The 2.22 release changed the default stanc compiler.
I added logic to the Makefile that would let users compile using the old stanc2 compiler, but I think that the way I handled this for the 2.22.1 release was quite clunky - it forces the user to keep recompiling bin/stanc in order to toggle between the old and new compilers.
I implemented a better solution which went in with PR https://github.com/stan-dev/cmdstan/pull/820. This solution allows the user to pass in make option STANC2=TRUE when compiling the model; this will build bin/stanc2 as needed and use it to compile the Stan program.
I think it would be good to put out another minor release of CmdStan now so that this is exposed to CmdStan users, because we are seeing bugs reported for the new compiler - e.g. https://github.com/stan-dev/stan/issues/2890 (note this should be a stanc3 issue). This would make it easier for CmdStan and CmdStanR and CmdStanPy users to work around bugs in Stanc3.
I think the way we handle stanc2 now on develop is much nicer to the user so this seems reasonable.
So this release would be 2.22.1 + PR 820 + PR 811 + stanc3 updated to the latest nighlty build. Or do we want to include anything else?
Not sure the make clean fix in #816 is as critical. Other than that there were changes to the random value and how the threading/mpi/opencl info is shown in stdout and csv. But that does not need to be included.
The cleanest thing would be to make a PR to master with all the PRs mentioned above. We then review that and @serban-nicusor makes a release tarball with the latest stanc3 build.
We must make sure that the Stan submodule reference is the same as in 2.22.1. Otherwise this is a 2.23 release.
Based on our current release schedule we are ~5 weeks away from the next release candidate coming out and hopefully 6 weeks from the 2.23 release. Do we then also move this dates back a bit or should we stick with Jan 18/Apr 18/July 18/Oct 18 release dates?