… looks like we (or better myself) have created a locked-in situation here:
- changes to the stan-math MPI makefiles were merged. cmdstan upstream tests were all good, because MPI was not yet merged into cmdstan.
- then we merged the cmdstan MPI PR which was “green” as the offending changes to stan-math were added after the cmdstan MPI PR were tested.
To resolve this, we will need to merge
Sorry for the mess, but this is another good reason to go for a mono-repo.