fix(cron): media-send failure logs an empty reason when the cause is a timeout

[AiwendilInTheWoods]

Problem

When a cron job's media delivery fails on a timeout, the log records no cause at all:

WARNING cron.scheduler: Job 'abc123': failed to send media /path/to/file.mp3:

Note the trailing colon with nothing after it. The operator cannot tell whether this was a timeout, an auth failure, a size rejection, or a network error — and no traceback is emitted on this path either.

Root cause

cron/scheduler.py bounds the send with future.result(timeout=30) and re-raises TimeoutError into the generic handler, which formats the exception with %s:

try:
    result = future.result(timeout=30)
except TimeoutError:
    future.cancel()
    raise                                   # -> outer handler
...
except Exception as e:
    logger.warning("Job '%s': failed to send media %s: %s", job.get("id", "?"), media_path, e)

TimeoutError is raised with no arguments, and str(TimeoutError()) is the empty string. So the single most likely failure on this path is precisely the one that erases itself from the log.

>>> import concurrent.futures
>>> repr(str(concurrent.futures.TimeoutError()))
"''"
>>> "failed to send media %s: %s" % ("/path/file.mp3", concurrent.futures.TimeoutError())
'failed to send media /path/file.mp3: '

Fix

Fall back to the exception class name when str(e) is empty, so an argument-less exception still identifies itself. Messages that do exist are unchanged.

Related

Same defect class as #63357 (computer_use: empty TimeoutError surfaces as blank "capture failed: "), which is still open — this is the cron/scheduler.py instance of the same pattern.

Scope

Logging only. No behaviour change, no new configuration.

How it was found

A ~7.4 MB audio attachment built correctly and then failed to deliver in production. The log line was empty by construction, so identifying the cause required reading the source rather than the journal.

[teknium1]

Merged via PR #88787 (merge commit bc76f62) — your commit landed with authorship preserved, together with your #87967 timeout work. A follow-up commit folds the class-name fallback into the delivery error recorded on the run status as well (post-#88631 those reasons surface in last_error, not just the log line), with regression tests for the empty-str(TimeoutError) case. Thanks for catching the trailing-colon-nothing log!