▋
docs/archive/tonight_spawn_roxas_plan.md

Tonight Plan: Spawn Extra Roxas (Local Prototype)

Scope: local single-room proof of concept only (no multiplayer sync tonight).

Session Goal

Spawn one extra Roxas in a fixed test room and capture enough evidence to make it reproducible.

1) Lock test environment (15 min)

  • Use one fixed KH2 build only
  • Use one fixed mod setup only
  • Pick one fixed test room (no transitions during tests)
  • Start a timestamped log file with:
    • build hash
    • mod hash
    • room label/id

Exit criteria:

  • Environment details are written down before any probing.

2) Baseline probe capture (20 min)

Capture baseline values every second for ~30 seconds:

  • world_id
  • room_id
  • map_program
  • battle_program
  • event_program
  • party slot references (0/1/2) if available
  • current actor/enemy count if available

Exit criteria:

  • You have a "known-good idle room" baseline log.

3) Confirm spawn path from an existing entity (45 min)

  • Identify the known Roxas spawn in the current room
  • Pick a "spawn sentinel" value A (something that flips when Roxas spawns)
    • Prefer simplest values you can observe reliably:
      • HP / alive flag (if you can damage Roxas)
      • OR position/rotation floats (always change after spawn)
    • In CE: find A by narrowing:
      • Load the room with Roxas already present (or wait until he appears)
      • Narrow by a value you can repeat (HP change, or x/y/z change due to a small movement/interaction)
  • Confirm A is Roxas-specific (validation by causing a change that affects only Roxas)
    • If you damage Roxas, A should change accordingly
    • If you teleport/move Roxas, A should track that movement
  • Set a breakpoint/watch on A and capture spawn parameters at the moment Roxas is created/activated
    • In CE: "Find out what writes to A" (or set breakpoint on write if available)
    • Reload the room and let Roxas spawn happen again
    • When the watch triggers, record:
      • the newly created actor/enemy pointer (spawn result pointer)
      • object id (spawn object/model id) if visible in the breakpoint/register context
      • spawn group id (the group/script/controller id used for the spawn) if visible
      • initial position/rotation for the spawned entity (read nearby floats around the result pointer)
  • Repeat / reload to verify repeatability
    • Roxas spawn should trigger the breakpoint again
    • The captured sequence should be consistent:
      • same sentinel A behavior
      • same "result pointer pattern" (new actor pointer changes per reload, but should be consistently produced by the same code path)
      • same object id and spawn group id (or extremely stable equivalents)

Exit criteria:

  • You can consistently observe one normal Roxas spawn end-to-end and produce:
    • a reproducible sentinel A
    • a reproducible spawn-result pointer source
    • stable object id / spawn group id (or the closest stable equivalents visible)

What to send me after each breakpoint hit (so we can deduce the actual spawn surface):

  • result_actor_ptr (address/value you identified as the spawned Roxas entity)
  • object id (if visible) and spawn group id (if visible)
  • pos/rot x,y,z read near result_actor_ptr (even if offsets are rough/temporary)
  • Any stack/callsite info CE shows (function name/module + the first few frames if available)

4) Attempt extra Roxas spawn (45 min)

  • Trigger one controlled Roxas spawn in the same room
  • Immediately capture:
    • spawned actor pointer
    • object/model id
    • HP/state flags
    • position/rotation
  • Observe for at least 20-30 seconds

Success criteria:

  • Roxas appears visually
  • no immediate crash
  • no instant despawn loop

5) Stabilize if partial success (30 min)

If he spawns but glitches/despawns:

  • Test small spawn position offsets
  • Test spawn timing (avoid transition/cutscene windows)
  • Test forcing a safe initial state (idle-ish behavior)
  • Change only one variable per attempt

Exit criteria:

  • Either stable behavior improves, or failure mode is clearly characterized.

6) Writeback and tracker update (15 min)

  • Update docs/pointer_map_v1.md with:
    • newly confirmed values
    • still-unknown values
    • exact pointers/offsets observed
    • reproducible steps and outcomes

Exit criteria:

  • Another session can continue from your notes with no guesswork.

Hard stop rules

  • If a step fails 3 times with distinct inputs, log and move on.
  • Do not add room transitions until single-room spawn is stable.
  • Do not start multiplayer sync work tonight.

End-of-night deliverables

  • Baseline + spawn-attempt log file
  • One reproducible "spawn Roxas" method OR one documented failure mode with captured evidence
  • Updated docs/pointer_map_v1.md