Reply To: JobDistributor Bug

[Anonymous]

For what it’s worth, the groups that use the race-condition-sensitive in_progress semaphore are the groups whose cluster structure does not lend itself to the race condition being a problem. Where the race condition is a problem, a wide variety of other solutions exist in the C++ job distributors; I don’t know how to use them in PyRosetta. (The standard solution is MPI communication of who does which job).

Using a directory is an interesting solution – good find!