Skip to content

LateBrake v0.5.94 — Physical Reverse Overspeed Recovery (SOURCE-ONLY pre-release)

SOURCE-ONLY experimental GitHub pre-release. This is NOT a playable game build. No compiled Win64 EXE, Windows installer, Android APK/AAB or installed game is included. The GitHub source ZIP and tar.gz archives are not installers. This release has zero executable binaries and zero attached game assets.

Immutable source identity

The v0.5.94 publisher must tag the exact reviewed, tested commit of unmerged development draft PR #295. The protected main branch does not acquire the candidate vehicle physics by publication. The previous immutable source-only milestone is v0.5.93.

The publisher reauthenticates the same-repository owner-created open draft PR, its full SHA and release-note bytes before and after the exact-source CI matrix. It refuses to rewrite an existing tag or release and publishes zero compiled assets.

Focused driving improvement since v0.5.93

Physical reverse overspeed recovery

  • Replace the real player/rival Chaos pawn's hard reverse-overspeed SetPhysicsLinearVelocity clamp with a finite, signed, planar Chassis->AddForce request. Preserve actual sideways skid and vertical gravity, instead of instantly replacing XY momentum after a collision.
  • Retain the original 48.6 km/h Arcade and 36 km/h Simulation Inspired reverse-speed values as soft targets, not guaranteed instantaneous hard caps. Transient collision and road/physics forces can legitimately exceed them.
  • Limit recovery to 850 cm/s² Arcade or 650 cm/s² Simulation before actual surface, tire, garage, damage and supported rear-axle reductions. Support loss or airborne travel never synthesizes a braking force.
  • Block new reverse-motor thrust near the measured reverse-speed target without blocking ordinary forward drive, low-speed reversing, genuine opposite-pedal braking, race time or scoring. Do not count blocked reverse input as engine work that generates artificial heat.
  • Give service/handbrake and opposite-pedal braking precedence over the governor. Budget the physical reverse-correction force against the same-frame lateral traction instead of stacking unrealistic braking and sideways force.
  • Preserve the old ArcadeReverseSpeedLimit.h as a pure historical numerical test reference; the accepted Unreal drive Tick does not call its velocity-projection function.

Regression and source binding improvements

  • Add a strict standalone C++20 ReverseSpeedGovernor regression: 100,000 deterministic heading/velocity/contact/grip/pedal samples; an additional 60,000 combined reverse-braking + lateral-grip force compositions against the shared friction budget; 30/60/120 FPS integration and settling; zero-contact, invalid-input, reverse-drift, service-brake and gear-transition conditions.
  • Synchronize the existing actual Unreal source-binding checks for reverse motor, gear power, yaw steering, planar speed management, human/rival telemetry order and the non-shipping handling log.
  • Document physical reverse-speed acceptance separately from the legacy instantaneous reference.

What source qualification proves — and what it does not

The publisher must independently compile and execute the entire strict portable C++20 test suite on Windows, Ubuntu and macOS, run all Python source/fixture tests, validate PowerShell packaging safety, and build/verify documentation at the exact pinned source SHA. Do not infer fixed totals from earlier PR merge previews; the actual publisher job logs establish the counts. If any gate fails, the source pre-release must not be published.

These isolated source-model and binding checks are not an Unreal Engine 5.8 Editor/UHT/Game build. They do not establish tire realism, measured Chaos deceleration, stable suspension, scene rendering, an original cooked race circuit or installed Windows/Android gameplay. The prototype remains a simplified chassis with experimental four-corner spring support, not a production game.

Required before physics merge or a playable release

  1. Run the manual trusted real Unreal 5.8 validation against this exact review commit: Editor/UHT first, then an authentic Win64 map cook and candidate.
  2. On installed Windows hardware, drive at 30/60/120 FPS. Verify physical reverse targets, collision recovery without XY snaps, speed-vs-slip trace, full/partial/no rear contact, service and opposite-pedal precedence, steering, rivals and camera.
  3. Capture original engine/circuit/source hashes, cooked archive, immutable test receipt, actual logs/video and independent hardware measurements. The hash-bound diagnostics report requested forces; they are not proof of gameplay or physics accuracy.
  4. Validate Android separately and retain the unresolved genuine playable-game gate #148. Do not call the experimental physics stable or merge it merely because CI passes.

The publication does not include executable binaries, a setup installer, completed art/audio, multiplayer or playable racing.

SOURCE-ONLY — NOT a playable game build — no executable assets. Source archives are not installers.