Bug in Stan 2.13 - new feature compound declare/define

I’m very sorry to say that there’s a bug with the just-release Stan 2.13: https://github.com/stan-dev/stan/issues/2138.
There is a fix which I will be submitting ASAP.

If no one minds, I’d like to undo release 2.13.0 since none of the interfaces have used them. Any objections?

I’ve submitted a PR for the fix.

This bug surfaced when I tried to run a model that used compound declare/define for a container type. I should have tried to run the test models for this feature by hand, but I thought that that was happening via the integration tests.

The test I came up with was to add statements into the Stan Bernoulli example model and run it via cmdstan. This might not work for all features added to the language, but it’s what I’m going to always do going forward.

I think it is okay to do that. If there are people out there tracking Stan on GitHub that we don’t know about, then retagging might mess them up a little bit. We could just call it 2.14.0, but either way the PR to expose the choose function to the Stan language should also go in.

Is there a downside to using 2.14.0 or 2.13.1? Seems like a good idea to me.

I’d go with 2.13.1 if it’s too much of a hassle
to retag the new one as 2.13.0. I don’t know if
anyone’s done anything dependent on this.

  • Bob

That’s easy enough. We’ll tag 2.13.1 on Monday.

how about we tag by the meeting on Thursday. I didn’t get to it today.