CRAN release 2.6-6 and the way ahead

[jarioksa]

Preparing vegan release needed more work than I expected. R 4.4.0 was released and Bioconductor upgraded to 3.19 meaning new tests and new dependencies. For changes and new features, see NEWS. This release will break (Warning or Error) at least 8/231 packages depending on vegan:

• adonis deprecation: see instructions adonis is deprecated: use adonis2 #641 on changing your package to avoid warning.
• Dropping scores from summary.cca (issue Provide less verbose output in summary.foo for ordinations #203) broke packages that used summary to get scores: see instructions Use `scores` to get ordination scores: `summary.cca` will fail in the next release #644 on changing your package to avoid error. The release will use old summary.cca with new print.summary.cca which gives a grace period these packages, but they should be fixed before the next release.

I have emailed maintainers of involved packages with patch files or made a pull request for github packages. Some of these packages may be abandoned and will be archived if not fixed.

The next vegan release will probably be 2.7-x because it will finally break things that we deprecate or obsolete in this release:

• orditkplot will be defunct and dependence on tcltk package will be removed. The function was moved to vegan3d and is still available. See vegan3d & orditkplot deprecation #632.
• adonis will be defunct (issue CRAN release & adonis deprecation: Request for Comments #523). This will break packages that do not adapt.
• Using summary to get scores will be disabled. This will break packages that do not adapt.

Other changes on queue are at least (but not limited to):

• Merge PR Matrix completion for robust CLR in decostand #619.
• Cajo ter Braak suggested that we merge his expanded and improved prc function in vegan.
• Function to add new points to NMDS. This has been done and tested in a non-published branch with Peter Minchin, but I'm not quite sure of its UI.
• Snappier (and probably faster) C code for permutation test on CCA.

[jarioksa]

You may have noticed that CRAN release has a cumbersome version number 2.6-6.1. This was not planned, but just happened. Soon after the release 2.6-6 I became aware of a critical bug #653 in compiled code. This was a cryptic bug: usually it had no impact, but occasionally and in non-repeatable pattern it can crash the session. At that stage, master branch had advanced under version number 2.6-7 and I needed to add the fourth digit to the release version number. The next major release will be 2.8-0 and master will move to 2.9-0, and if there are minor releases they will be 2.8-x, so that the last number after dash will clearly refer to minor releases with mainly bug fixes.

I have cherry-picked bug fixes and fixes to unexpected behaviour (not necessarily a bug, but against reasonable expectation) to cran-2.6-6 and these are already so many that we could release 2.6-6.2 any day. I think that 2.8-0 is so far away that there are good reasons to have minor 2.6-6.x releases. I think I'll submit a new release to CRAN in late August before University semesters, or perhaps even in early July if I'll switch to summer mode with slower development.

You may check the latest NEWS in the cran-2.6-6 branch to see what's happening.