Stan 2.18.1 release before RStan and PyStan 2.18 releases?

Developers
1

two bug fixes just went into develop:

should we do a 2.18.1 release?
could RStan and PyStan go straight to 2.18.1 release?
@bgoodri, @ariddell

2

PyStan 2.18 is already released, but not in pypi/conda.

3

I don’t think it is necessary. The first does not result in any errors and the second I reimplemented for 2.18 anyway.

4

does PyStan 2.18 swallow the “unknown variable” message? worried that this is going to confuse lots of users.

5

I’m not sure, I can test when I’m on a computer.

6

It shows

 DIAGNOSTIC(S) FROM PARSER:
Unknown variable: foo_lp
Unknown variable: to_vector

DIAGNOSTIC(S) FROM PARSER:
Unknown variable: foo_lp
Unknown variable: to_vector

edit. I had something wrong with my environment, so the FileNotFindError is not related to the parser warning.

My guess is that users won’t notice that unless something goes wrong with their model and they can’t find anything meaningful.

7

that is unfortunate, but I suspect you’re right about users not noticing.

8

@Bob_Carpenter mention that we weren’t doing hotfixes correctly? Should we talk about in the meeting? I’m a non-package manager and I don’t understand what’s going on, but I could treat bugfixes/features separately if it’d help with the hotfixing.

9

We decided a while ago it wasn’t worthwhile given that we keep the develop branch in a working state.

We could start hotfixing bugs if it’ll make it easier on the RStan and PyStan patch processes. It’s not necessary for CmdStan since we keep it in a pretty much releasable state all the time.

10

I can understand why PyStan isn’t available from the Anaconda distribution yet, since it usually takes a fair bit of time for new package versions to get there, but why hasn’t the release shown up in PyPi?

11

We build wheels and upload to pypi in another repo

osx + c++11 is failing, and also some problems with py37.