fix(login): fail the run when login can't reach home

login.run() only printed a warning and returned when it gave up, so the
game window still existed, _run_sequence kept going, every later task
aborted at its own first click-verify, and the run exited 0. Cron then
logged and Discord-alerted "finished OK" for runs that did nothing --
18 of them between 2026-10-26 and 2026-10-29 while the client itself
was hanging at its "Initialize Catalog" startup screen.

Now login kills the stuck game and raises RuntimeError, so the run exits
non-zero and ba_cron_run.sh logs FAILED and sends the error alert.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Nik Afiq 2026-10-29 14:50:56 +09:00
parent 598e1ef65b
commit 8bd76b88f8

View File

@ -553,4 +553,17 @@ def run(driver, config):
if attempt < config.LOGIN_MAX_RELAUNCHES: if attempt < config.LOGIN_MAX_RELAUNCHES:
_recover(driver, config) _recover(driver, config)
print(f"[login] warning: could not reach home after {config.LOGIN_MAX_RELAUNCHES} attempts -- manual check needed") # Raise rather than print-and-return (real incident, 2026-10-29): the
# game window still exists here, so _run_sequence's own window_exists()
# check never fired -- every later task then failed its own first
# click-verify step and aborted safely, the run exited 0, and
# ba_cron_run.sh logged and Discord-alerted "finished OK" for runs that
# did nothing at all (18 of them over 10/26-10/29). Raising makes the
# run exit non-zero, so cron logs FAILED and sends the error alert. The
# game is killed first: it's stuck somewhere short of home, and this
# abort skips the exit_game task that would otherwise have closed it.
driver.kill_game()
raise RuntimeError(
f"[login] could not reach home after {config.LOGIN_MAX_RELAUNCHES} attempts -- "
"aborting the rest of this run (check nik-gpu manually)"
)