UnboundLocalError in job_status when status command is killed by signal

[johnyaku]

Bug

When the cluster status command is killed by a signal (negative return code) fewer than 10 times, job_status() falls through the except block without assigning ret, then crashes trying to access it.

Traceback

File ".../snakemake_executor_plugin_cluster_generic/__init__.py", line 236, in job_status
    ret = ret.strip().split("\n")
          ^^^
UnboundLocalError: cannot access local variable 'ret' where it is not associated with a value

Root cause

In __init__.py, the job_status closure (around lines 196–236):

def job_status(job_info, valid_returns=...):
    try:
        ret = subprocess.check_output(...)       # line ~201
    except subprocess.CalledProcessError as e:
        if e.returncode < 0:                     # killed by signal
            self.status_cmd_kills.append(-e.returncode)
            if len(self.status_cmd_kills) > 10:
                # logs info and clears list
                pass                             # falls through
            # <-- no return or assignment when count <= 10
        else:
            raise WorkflowError(...)

    ret = ret.strip().split("\n")                # ret is unbound!

When e.returncode < 0 and len(self.status_cmd_kills) <= 10, the handler does nothing — no return, no assignment to ret. Execution continues past the except block to ret.strip() where ret is unbound.

Suggested fix

Return "running" when a signal-killed status check is ignored (count ≤ 10):

if e.returncode < 0:
    self.status_cmd_kills.append(-e.returncode)
    if len(self.status_cmd_kills) > 10:
        self.logger.info(...)
        self.status_cmd_kills.clear()
    return running  # <-- treat interrupted check as still-active

Environment

• snakemake: 8.30.0
• Executor: cluster-generic
• Cluster: PBS (Gadi NCI, Australia)
• Status command: qxtat check --snakemake

The status command was intermittently killed by a signal during normal workflow execution, causing the entire snakemake session to crash even though the underlying PBS jobs were still running or had completed successfully.