Ensure wakeup() is called even when add_job() raises an exception

[nuglifeleoji]

Problem (fixes #1092)

If the job store's add_job() call raises an unexpected exception (e.g. a SQLAlchemy error during a database write), _real_add_job() propagates the exception before reaching the self.wakeup() call at the bottom of the method.

# From _real_add_job():
try:
    store.add_job(job)          # <-- raises here
except ConflictingIdError:
    ...

# ... mark job, dispatch event, log ...

if self.state == STATE_RUNNING:
    self.wakeup()               # <-- never reached on unexpected exception

This leaves the scheduler's event loop without a wakeup signal. Already-scheduled, unrelated jobs stop being triggered until the scheduler is manually woken up.

Fix

Wrap the body of _real_add_job in a try/finally so that wakeup() is always called when the scheduler is STATE_RUNNING, regardless of whether add_job() succeeded or failed.

The job-added event dispatch and log message are left inside the try block and are therefore only emitted on success — the semantics of those operations are unchanged.

try:
    try:
        store.add_job(job)
    except ConflictingIdError:
        if replace_existing:
            store.update_job(job)
        else:
            raise
    # mark job, dispatch event, log ...
finally:
    if self.state == STATE_RUNNING:
        self.wakeup()

Made with Cursor

[agronholm]

Why should the scheduler be woken up if adding the job failed? The whole premise seems odd to me. If the job store raises an exception from add_job(), then there is either a catastrophic failure or a bug in the job store. Either way, it is not the scheduler's job to catch such errors.

[agronholm]

Judging by the fact that you force-pushed a revert which makes this PR a no-op, I conclude that you were not serious about this.