Brian Ward

This would get messy if users changed the `percentiles` arguments between calls, I think. Same if we exposed the [`--include_param` argument](https://github.com/stan-dev/cmdstan/pull/1143) for filtering down to specific variables (which I thought...

Your two questions are part of why we did not implement this when first exposing pathfinder -- it's not entirely clear when users want these, and it introduces a whole...

On the most recent couple versions of stan, fixed_param is actually never necessary, the behavior where it would use it with 0 parameter models is purely legacy at this point....

A program that was previously manually specifying fixed_param=True would probably be at least as upset if we started transparently returning normal draws from HMC as it would at us erroring....

> A client specifies fixed_param=True for a Stan program with a parameter vector of size greater than zero. In this case, their program will fail in the existing setup (I...

This is cool, it sounds like you may have been building this around the same time we were building [stan-playground.flatironinstitute.org](https://stan-playground.flatironinstitute.org), which also compiles Stan models to WASM. My guess is...

I can give the config changes and migration code a once over soon if you’re hoping to do this release during the offseason still

There is this in the code: ``` T_p pi_div_p = pi() / p; ``` If T_p is deduced as `int`, then this will also be an int and you'll get...

I think making `pi_div_p` have type `auto` might be enough. I'm sure there are other places where this would be an issue, though, so maybe we should just make the...

Users of the Stan language won't run into this starting in the next version, but we should probably still fix it for users of the C++ lib directly. A similar...