[JobRunner] Uncaught exceptions leave jobs in "running" state

[msheakoski]

The docs say to not catch exceptions and fail fast when processing messages, however, if an exception occurs within a running job, things do not properly clean up and jobs are left in a "enqueue.job_queue.status.running" state. You are also no longer able to execute a job with the same name again because it is stuck in the "enqueue_job_unique" table (DBAL transport).

I think in

enqueue-dev/pkg/job-queue/JobRunner.php

Line 49 in e2a950c

$result = call_user_func($runCallback, $jobRunner, $childJob);

this line should be wrapped in a try/catch block, failed, and then re-throw the exception so it bubbles up to the message processor. Something similar to:

try {
    $result = call_user_func($runCallback, $jobRunner, $childJob);
} catch (Exception $e) { // for PHP 5
    $this->jobProcessor->failChildJob($childJob);
    throw $e;
} catch (Throwable $e) { // for PHP 7+
    $this->jobProcessor->failChildJob($childJob);
    throw $e;
}

I am not entirely familiar with the codebase so I may be missing something, but hopefully you get the right idea from my example.

[makasim]

It makes sense to me. @ASKozienko what do you think about this?