expose_functions errors after update to 2.33

[jgabry]

Describe the bug

@andrjohns I can't expose stan functions anymore after installing 2.33. Have you run into this? Or any ideas what's going on?

To Reproduce

The example at ?expose_functions now errors when it didn't before:

stan_file <- write_stan_file(
  "
 functions {
   real a_plus_b(real a, real b) {
     return a + b;
   }
 }
 parameters {
   real x;
 }
 model {
   x ~ std_normal();
 }
 "
)
mod <- cmdstan_model(stan_file)
mod$expose_functions()

Compiling standalone functions...
Error in 1:first_decl : NA/NaN argument

Expected behavior
No error

Operating system

Mac Big Sur

CmdStanR version number

Current master branch (v 0.6.1.9000)

[jgabry]

The error seems to be coming from get_plain_rtn() in utils.R. Is that because something in the generated c++ code changed that cmdstanr was assuming wouldn't change between cmdstan versions?

[jgabry]

Not sure why but finding the first function declaration returns NA now:

# this now returns NA_integer_
first_decl <- grep(paste0(fun_name,"\\("), model_lines)[1]

Edit: the reason is that fun_name is now "double a_plus_b" but that doesn't appear anywhere in model_lines. In model_lines we only have

a_plus_b(const T0__& a, const T1__& b, std::ostream* pstream__);

but double a_plus_b doesn't occur.

[jgabry]

we should probably run cmdstanr unit tests against the release candidate version of cmdstan going forward so we can catch this sooner

[WardBrian]

I'm guessing this is because of stan-dev/stanc3#1313 which @andrjohns requested to make this feature easier for cmdstanr

We do run the cmdstanr unit tests during the release candidates, it's not always clear if the failures are significant - this time a few looked like they were due to the deprecations

[jgabry]

Yeah that looks like it would cause the issue I'm seeing. Thankfully this doesn't affect running any models. But it does cause a bunch of unit tests to fail that aren't just due to the deprecations:

https://github.com/stan-dev/cmdstanr/actions/runs/6114306271/job/16595596153?pr=843#step:13:3398

[jgabry]

Closing now that #843 was merged