Ledger: Our own by design (not from the game)

These are ours on purpose and are not meant to be decoded: the debug HUD (SlotHud, F3, with the dummy and capsule name labels), the debug overlays, netplay (protocol, prediction, snapshots), the sandbox and test dummy, and the test fixtures under godot/tests/data/.

The setting “Xbox effect sprites” (effects1) (pd/render/xbox_effect_sprites, default on; command line --pc-effect-sprites / --xbox-effect-sprites; pd_effect.gd xbox_sprites) is ours: it picks the point-sprite size, the Xbox’s (the decoded target) or the PC port’s smaller one. No menu entry yet.

Known test flake (not a placeholder): --autotest=versus needs both bots to land a hit, and the bots’ skill rolls sometimes give one side no damaging press, so a run fails with “took 0”. Measured 11-04 on main 3efcc8e with the same data: st03 1 of 8 runs failed, st01 0 of 10 (perfcam head: st03 2 of 8, st01 0 of 10). Lead: seed the bots’ rolls, or let the check require hits only when a damaging skill was pressed.

The same kind of flake hits --autotest=net on st03 (“hits host 0”): the test’s bot runs into the arena’s west wall (x ≈ −11.6) and keeps pressing into it, so the fighters can stay out of range. Measured 11-04: on main a7074ff it parked at that wall and landed 3 hits (pass). On the collision merge a20e58f, the run of the full suite parked at the decoded set’s corner (−11.4, −20.8) for the whole match (0 hits, fail), and two re-runs passed with 15 and 7 hits. Lead: the bot steers off a wall contact (floor=true wall=true in its tm lines).