manually checked to pass tests with R 3.4.0 – no need to upgrade dependencies in DESCRIPTION.

[jarioksa]

I checked the reverse dependences against CRAN release vegan 2.5-7:

The deprecation of adonis in favour of adonis2 will trigger warnings in examples of the following packages:

betaper dispRity GUniFrac microeco smartsnp

I have contacted the authors of these packages. I will also delay the deprecation of adonis for this minor release so that these package authors can adapt to the deprecation (but print a message of the future deprecation).

Several packages failed their tests, but for reasons unrelated to vegan. The only packages that had different test results in 2.5-7 and 2.6-0 are those listed above.

Reverse dependences were checked in 170 packages and 609 packages had to be installed to satisfy the dependences of these tested packages. I'm happy that I have a fast computer with several cores for parallel testing.

[jarioksa]

Version 2.6-2 is now in CRAN. First release candidate (2.6-1) was rejected, because it broke CRAN package microbial. That was not our fault, but the error was in BioC package phyloseq which assumed vegan pre-2.0-10 (Dec 2013) data structure, if stats::uniroot issued a warning in vegan::fisherfit. A quick trick was to go back to the old bracketing strategy in stats::uniroot which was changed for other reasons in commit 69b59dc. Now we will get an error in phyloseq from a warning with all-singleton communities, such as a "community" made of one individual: try fisherfit(1) or fisher.alpha(1). The only sustainable solution is that phyloseq is fixed so that it does not assume a pre-2014 result structure when fisher.alpha issues a warning (and warning leads to an error in phyloseq). I have published an issue and provided a pull request in phyloseq, but I do not expect that these lead to any changes very soon. However, the change made for 2.6-2 allowed acceptance of vegan in CRAN, since the other cases with new warnings were not used in any dependent package.