Skip to content

Development Validation

smallketchup82 edited this page Oct 29, 2026 · 1 revision

The current target is Apple Silicon with macOS 26+ and Metal 4. The app is built from version-pinned source and dependency inputs; byte-for-byte identical DMGs are not currently claimed.

  • Test a downloaded DMG on another Apple Silicon Mac with no developer tools, Mono, .NET or Homebrew installed, including the normal macOS trust process.
  • Validate a full multiplayer match against the matching Windows/Linux engine; reporting the correct version and connecting are not synchronization proof.
  • Test multiple displays, changing Retina density, menu bar/notch placement and cursor capture on more hardware. macOS fullscreen now avoids redundant window resizing and framebuffer readbacks when hiding/restoring the persistent Mesa context. Still validate repeated lobby/game transitions, Command-Tab, native titlebar fullscreen and display disconnection during a real match. Run uv run --no-project --python python3 python tests/check-fullscreen.py for transition/geometry regression checks; add --native for a short real SDL/Cocoa fullscreen and window restoration probe (opens a temporary window).
  • Validate automatic notch-HUD switching on more MacBook models, external monitors and UI scales. The gap follows the active display's native camera boundaries. Native Spaces fullscreen, notch-covering fullscreen and windowed restoration have local probes; multi-display disconnection and full-match transitions still need broader testing. Run luajit tests/check-notch-spacing.lua and luajit tests/check-notch-settings.lua for layout/settings regressions. Native coverage probe: compile tests/check-notch.mm with the engine's Mac WindowManagerHelper, SDL2 and AppKit, then run with and without custom.
  • Investigate intermittent missing Chili panel backgrounds after joining a multiplayer match. Reported again on Onyx Cauldron 2.2.2 in TAW on October 28: rejoining the same live match retained the problem, while the owner's replay displayed normally. Saved HUD opacity settings were 0.95–1.0; all 1,161 inspected UI Lua files and skin assets passed pool checksums. The affected sessions used old Chili, with macOS HDR disabled, and logged no Chili errors. An isolated replay also retained visible panels after starting and skipping to frame 3,600. This does not reproduce a live join. Root cause remains unconfirmed. Compare live join/catch-up rendering with replay playback, including Chili's deferred texture loading and cached display lists. If it recurs, capture it before trying /luaui reload; that command rebuilds LuaUI, but recovery of this particular symptom is not yet verified.
  • Optional graphics effects still need individual compatibility work, including geometry-shader selection, Xray glow, nightvision and a deferred widget's shutdown path. Do not advertise complete GL 4.6 feature support.
  • Measure long-game memory, frame pacing and power use with shared GPU textures and the CPU compatibility path. GPU event synchronization and reusable Metal uploads now have local native checks and short measurements; validate sustained behavior and mixed-display transitions on more Apple Silicon models.
  • CAMetalDisplayLink is experimental and disabled by default. Initial static and normal-replay samples regressed. Investigate missed presentation deadlines and compare actual displayed intervals before considering it a default. Validate sustained behavior, external-display rate changes and one-frame versus two-frame scheduling on more hardware.
  • Revisit the legacy AnyCPU shared-library build so developers no longer need Mono MSBuild. The distributed ARM64 app already bundles its native .NET runtime.
  • Keep upstream source/third-party notices and exact build pins current when updating the game-compatible engine branch or helper runtime.

Clone this wiki locally