Stabilization batch 07¶
Scope: investigate GAME-01's long local logo interval with focus controls and native startup timing. Reuse the batch-06 Linux IL2CPP artifact; no game source, scene, settings, package, or feature change, and no new build.
Probe identity and safeguards¶
Both probes used /tmp/tinychaos-batch06-linux/TinyChaos.x86_64 and verified GameAssembly SHA-256 c9990b14d8445504f87d9808d0b603427248c94625fc145f0dd2c1b9fcbbd957, matching batch 06. Fresh child-process configuration/cache roots were /tmp/tinychaos-batch07-profile and /tmp/tinychaos-batch07b-profile. The existing local-only development Steam app-ID file prevented switching to an installed Steam copy; Steam authentication was unavailable. Normal game-profile hashes remained unchanged.
A debugger measured entry/return at the exported native DiscordCreate function. The entry backtrace places that call on the Unity player main-thread path through GameAssembly and the Discord SDK library. This measures that constructor entry point in these runs, not every later SDK method or the entire integration. Breakpoint/backtrace overhead is instrumentation, not an uninstrumented performance benchmark.
An external sampler recorded the candidate PID, active window/PID, and main-thread kernel wait channel about every half second. Only the PID-verified candidate window was minimized/restored for the background control. No production run-in-background policy was changed. A guarded Space-key helper found the first probe's transition already started and sent no synthetic keypress; subsequent mouse interaction was not a controlled click test.
Results¶
| Measurement | Foreground probe | Background control |
|---|---|---|
| Logo-start marker UTC | 04:20:10.257 | 04:24:46.656 |
| Logo-complete marker UTC | 04:20:13.620 | 04:24:58.208 |
| Wall-clock logo duration | 3.363 seconds | 11.552 seconds |
| Focus samples during the logo interval | 7/7 candidate-focused | 7/23 focused; temporary minimize accounts for the background interval |
| Controlled window action | None | Minimize at probe elapsed 4.063s; restore starts at 12.126s, with small restoration overhead |
| Native DiscordCreate duration | 0.000724 seconds | 0.002298 seconds |
The controlled minimize adds 8.189 seconds to the logo duration, consistent with normal suspension while the player is backgrounded. PlayerSettings has runInBackground: 0, and this candidate demonstrates the corresponding timing behavior. This establishes why the earlier 61.867-second wall-clock measurement alone cannot prove a frame-freeze defect. It does not retrospectively establish what caused that unobserved interval or resolve the owner's Windows/Steam symptom.
No minute-long delay occurred in the foreground logo control, and DiscordCreate was fast in both runs. Do not infer that all graphics, saves, network calls, or Discord methods are cleared. A controlled affected-build reproduction and frame/main-thread evidence are still needed. The uncertain historical boundary remains the owner's 0d56ec2c engine/package update.
Both probes were bounded (100 and 45 seconds) and reached terminal debugger/launcher results; no debugger/player process remained. Debugger exit 0 means the diagnostic session ended, not that a playthrough passed. Private raw evidence and temporary probe scripts remain under /tmp/tinychaos-batch07*; /tmp/tinychaos-batch07-summary.json contains the sanitized local aggregates. Raw logs/focus PIDs/backtraces are not committed.
Additional confirmed menu finding¶
The foreground player reached [MainMenuManager] Menu initialized at 04:20:28.520. At 04:21:04.390 a mouse-release-driven UnityEvent logged a NullReferenceException in MenuController.PushPage. The exact clicked button was not recorded.
Serialized inspection finds ten MainMenu PushPage bindings; TutorialButton (GameObject 865741372, Button 865741375) is the one with a null object argument. PushPage immediately dereferences page.Enter(true). This is a concrete broken binding and matches the runtime failure mechanism; attributing that particular click to TutorialButton remains an inference. GAME-18 owns the finding and the required existing-route repair. The current MainMenuManager.StartTutorial handler already loads the Tutorial scene; verify the intended route before editing the persistent event. This does not authorize a tutorial redesign.
Checkpoint¶
The prior batch made verified source/player progress. This batch changes the next action with measured foreground/background evidence, narrows one native-call hypothesis, and identifies another reachable menu defect. It makes no gameplay fix or freeze-resolution claim. The personal Unity skill was updated with the focus-measurement lesson at its canonical source and validated; its installation remains a symlink.
All existing project edits, missing audio, version-control artifacts, and player data were preserved. Next small menu work is GAME-18 and navigation verification, while GAME-01 remains open for controlled affected-build/Windows investigation. Pause at the owner's one-batch check-in; keep the full stabilization objective unfinished and feature ideas deferred.