Today's work advanced MW3 campaign playtesting, repaired the native Mech Lab's ultrawide canvas, and brought Pirate's Moon to a native Vulkan menu. HUD, zoom and fog fixes improved the local Remaster build. Model and texture work continued alongside GPU geometry experiments, with failed comparisons kept separate from the playable build.
What changed
- The user reported completing MW3's first campaign mission in the local Remaster profile. Automated checks separately covered mission entry and bounded gameplay; full-campaign and save/load equivalence remain unverified.
- A video-startup divide fault was traced to an optimized decoder calibration loop. The correction preserves the original measured loop rather than inventing a delay or clock result.
- The native Mech Lab now places its original controls and 3D preview on one centered, aspect-preserving canvas with matching pointer input. At 3440 by 1440, the Lab uses a centered 1920 by 1440 area.
- An authentic Instant Action briefing-to-Lab test removed a weapon, closed the Lab and carried the edited loadout into a native mission. Widescreen world rendering resumed after leaving the Lab. The exact post-campaign route and interaction across different frame rates still need testing.
- HUD repairs corrected overflowing status backgrounds and the oversized weapon container, and restored original condition schematics while retaining replacement models in the world. Zoom corrections aligned the magnified scene with its movable window and repaired target-triangle hiding and clipping.
- A fog-cache repair restores scene-defined fog after the gameplay display-mode switch. The regular and separate alpha GPU profiles use the original 800-unit scenery range. Longer distances remain an optional, bounded experiment with higher rendering cost and map-capacity limits.
Pirate's Moon reaches a native menu
The experimental Windows x64 build displays the animated expansion menu through the shared Vulkan renderer. A scripted test opens Options, returns, confirms Quit and exits normally after native cleanup at frame 210.
All 210 scene draw counts and fingerprints match a separate original-game reference under the same input and clock settings. Seven paired images are visually close, with mean RGB channel differences below 0.76 out of 255; they are not pixel-identical.
This reused MW3's translation pipeline, native host, memory and service contracts, input, timing and renderer. Pirate's Moon still required its own executable identity, addresses, dependencies and translated callbacks. The native build executes translated code without an interpreter or original-executable fallback. Physical input, missions and cinematics remain unverified.
Models and visual development
- Orion, Daishi and Shadow Cat joined the four existing replacement-model candidates. Bounded native runs exercised each addition and its distance model; complete distance-transition and damage coverage remain open.
- An Orion-versus-Daishi combat test exposed whole-model fallback after arm loss. The repair retains surviving replacement parts while preserving original stump and debris draws. Other live limb-loss cases still need testing.
- Daishi arm placement, Shadow Cat thighs and Bushwacker waist fitting received corrections. The latest Bushwacker waist fit passed offline animation and adapter checks; live visual review remains pending.
- Worn base-color sets for Bushwacker, Mad Cat, Orion, Thor and Vulture were installed in the local launcher registry and verified through asset preparation. This does not establish in-game seam, mip or performance acceptance.
- Normal-map walking and lighting previews remain offline. The selected stronger MW5 material appearance is a future renderer target; live normal mapping, dynamic lighting and shadows are still planned work.
GPU experiments and validation
A separate alpha launcher exercises GPU world geometry and faster binding while retaining the original CPU producer. Controlled checks support continued investigation, but recurring divergence remains unresolved.
The deeper experiment that skips original CPU perspective iterations is not promoted. Its watched reference and bypass runs completed, but all five gameplay captures differed and original draw counts first diverged at frame 489. Memory-observation guards passing does not establish rendering equivalence, and no performance gain is claimed for that candidate.
Recorded regressions for the HUD and model repairs retained 116 Training captures, 1,150 original GPU draw records and 116 Instant Action captures in their stated comparisons. The final Lab repair separately retained 116 Training captures and 1,150 original draw records. These checks cover bounded scenarios, not every mission or interaction.
What comes next
Continue campaign and Mech Lab playtests, including the post-mission route and frame-rate interaction matrix. Compare radar behavior and controlled per-part damage against the original. Investigate the GPU divergence before attempting to promote CPU-work removal, and correct replacement-model fog separately. Adaptive menu layout and settings for implemented graphics options remain future frontend work.
The accepted startup-root count remains 131 of 144, with strict acceptance paused. These results add no formal milestone credit. Game media, replacement assets, music, generated translations and native executables remain local. This update is not a downloadable game release.
The source repository is currently private; GitHub links require collaborator access.