Jonathan Clark
Jonathan Clark
This was originally observed as of revision af3cd96, before the deadlock detection code in the unpacked walker unit test was fixed. To reproduce (for that revision): ``` sbt test ```...
Currently, jobs are submitted (e.g. to qsub) as soon as they become available. Especially when a workflow begins to run, this can result in a very large number of calls...
Some packages with stable release cycles may be installed system-wide and have a well-established version number (e.g. SRILM). These are generally used with the disk "versioner", which does not support...
Currently, submitters only allow a "run" action, but they should also allow the following actions (mutually exclusive with "run"): - submit - get_q (static, shared among all tasks using this...
Currently, the "env" and "commands" modes, calculate and unpack the full plan, before outputting the commands. This is unnecessary and slow. The fix is to generalize Plans.getPlannedVertices() to take a...
It would be useful to be able to "diff" the entire history of a workflow leading to some task, comparing two specified realizations. This would tell the user exactly what...
This work has already begun, but should be updated after the CLI stabilizes. This issue might need to get broken up into smaller issues once we know the scope of...
The versions mode should show: - The timestamp at which each workflow version was run - The tasks that were planned for that version and whether each completed or failed...
Currently, our visualizations show packed representations of the workflow. It would be useful for debugging if we were able to show a visualization of a single unpacking of the workflow....