Alpha 0.6.3 — Revert 0.6.1/0.6.2 over-eager cleanups + lobby BG leftover fix

MicroMotorMayhem NetLink — Online Multiplayer (Alpha 0.6.3)

Hotfix on 0.6.2. Reverts the 0.6.1/0.6.2 "defensive cleanup" attempts which made rendering WORSE in real testing. Keeps the lap-count off-by-one fix from 0.6.2 and the per-race instrumentation, and adds a focused fix for the lobby-background leftover-artifacts bug.

💝 Massive thanks to jberetta for the original MicroMotorMayhem.

How to play

Load game.cue. Title → Online Play → name → dial 199406.

Changes since 0.6.2

Change	Why
Reverted: PHASE C zero-out of xpdata_/pdata_LP_/cdata_/map_section[] entries	User reported that 0.6.2 made rendering MORE broken (even in race 1) instead of better. The zero-out was nulling pointers that upstream code apparently dereferences for indices >= model_total — getting away with it when those entries held random garbage from un-init'd memory, but crashing or producing worse corruption when set to NULL.
Reverted: xpdata_/pdata_LP_/cdata_ linkage from file-scope back to static	Promoting them to file-scope shifted BSS layout — the cascading offset of all subsequent statics may have been the actual cause of race-1 rendering corruption in 0.6.2. Putting static back restores the original BSS placement that the upstream HUD/PUP/sprite-id math depends on. The DIAG instrumentation now uses a helper function defined where static xpdata_ is visible, instead of an extern reference.
Reverted: extra jo_disable_background_3d_plane before init_3d_planes	Cargo-culted theory; both disable and enable only flip a bit, neither actually clears VDP2 RBG0 VRAM. Removed.
Reverted: skip-display-init-when-current_players-unchanged optimization	Wasn't a fix, just an optimization, and may have contributed to viewport mismatch on subsequent races. Always re-init for safety.
Kept: server-side lap-count > instead of >= (from 0.6.2)	The "race ends after 2 laps not 3" bug. Already-deployed server fix.
Kept: DIAG_RACE + DIAG_XP instrumentation per-race	Still need ground-truth on what xpdata_ looks like in race 2 vs race 1. Now via helper function so the data is captured without disrupting linkage.
NEW fix: disable RBG0 + clear NBG1 when end_level transitions to LOBBY	User reported "lobby screen rendering background is broken and shows left over artifacts from the previous race". Race-end → LOBBY path didn't reset the RBG0 plane that held the previous race's sky/floor. Now does. mmm_online_start_race's init_3d_planes re-enables for the next race, so this is contained to the lobby's lifetime.

What's IN alpha 0.6.3

• All 0.6 functionality (race ends correctly, P2 controls work, leaderboard polished)
• Lap-count fix (race ends after 3 real laps, default lap_count=3)
• Instrumentation for one-shot diagnosis if anything is still off
• Lobby background no longer shows previous race's leftover sky/floor

Test plan

One race, one return-to-lobby:

1. Cold boot → online → start race → drive at least 4 START crossings (= 3 real laps)
2. Watch for rendering corruption (3D background, track geometry, sprites)
3. Race ends with Race finished! (not stall) — server log confirms via DIAG pid=0 lap N->M lines
4. Back to lobby → confirm clean lobby background (no leftover sky/floor)
5. Start race 2 → drive briefly → confirm rendering

The diagnostic logs (DIAG_RACE, DIAG_XP, DIAG_P1 INIT/HB/LAP/CP, DIAG_PER, DIAG_GP, DIAG_REPIN) capture enough state for any remaining issue to be diagnosed without another test cycle.

Disclaimer

Online NetLink functionality implemented with assistance from AI (Claude). The underlying MicroMotorMayhem game — engine, tracks, cars, physics, powerups, AI, audio — is entirely jberetta's work.