docs: Campaign CT gate PASSED; file #441 (death/lifestone return, intermittent); retire the CT-era hover probe
The ACDREAM_PROBE_UI_HOVER probe dies with its closed investigation
(scrollbar hover, fixed 2d6333f8) per the launch-options rule; #441
stays filed with its probe recipe for the next occurrence.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
parent
752782d0a9
commit
9ffc869925
1 changed files with 35 additions and 0 deletions
|
|
@ -24,6 +24,41 @@ What does NOT go here:
|
|||
- Every session: scan OPEN issues at start; promote/close anything we touched during the session before ending.
|
||||
- Promoting to a Phase: mark as `DONE (promoted to Phase X)` + commit SHA where the Phase entry landed.
|
||||
|
||||
## #441 — Death return to lifestone: stuck in portal once; arrived once with world not ready (missing doors/portals)
|
||||
|
||||
**Status:** OPEN (intermittent — did not reproduce under probes 2026-08-24).
|
||||
**Component:** reveal generation / death teleport / entity hydration
|
||||
|
||||
Owner observed twice (before probes were armed), then could NOT reproduce in a
|
||||
probed session the same day:
|
||||
|
||||
1. **Run 1:** died, and the death teleport left the client **stuck in portal
|
||||
space** (the authored tunnel never released).
|
||||
2. **Run 2:** arrived at the lifestone while the world was **not ready** —
|
||||
"missing doors, portals and stuff". Note: doors and portal objects are
|
||||
server-spawned WEENIES, not dat-baked statics
|
||||
(`claude-memory/feedback_weenie_vs_static.md`) — so this symptom may be
|
||||
**entity hydration lagging the viewport reveal** (CreateObject stream not
|
||||
yet drained when the reveal completed), not a composites/terrain gap.
|
||||
Possibly related to #433 (stale entities near Holtburg after /ls-style
|
||||
teleports — the flush direction of the same coin).
|
||||
|
||||
**Owner's repro protocol (worked twice pre-probe, then went cold):** log in,
|
||||
position far from the lifestone, log off, log on again, get killed. The
|
||||
fresh-login-then-die sequence appears to matter.
|
||||
|
||||
**Probe recipe (armed and documented, ready for the next occurrence):** launch
|
||||
with `ACDREAM_PROBE_REVEAL=1` + `ACDREAM_PROBE_REVEAL_TIMING=1` and keep the
|
||||
log. The always-on `[world-reveal]` lines plus `[reveal-timing]` per-edge
|
||||
attribution distinguish the two shapes directly: symptom 1 = a readiness edge
|
||||
that never arrives for the death generation; symptom 2 = `complete` firing
|
||||
with the object stream still behind.
|
||||
|
||||
**Healthy baseline (2026-08-24, `launch-death-repro-3.log`):** login gen and
|
||||
two portal gens all converge; readiness edges complete in 83-1400 ms;
|
||||
`totalMs` ≈ 5.5-5.7 s of which the balance after `materialized` is the
|
||||
authored portal-tunnel presentation, not a stall.
|
||||
|
||||
## #440 — CLOSED: Trained skill did not move to its section (and raise buttons never visually un-ghosted) until the next click
|
||||
|
||||
**Status:** CLOSED 2026-08-24, found by the owner during the CA5 drive.
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue