Add passive F2 diagnostics overlay with persistent session visibility
This commit is contained in:
1 parent
20aa1bc2b9
commit
113e9a41cc
17 files changed
+414
-24
No files matched your search
+41
-1
@@ -1,4 +1,4 @@
|
||||
# F1 debug menu
|
||||
# Debug menu and live overlay
|
||||
|
||||
User-requested modern investigation tools, separate from recovered original UI.
|
||||
F1 works in the game, during cutscenes and practice prompts, and in scene inspection.
|
||||
@@ -8,6 +8,7 @@ audio. Closing resumes the previous mode. No new assets or dependencies are need
|
||||
| Key | Action |
|
||||
| --- | --- |
|
||||
| F1 | Open/close |
|
||||
| F2 | Toggle live diagnostics without pausing |
|
||||
| Up/Down | Select a row |
|
||||
| Left/Right | Change selected level, startpoint or staged SCENE task value |
|
||||
| Page Up/Down | Change staged SCENE value by10 |
|
||||
@@ -62,6 +63,45 @@ It also enables Force Control through the menu and verifies actual movement whil
|
||||
Carl's MCA is still playing. Director checks cover disabling the override, shot
|
||||
changes, scene reset and the normal completion music handoff.
|
||||
|
||||
## Live diagnostics (F2)
|
||||
|
||||
F2 toggles a compact panel in the upper-left corner while gameplay, cutscenes and
|
||||
audio continue. The F1 menu also offers Live Overlay; opening F1 replaces the live
|
||||
panel temporarily, preserving its setting. F2 works while F1 or practice is open
|
||||
without changing their pause state. Visibility persists through resets and scene
|
||||
loads within the process; it is not stored in TSK. `--debug-overlay` starts enabled.
|
||||
|
||||
Twelve bounded rows show measured FPS/frame milliseconds; level/start; root XYZ;
|
||||
velocity XYZ; yaw/speed/turn rate; ground/pose/energy; control/force/pause; SCENE/
|
||||
score/jetpack state and burst timer; actual camera mode/far plane; cutscene/shot/
|
||||
blocked state; movement counters; and reported issue/Vulkan-error/frame counts.
|
||||
STRT, orbit, gameplay and cutscene modes use the actual rendered view selection;
|
||||
orbit reports its own radius-derived far plane. A paused player/control lock is
|
||||
separate from pausing the whole simulation.
|
||||
|
||||
Timing measures rendered-loop wall intervals, including rendering/presentation,
|
||||
independently of the simulation clock. It shows an initial running mean, then
|
||||
publishes half-second window averages. Text uses the preceding timing sample;
|
||||
the first frame has no measurement yet. This is frame timing, not GPU profiling.
|
||||
Timing continues while hidden so enabling the panel has current samples.
|
||||
Motion fields describe the player simulation; standalone mesh inspection has no
|
||||
running player and displays zero/default motion values.
|
||||
|
||||
DebugDiagnostics is read-only and retains timing/visibility across scene rebuilds.
|
||||
DebugMenuOverlay reuses the existing font and fixed vertex ranges with a smaller,
|
||||
content-sized panel. Gameplay state, pause policy and original HUD stay separate.
|
||||
There are no added assets or dependencies.
|
||||
|
||||
DebugDiagnosticsTests verifies timing, finite zero/invalid cases, live snapshots,
|
||||
camera selection, culture-independent formatting and label bounds. The24-frame
|
||||
`--smoke-debug-overlay` SDL check verifies F2/repeat handling, movement without
|
||||
pause, hide/show, F1 priority, visibility preference, menu action and800x600 resize.
|
||||
Native Linux evidence: artifacts/live-debug-window.log; zero Vulkan validation
|
||||
errors. The existing F1/force-control SDL check also passes. Full regression and
|
||||
cross-scene results are in progress.md; no new Windows runtime check is claimed.
|
||||
The headless lab/VR04/lab check preserves enabled state and frame samples1/2/3,
|
||||
with zero Vulkan errors (artifacts/live-debug-return.log).
|
||||
|
||||
Native Linux GTX1070 window test passes with zero Vulkan validation errors:
|
||||
artifacts/debug-window.log. GPU screenshots debug-menu.png and debug-window.open.ppm
|
||||
show the menu. Source/build/regression evidence does not constitute new native
|
||||
|
||||
Reference in new issue
Block a user