Repository navigation
daemon truncates its own log on every start, erasing the previous daemon's output #991
Description
open(LOG, 'w').close() in __main__ truncates the log on every daemon start. When ensure_daemon() respawns a daemon, the previous daemon's last output (WS close code, fatal error) is destroyed before it can be read.
Observed on 0.1.13 (Linux, native Brave, toggle flow): a daemon served fine; a later call found it replaced; the log contained only the new daemon's history — the previous exit was undiagnosable.
Fix is one line: rotate (os.replace(LOG, LOG + '.1')) or append. Related: #920, #865, #784.
Activity
[-]daemon truncates its own runtime log on every start (open(LOG, 'w')) — previous daemon's death evidence is erased[/-][+]daemon truncates its own log on every start, erasing the previous daemon's output[/+]