You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Wipe and leader-death accounting: reports show 0 wipes and 0 leader deaths after four lead deaths and two near-total party collapses #264
Found in the 2026-10-20 full observer runs (ZF #257, RFC #258, Uldaman #261, WC); data in ~/TortoiseBots-audit/tests/full-obs-*-19z16/, lessons entries zf/rfc/uldaman/wc-full-19z16 in RUN-LESSONS.md, analysis ~/TortoiseBots-audit/RFC-DEATHS-ANALYSIS.txt. Module code is Sagiroth/TortoiseBots feat/dungeon-self-clear (ours to merge into); tracked here.
Evidence: RFC run 2b report: wipes 0, leader_deaths 0, resurrections 0 although the lead died 4 times (15:03:17, 15:06:56, 15:13:42, 15:36:34) and 5 of 5 died within 63 s at step 34. Agent side: internal/dungeon/observe.go:280 throws away the 'all dead' result of countDeaths, so observe mode never records a wipe or a lead death; module side the lead self-revives within 10-13 s so all five are never dead in one tick. Fix: count a wipe when every member dies within one fight episode (a window, e.g. 90 s without a fight end), count the lead bot's deaths into LeaderDeaths, both in the agent report and the module's run record (the module counted '2 near wipes' in Uldaman: surface that too). Tests with a fixture trace of the RFC episodes.
Progress: the agent side is merged into dev via PR #294 (79c52a6): observe runs now count the lead bot's deaths and revives (leader_deaths, resurrections, seconds_dead, leader_death trace events) and a wipe spread over 90 s even when the lead self-revives (wipe event with "window": true, no double count with the all-dead latch). Still open, so this issue stays open: the module-side record of leader deaths and wipes and the Uldaman near-wipes (TortoiseBots), and a fixture test from the real RFC trace.
Module side done: TortoiseBots fix/ta264-wipes (1d63488a) counts a wipe spread over 90 s, counts lead deaths ('N lead deaths' in the done note) and adds spread wipes to the server trace summary. In integ/20d, reviewed, deployed as all-fixes-20d2. Still open here: part C (module_wipes / module_near_wipes / module_lead_deaths in the observe report) and part D (the RFC fixture).
Activity
Merge pull request #294 from pjerra/fix/ta264-observe-leader-deaths-w…
pjerra commentedon Oct 22, 2026
Progress: the agent side is merged into dev via PR #294 (79c52a6): observe runs now count the lead bot's deaths and revives (leader_deaths, resurrections, seconds_dead, leader_death trace events) and a wipe spread over 90 s even when the lead self-revives (wipe event with "window": true, no double count with the all-dead latch). Still open, so this issue stays open: the module-side record of leader deaths and wipes and the Uldaman near-wipes (TortoiseBots), and a fixture test from the real RFC trace.
pjerra commentedon Oct 26, 2026
Module side done: TortoiseBots fix/ta264-wipes (1d63488a) counts a wipe spread over 90 s, counts lead deaths ('N lead deaths' in the done note) and adds spread wipes to the server trace summary. In integ/20d, reviewed, deployed as all-fixes-20d2. Still open here: part C (module_wipes / module_near_wipes / module_lead_deaths in the observe report) and part D (the RFC fixture).
Merge pull request #305 from pjerra/fix/ta264-agent
pjerra commentedon Oct 27, 2026
Fixed by #305 (merged into dev, cfde9ac): the observe report carries the run leader's wipe, near-wipe and lead-death counts.