Battlefield flower warp - 2026-10-22
Status: VERIFIED in the selected automated native gameplay paths.

Final EXE SHA256 a5f27c7fc6e9880dc2000da9fa7f5dbea918fba6684256772ded0aa9bb739228
Build/link, two adapter checks, five focused CTests and native data audit pass.
The normal-input route reaches the original flowers; no player/actor placement,
forced hit, teleport, camera, reward or save-state injection is used.
Two actual teleport-state episodes select original destinations3 and4. Both
fade opacity31->0->31, relocate to original local-ROM destination records,
return control and set the original60tick cooldown. Holding on arrival causes
no extra teleport. Leaving and re-entering the peer cylinder enables return.
Original pause Exit Course returns to the castle; original painting/mission1
selection re-enters Battlefield. All4 warp actors are removed then recreated.
Final13000tick case exits0, muted native audio, mixer errors0, grounded Wait
and normal control. A prior7000tick repeat-only case also passes. Screenshots
reviewed locally. Mountain warps/all courses were not newly tested.

Production ground/line queries agreed with static KCL. Early QA routes advanced
an airborne waypoint then drifted off the narrow raised-field approach. Walking
height-aware waypoints now use mode0 and their existing landing/height check.
A wider measured route around the fence works. No game-floor defect was found
and no collision workaround was added. The arrival-side request during an
existing teleport is original busy-latch/input bookkeeping, not another state
episode. No flag-table change was needed. warp_probe.h/warp_replay.h are opt-in
QA observations/controls; ordinary play does not use the probes or routes.
Physical controller/listening fidelity remain separate; mixed audio can clip.
Full port remains experimental/incomplete. Full EXE and private data stay local.

Historical investigation below (superseded by the verification above):

The PC registry omitted WARPKUN (actor347). Added its recovered factory and
destructor to the native source closure and connected its original profile
with a native factory pointer. The complete profile metadata is checked against
the user's original overlay by check_native_actor_data.py. Matched src/ and
shared include/ files remain unchanged. warp_host.py adds only optional tracing
and host layout assertions; original cylinder, CanWarp, latch, destination
selection and player teleport state logic are retained.

The build links successfully. Two adapter checks, five focused CTests and the
native data/profile audit pass. A muted Battlefield run spawns all four original
warp triggers and exits normally. This does NOT prove a working teleport.

Input-only flower approaches have not reached a valid stationary trigger
contact. One was blocked along a fence and ended in defeat. A second jumped
to the lower side of the raised field. A static-KCL-derived walking route also
fell to lower ground after an airborne waypoint/Goomba encounter and stayed in
OnWall. Every replay exited0 with mixer errors0 and muted device output. No warp
request or St_Teleport_Main was observed; reject these as teleport acceptance.

Next: inspect the actual production ground queries and collision/route state
near the bridge/raised-field approach, comparing against the private static
KCL reconstruction. The static geometry probe may itself be incomplete; no
native floor-collision defect is established. The route can advance in the air,
and enemy avoidance changes its trajectory. Do not modify game collision or
warp rules to make a replay pass. Then prove stationary flower contact,
original fade/destination/control return, cooldown/repeat and course reload.
No QA player travel or actor/state injection was used in these runs.

Candidate stays local and is not installed over the accepted F checkpoint.
Current F pause-music checkpoint and public source/build kit are unchanged.
This actor integration has no new generated ROM-data object. Any future kit
still requires the existing format2 guards and fresh reconstruction/runtime.
