- Fixed ClusterLab.updateResults() not to bail out completely in response
to a low-level crashed job (which can happen because of engine failure)
- Slightly more aggressive closing of connection to cluster, since connections
held open for a long time seem a lot more prone to failure
- Slight change to the semantics of lambda-expressions in Python3.8 broke
ClusterLab
- Dropped support for dill, replaced with automatic use of cloudpickle
- Fixed issue with handling RepeatedExperiments
- Reverted to LabNotebook.addResult() for all additions (splitting between
addResult() and addResults() caused confusion)
- Forced pending results job ids to strings, not bytes (needed by ipyparallel)
- Fixed slightly cavalier attitude towards default (missing) values in ResultSet
- Added epyc-specific exceptions in places where a programmatic response might
be appropriate
- Modified cancellation behaviour to record the job as cancelled rather than
silently dropping it, to maintain the audit trail
- Added locking of result sets to prevent further changes
- Added Travis CI integration