Release CmdStan & Stan 2.28.2

[rok-cesnovar]

Description:

There were two bugs that were discovered in the 2.28.1 release:

• Performance regression 2.28 #1055
• Unable to set init files with num_chains > 1 #1058

These have now been addressed and we can do a patch release. We don't want to release on a Friday, so we should do it Monday or whenever @serban-nicusor-toptal has time (but not before Monday).

Here is a short checklist:

• merge diagnostics writer now has a nullptr if we are not using it #1060
• check that Stan submodule updates on CmdStan master after https://jenkins.mc-stan.org/blue/organizations/jenkins/Stan/detail/master/30/pipeline/ finishes
• update the release numbers of Stan on the master branch (not on develop!)
• update the release numbers of CmdStan on the master branch (not on develop!)
• release Stan 2.28.2 (release notes are posted below)
• release CmdStan 2.28.2 (use 2.28.1 stanc3 tarballs - there were no changes there, release notes are posted below)
• run CmdStanPy/R tests to make sure we didn't miss anything (CmdStan tests arent really that rich)
• create and merge Stan PR to merge master to develop
• create and merge CmdStan PR to merge master to develop

Stan release notes:

• Fixed a performance regression caused by using the diagnostic writer even when diagnostics were not requested.

CmdStan release notes:

• Fixed a performance regression caused by using the diagnostic writer even when diagnostics were not requested.
• Fixed a bug that prevented the use of init files when num_chains > 1.

Current Version:

v2.28.1

[serban-nicusor-toptal]

Release for Stan and CmdStan is now ready.
CmdStanR tests are running here

[serban-nicusor-toptal]

@rok-cesnovar cmdstan is failing on master, did I mess something up or is this related to #1062 ?
Sadly Stan failed too

[rok-cesnovar]

Yes, it related to 1062. Start a run with master branches for submodules and it will pass.

[serban-nicusor-toptal]

Argh I think Stan failed because of this develop here
PS: I've fixed #1062 so there's no need for manual action.

[rok-cesnovar]

The releases look good, the cmdstanr tests seem to be passing. @WardBrian can you run cmdstanpy when you have the chance?

PS: I've fixed #1062 so there's no need for manual action.

hm, but wont this now be a problem with develop cmdstan? I guess we will see :) But we can address that later.

[rok-cesnovar]

Only thenon-x86 binaries are missing, then should be good if cmdstanpy tests pass.

[WardBrian]

Cmdstanpy tests are running now as a result of my pr to allow us to run them on arbitrary versions: stan-dev/cmdstanpy#502

If either of you could take a look at the GHA file and approve, in the future this should allow us to manually run all the cmdstanpy tests on any version (including RCs, for example)

[serban-nicusor-toptal]

I'm not entirely sure why Stan fails, I've ran it with master here with no success.
PS: Working on the non-x86 tarballs

[WardBrian]

CmdStanPy tests are green

[serban-nicusor-toptal]

non-x86 are up now! The only thing left is to make Stan tests pass green on Jenkins and then the PRs

[rok-cesnovar]

Order seems to be restored and the patch release process is done. Thanks @serban-nicusor-toptal !