Currently, debugging subprocesses run by kurobako is rather painful.
There can be multiple possible improvements.
- We could show that "During study <study setting>, subprocess <id> finished with an error <error_code>" instead of "kurobako received an empty line".
- We could capture stderr messages of the subprocesses, and only show that of the failing subprocess. This prevents the important error message from being buried in other logs running in parallel.
Currently, debugging subprocesses run by kurobako is rather painful.
There can be multiple possible improvements.