squeue fails if a job id is passed to -j for a job that is not in the

queue (any more). Because we cannot guarantee that the query intervals
are smaller than the intervals at which jobs get purged from the queue,
failed `squeue` invocations can happen.

This change replaces the `-j` option with `--me`, which lists all jobs
for a user. This has the potential of listing more jobs than necessary
(since a user can have multiple PSI/J instances managing jobs).

Author: hategan

src/psij/executors/batch/slurm.py

@@ -147,12 +147,11 @@ def process_cancel_command_output(self, exit_code: int, out: str) -> None:
 
     def get_status_command(self, native_ids: Collection[str]) -> List[str]:
         """See :meth:`~.BatchSchedulerExecutor.get_status_command`."""
-        ids = ','.join(native_ids)
 
         # we're not really using job arrays, so this is equivalent to the job ID. However, if
         # we were to use arrays, this would return one ID for the entire array rather than
         # listing each element of the array independently
-        return [_SQUEUE_COMMAND, '-O', 'JobArrayID,StateCompact,Reason', '-t', 'all', '-j', ids]
+        return [_SQUEUE_COMMAND, '-O', 'JobArrayID,StateCompact,Reason', '-t', 'all', '--me']
 
     def parse_status_output(self, exit_code: int, out: str) -> Dict[str, JobStatus]:
         """See :meth:`~.BatchSchedulerExecutor.parse_status_output`."""