Luma/docs/VALIDATION.md

282 lines
50 KiB
Markdown
Raw Permalink Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# Device validation
Validated on the user’s onn. 4K Plus Streaming device over ADB, October 17, 2026 (America/Phoenix). Android 14 / API 34, approximately 1.88 GiB physical RAM, 1920 × 1080 render override on a 4K display.
## Functional checks
- Native app discovery, D-pad focus across rows, OK launch, long-press management menu, and settings navigation.
- Artwork only, Dynamic, and Minimal banner modes; original spacious layout restored as the default. Focused-only labels and hidden app-section text are the defaults.
- Smooth focus scrolling across all app rows, including the All apps footer, and back to the first row.
- Plex connection, bounded rows, artwork, encrypted token storage, and disconnect verified against a temporary local fixture. Fixture credentials were removed. A real Plex account/server has not been connected; actual Plex authorization and title playback remain unverified.
- Art screensaver: all three bundled paintings decoded and displayed, sampled for nonblank pixels, with pure-black side mattes. Captions have a dark gradient and text shadows. Paintings retain their full aspect ratio in Auto/Fit.
- Scenic photo preview, single-button wake, and real one-minute idle activation passed. Idle timeout was restored to ten minutes, photo duration to thirty seconds, and mode to Scenic.
- NPS bird video downloaded to private device storage (about 40 MB), played locally in a loop, and stopped on leaving Luma. The local file remains available in Nature video mode; no ongoing stream/download is active.
- Home routing: Android reports Luma as HOME role holder but the OEM resolver still chooses Google Home. The enabled Luma accessibility service routes Home to Luma. Repeated Home presses, return from Android settings and SmartTube, and screensaver dismissal were checked.
- The previous launcher fallback stays open when explicitly selected through its app tile or Settings. Back stays on Luma; from lower rows it returns to the first row. Home returns to Luma. The previous launcher remains installed and enabled; only its competing “Override current launcher” preference is off. Its accessibility service and Button Mapper’s service are retained.
- Luma’s override Off control removes only its own accessibility service. Re-enabling through Android’s actual Accessibility settings was verified. The service is system-bound in a separate `tv.luma.launcher:home` process. No view-content retrieval capability is requested.
## Wake regression — October 18, 2026
The original blank state was captured before sending input or restarting Luma. Android reported the existing activity as resumed; tiles were displaying their text placeholders and the alpine bitmap was absent. The recorded wake sequence included start/resume, a brief pause, and then resume without another start. Image work scheduled only from `onStart()` could be skipped or discarded while paused, with no retry when the activity resumed.
Version 1.0.1 moves restoration to `onResume()`, retries interrupted app discovery, and reloads missing wallpaper, app icons, and optional media artwork. Pausing invalidates stale asynchronous results; stopping still releases image references and caches. A pause-only resume retains existing tile views and decoded app images. This follows Android's documented [pause/resume lifecycle](https://developer.android.com/guide/components/activities/activity-lifecycle).
A separate same-signature instrumentation APK reproduces the interrupted-load sequence deterministically. The installed 1.0.0 build failed with missing artwork; the fixed build passed the same case, recovery of a partially rendered grid whose icon jobs were interrupted, pause-only memory trimming, reuse of existing loaded tiles/bitmaps, and five complete stop/start cycles. No preferences are rewritten by these tests. The temporary diagnostic APK was removed afterward.
One real sleep/wake cycle also passed screenshot checks for the alpine wallpaper and Plex banner without pressing Home or restarting the launcher. A subsequent automated wake was interrupted by HDMI standby (`mLastSleepReason=hdmi`), and One Touch Play timed out; no CEC settings were changed. The user turned the TV on again for the final screen check. The final screenshot shows the alpine wallpaper and app banners restored, with the user’s five-column layout preserved. Both the launcher and Home helper recorded zero CPU ticks over a ten-second background sample. Home returned correctly from Android settings and restored the UI after its cached process was explicitly reclaimed. The Home helper was re-bound after instrumentation cleanup, preserving the original enabled services. This is not an overnight endurance test.
## October 17 performance baseline
Final device sample with the Home override enabled:
| Measurement | Result |
| --- | --- |
| Launcher UI foreground PSS | 95,566 KiB (93.3 MiB) |
| Launcher UI cached/background PSS | 66,635 KiB (65.1 MiB), reclaimable |
| Separate Home router PSS | 6,348 KiB (6.2 MiB) |
| Launcher background CPU | 0 CPU ticks over 10 seconds |
| Home router idle CPU | 0 CPU ticks over 10 seconds |
| Focus navigation rendering | 84 frames, 4 janky (4.76%); 95th percentile 24 ms |
The separate Home router avoids retaining the UI image heap. Android can reclaim the UI while the router stays bound; Home then creates the UI again. This is not a guarantee of perfect 60 fps.
Raw measurements are recorded in `../output/luma-*-memory.txt`, `../output/luma-background-cpu.txt`, and `../output/luma-focus-frames.txt`. PSS includes framework/graphics allocations and fluctuates with system state. Short CPU samples indicate idle behavior, not a guarantee covering all usage.
## Scope
No launcher application was disabled or uninstalled. Screensaver selection at the Android system level was not changed. Standard Android Dream settings are unavailable on this TV, so the customizable idle screensaver runs while Luma is foreground. No real Plex credentials were extracted or used. The UI-reclaim test explicitly killed the cached UI process while leaving the Home router bound, then verified that Home restored Luma. No reboot test was performed.
Build validation uses aapt2, javac, D8, zipalign, and apksigner signature verification. There are no third-party runtime libraries.
Final APK (1.1.10): 2,761,885 bytes. SHA-256: `d0bdcb6cecc43ecd6652d9c3da6b704f6f4586f07c2273e8694f31b6a96e0f9a`.
## Updater regression — October 18, 2026
The same-signature updater test ran against the installed 1.1.0 bootstrap and a newer signed 1.1.1 APK. It passed persisted daily network scheduling, Off cancellation, branch-to-commit pinning, a real APK download and certificate/package/version/checksum verification, and private provider read access. It rejected an incorrect checksum, another signing key, a different package, a mismatched version code, oversized metadata declarations, and public HTTP sources. The provider rejected write requests and unknown paths. The current version was not offered as an update. Test preferences were restored afterward.
The production default is the public HTTPS Forgejo feed in `output/update.json`, generated with the APK by `scripts/build.sh`. It requires no credentials on the TV. Native installer and unknown-source permission activities resolve on the onn. Android 14 device.
## Production self-update verification
A forced scheduled check against the real public HTTPS repository found a newer version while Android settings stayed foreground. Android recorded the Luma update notification, and Home showed the Update button. Downloading and verification used the commit-pinned APK from Git, not ADB. The first-use Android install-source permission screen was exercised.
Testing exposed a device-specific consequence of legacy APK/file installation: Android restricted settings reset the accessibility service list. The final flow uses a PackageInstaller session for a publisher-verified self-update (`PACKAGE_SOURCE_OTHER`), always requiring user confirmation. It does not claim to be an app store or request silent installation. Unlike a general APK file installer, Luma accepts only its own package signed by the installed key.
The final real upgrade, **1.1.4 → 1.1.5**, was downloaded from production Git and approved with the TV remote. The complete enabled-accessibility-service list matched before and after, including Luma Home, the previous launcher, and Button Mapper. Restricted-settings authorization remained allowed, and the TV returned directly to Luma. No ADB install or accessibility repair was used for that final upgrade. Prior temporary configuration changes from failed installation-method tests were restored.
The source classification follows Android's [PackageInstaller source definitions](https://developer.android.com/reference/android/content/pm/PackageInstaller#PACKAGE_SOURCE_OTHER); the platform's [file-source restriction implementation](https://android.googlesource.com/platform/frameworks/base/+/143db55921dd%5E%21/) explains the behavior encountered during the earlier file-install tests.
## Integrated TV rendering — October 19, 2026
The authorized second device (`192.168.50.200:5555`) is a TCL Smart TV Pro / G08, Android 14 (API 34), ARM Cortex-A73 cores, Mali-G52 OpenGL renderer, approximately 2.34 GiB physical RAM. Its existing 1920 × 1080 render override and density 320 were retained. About 1 GiB was available at diagnosis; Android reported no thermal throttling. The original installed version was 1.1.5.
A repeatable ADB D-pad sequence on 1.1.5 reported 627 rendered frames, 310 missed frame deadlines (49.44%), median 57 ms and 90th percentile 69 ms in `dumpsys gfxinfo`. Frame timing was recorded on the TV itself and is independent of ADB/Tailscale latency.
A separate same-signature `PerformanceInstrumentation` diagnostic isolates the scene and tile shadows. It requests the same first-six-tile sequence 36 times with 350 ms between selections; this includes moving to/from the next row. Android `FrameMetrics.TOTAL_DURATION` includes renderer/queue waits and differs from the histogram summary in `gfxinfo`. These are comparable A/B samples from the same TV, rather than benchmarks across different devices:
| Rendering variant | Frames | Median | 90th percentile | Frames over 16.7 ms |
| --- | ---: | ---: | ---: | ---: |
| 1.1.5 unchanged | 433 | 85.97 ms | 89.50 ms | 433 (100%) |
| 1.1.5 background hidden | 366 | 9.22 ms | 10.46 ms | 0 |
| 1.1.5 tile shadows removed, background restored | 433 | 84.90 ms | 87.39 ms | 433 (100%) |
| Precomposed background, original shadows | 638 | 14.40 ms | 25.79 ms | 185 (29.0%) |
| 1.1.6 background and focused shadow enabled | 639 | 12.99 ms | 17.00 ms | 64 (10.0%) |
| 1.1.6 shadows removed for diagnosis only | 640 | 12.67 ms | 13.59 ms | 0 |
The full-screen shader/transparency passes were the dominant bottleneck on this renderer. Removing redundant shadows after fixing that bottleneck reduced the remaining tail latency. This supports an application rendering optimization issue exposed by this TV's graphics implementation; there is no evidence here that replacing the TV or adding RAM is necessary. This is a short controlled navigation sample, not a claim of universally perfect 60 fps.
Version 1.1.6 composes the original artwork, accent glow, darkening, and gradient protection on one background worker into opaque textures no larger than 960 × 540. Each texture is about 1.98 MiB. The worker has one pending-request slot, discards stale results, and times out after two idle seconds. Navigation crossfades the finished textures rather than regenerating full-screen shaders each frame. Only the focused tile casts a shadow, and its parent no longer casts a duplicate shadow. Focus enlargement, smooth scrolling, app-colored accents, and user preferences remain enabled. This follows Android's guidance on [reducing overdraw](https://developer.android.com/topic/performance/rendering/overdraw).
The enhanced wake regression passed cold launch, interrupted loading followed by pause/resume, memory trim recovery, existing-image reuse, partial-grid recovery, composed-texture resolution/nonblank checks, release of all composed/source textures on stop, screensaver dismissal resuming rendering, and five stop/start cycles. No preferences were rewritten by the diagnostic. Temporary test APKs are not part of the published launcher.
Raw local captures are in `output/performance/` (ignored by Git), including the original screenshot, A/B timings, final navigation frames, memory/idle samples, and final screenshot.
The original D-pad workload was then repeated against the final installed APK: **853 frames, 12 missed deadlines (1.41%)**, median 13 ms, 90th percentile 24 ms, and no slow bitmap uploads. The original missed-deadline rate was 49.44%. Screenshots retain the mountain artwork, focus enlargement, five-column layout, and partial second row. The difference from the isolated benchmark reflects using actual D-pad input and more repeated scrolling.
Foreground PSS was 43,742 KiB (42.7 MiB), and background PSS was 35,803 KiB (35.0 MiB). The launcher consumed zero CPU ticks over five seconds while Android settings was foreground. Source/composite caches are cleared when stopped; actual heap reclamation remains Android's responsibility. These short samples do not establish overnight endurance.
The TV's Home override was already off (no enabled accessibility services) and was retained. A physical Home-key injection returns to Google Home on this TV; opening Luma directly and repeated Back stays in Luma. No default-launcher, picture, power, resolution, or accessibility settings were changed. The optimized APK was installed only on the authorized TV at `.200`; the onn was not operated during this investigation. Version 1.1.6 is published in the existing Git update feed for the other devices.
## Further focus optimization and onn comparison — October 19, 2026
The onn was still running 1.1.5. After waking it and allowing its images to settle, the same 42-key D-pad workload reported 603 frames, 34 missed deadlines (5.64%), median 20 ms, 90th percentile 23 ms, and 95th percentile 24 ms. Thus the integrated TV's 1.1.6 results were already comparable to the working onn installation, with a lower median and missed-deadline rate. This compares the devices' actual installed versions and app grids; it is not a same-version hardware ranking. The onn's package, preferences, and Home override were not changed; its original standby state was restored after measurement.
On the TV, the 1.1.6 diagnostic showed that the remaining native focused-tile shadow increased tail latency. Version 1.1.7 replaces it with one shared 192 × 122 ARGB soft-shadow texture (93,696 bytes), tinted for each app and faded over the existing 160 ms focus transition. The existing 6.5% focus scale, glow, background colors, and scrolling remain enabled. This avoids animating live elevation-shadow geometry and does not require reducing image resolution further or removing animations.
The same TV-local focus benchmark with the new shadow enabled produced **640 frames, median 13.48 ms, 90th percentile 14.31 ms, and zero frames over 16.7 ms**. Version 1.1.6's corresponding sample was 639 frames, median 12.99 ms, 90th percentile 17.00 ms, and 64 frames over 16.7 ms. The gain is consistency/tail latency rather than a faster median. Disabling the cached shadow for diagnosis barely changed the result (median 13.33 ms, 90th percentile 14.10 ms), indicating its remaining rendering cost is small.
The enhanced wake regression passed with the cached-shadow build, including checking that its animation stops on leaving Home, nonblank bounded backgrounds, interrupted-load recovery, screensaver dismissal, and five stop/start cycles. An initial attempt began while the TV was asleep; later ADB disconnects interrupted streamed test reports. Running the diagnostic locally with saved output while the TV was awake completed successfully. The temporary diagnostic APK is removed after verification.
The broader 42-key D-pad sequence on 1.1.7 recorded 861 frames, 16 missed deadlines (1.86%), median 14 ms, 90th percentile 25 ms, and 95th percentile 25 ms. This is roughly level with the 1.1.6 sample (1.41%, 13/24/26 ms); the cached shadow's isolated-focus consistency gain did not produce a measurable improvement to the whole navigation workload in these samples. It is not presented as a new overall performance leap. Screenshots confirm the soft app-colored shadow, scale/outline, wallpaper, and spacing remain visible. The remaining navigation time is primarily GPU/compositing work; main-thread draw/traversal was already sub-millisecond in the sampled 1.1.6 trace. The onn comparison remains 1.1.5, rather than claiming a same-version hardware comparison.
## Portable tile optimization — October 19, 2026
Version 1.1.8 uses the same default rendering path and settings on every device. No model, manufacturer, product, IP address, or device-name selection is used. App banners are marked opaque after composition over an opaque background. Rounded app artwork uses a cached bitmap shader and matrix, avoiding repeated path clipping and redundant background draws. Transparent artwork retains its underfill and media progress bars retain their rounded clipping. The shared focus shadow is now ALPHA_8: 23,424 bytes instead of 93,696 bytes. Focus enlargement, color glow, 160 ms focus transitions, background transitions, and smooth scrolling remain enabled.
The onn and integrated TV were first brought to the same released 1.1.7 baseline. The following comparisons use a same-signature diagnostic with real D-pad key injection, Android FrameMetrics, and identical 42-key sequences across the app rows. Normal navigation waits 350 ms between keys; rapid navigation waits 80 ms. This avoids the extra process/VM startup caused by invoking Android's shell `input` command for every key. Metrics were collected locally on each device, rather than timing the ADB connection.
| Device / pace | 1.1.7 median / p90 | 1.1.8 median / p90 | Missed deadlines, baseline → optimized |
| --- | ---: | ---: | ---: |
| onn, normal | 12.09 / 15.08 ms | 10.49 / 13.30 ms | 8/798 → 7/801 |
| onn, rapid | 11.60 / 14.89 ms | 10.76 / 13.63 ms | 3/254 → 1/255 |
| TV, normal | 25.03 / 26.75 ms | 13.28 / 24.24 ms | 32/828 → 17/864 |
| TV, rapid | 25.36 / 26.91 ms | 24.39 / 25.17 ms | 28/212 → 25/210 |
The TV improves substantially in normal-navigation median duration, but its rapid-navigation GPU cost remains greater than the onn's. These are short samples of each device's existing app grid, not proof of identical hardware performance or perfect 60 fps. FrameMetrics total duration and deadlines are reported directly; they should not be mixed with earlier gfxinfo histogram values or an isolated first-six-tile benchmark. All samples above had zero dropped FrameMetrics callbacks. Raw reports are saved locally under `output/performance/round2/` and are ignored by Git.
**Settings → Appearance → Memory saver** is available on every device and is **off by default**. It separates accent tint into a bounded alpha mask, retaining the same composed background when only the selected app's color changes. This reduces repeated full-image allocation/uploads, with a different GPU tradeoff. Earlier shell-workload onn PSS samples were 132,417 KiB for 1.1.7, 109,413 KiB for optimized tiles with the default background, and 67,578 KiB with Memory saver and optimized tiles. PSS is variable, includes framework/graphics overhead, and is not a guaranteed reduction on every device.
With Memory saver enabled, the same native-input test measured normal/rapid p90 durations of 13.03/14.88 ms on the onn and 24.44/24.81 ms on the TV. Missed-deadline counts were 12/681 and 1/259 on the onn, and 20/732 and 27/208 on the TV. It is not consistently faster than the default, so it is an explicit preference rather than an automatic device classification. Both modes work on both devices. GPU-specific single-pass shader experiments were discarded after they regressed performance.
GPU PixelCopy checks passed on both devices for wide opaque banners, portrait center cropping, transparent artwork, media progress bars, and rounded corners. Sampled tile interior differences against the previous drawing implementation were at most 1/255 per color channel. Clearing a tile image also clears its cached shader reference. Memory-saver backgrounds matched the original composited colors within 2/255, and artwork crossfades at 25%, 50%, and 75% matched within 3/255. Accent changes reused the existing background and mask.
Wake regression passed on both devices in both rendering modes: cold launch, interrupted loading/pause/resume, memory trim, existing-tile reuse, partial-grid recovery, bounded nonblank backgrounds, texture/queue release on stop, screensaver dismissal, and five stop/start cycles. Memory-saver checks also verified loading the saved preference and switching back and forth at runtime. Temporary preference changes were restored by the test runners. The first tile-comparison test had an uninitialized reference bitmap in the diagnostic overlay; fixing that test setup allowed the comparisons to pass without changing production rendering.
The final signed 1.1.8 APK was installed on both devices and retains the previous certificate. Final screenshots show the original five-column spacious layout, mountain background, rounded banners, and app-colored focused shadow. Temporary diagnostic packages were uninstalled. The onn Home helper was re-bound after instrumentation using its original enabled-service list, retaining Button Mapper; Home from Android settings and repeated Back returned/stayed in Luma. The TV's Home override remains off with its original empty service list. The previous launcher remains installed and enabled.
With Android settings foreground, the onn UI and Home helper each consumed zero CPU ticks over five seconds. The TV initially consumed five CPU ticks over five seconds after installation; a later five-second sample after settling consumed zero. These short samples show idle behavior after pending work completes, rather than establishing overnight endurance. Both devices were returned to Luma after verification.
## Focus-color coalescing — October 19, 2026
An earlier review of another launcher's public release notes motivated testing fewer unnecessary accent transitions. The available material did not expose that launcher's rendering source, so no implementation was copied or inferred. Luma's changes were validated against its own code and device measurements.
Version 1.1.9 retains immediate app-tile focus scale/glow/labels, but coalesces app accent changes until 160 ms without another different color. It then uses the existing 280 ms background transition. Reduced motion and initial rendering apply immediately. A single reusable callback retains only the latest color, is removed on pause/screensaver/stop, and restores the selected color when resumed. Both devices use the same code and behavior. Bitmap resolution, image quality, and the spacious layout are unchanged.
Fresh same-device 1.1.8 versus candidate comparisons used the native-input 42-key NavigationInstrumentation workload described above, with no shell input process per key. The test ran normal navigation (350 ms between keys), then rapid navigation (80 ms between keys), on each device's existing grid:
| Device / pace | 1.1.8 median / p90 | 1.1.9 median / p90 | Missed deadlines, baseline → optimized |
| --- | ---: | ---: | ---: |
| onn, normal | 10.02 / 13.29 ms | 9.90 / 12.92 ms | 5/800 → 10/737 |
| onn, rapid | 11.02 / 13.61 ms | 10.76 / 14.02 ms | 1/255 → 1/252 |
| TV, normal | 16.52 / 24.34 ms | 12.53 / 24.30 ms | 18/853 → 18/799 |
| TV, rapid | 24.37 / 25.07 ms | 20.64 / 21.56 ms | 25/214 → 21/229 |
On the TV, rapid-navigation median/p90 duration improved about 15%/14%; GPU p90 fell from 20.62 to 17.29 ms. On the onn, frame durations remained roughly level and the normal-navigation missed-deadline count increased in this short sample. This is a reduction in unnecessary background work and an improvement on the TV, not a consistent frame-time improvement on every device. TV rapid frames still exceed a 16.7 ms budget. All captures had zero dropped FrameMetrics callbacks. These samples are not an endurance test or a hardware ranking; raw results are under ignored `output/performance/round3/`.
The new scheduling regression passed on both devices in both rendering modes. Three colors 60 ms apart produced exactly one final full-color composition (or reused the existing Memory-saver textures), retained the final selected color, canceled pending work on pause, restored on resume, and applied reduced-motion changes immediately. The complete wake regression also passed on both devices in both modes: interrupted loads, partial-grid recovery, memory trimming, image reuse, nonblank bounded scenes, release on stop, screensaver dismissal, five stop/start cycles, and runtime renderer switching. Diagnostic preference changes were restored.
The final signed 1.1.9 build was installed on both devices. Screenshots confirm the wallpaper, app banners, focus appearance, and original layout. After settling with Android settings foreground, both launcher UIs and the onn Home helper consumed zero CPU ticks over five seconds. Home/Back returned/stayed in Luma on the onn; the TV retained its original disabled Home override. The onn's original accessibility list was restored after instrumentation, including Button Mapper. Both diagnostic APKs and the round's temporary on-device reports/scripts were removed; the previous launcher remains installed and enabled on both devices. Both devices were returned to Luma.
## Capability-based TV input button — October 19, 2026
Version 1.1.10 adds a focusable input icon immediately left of Settings. It is shown only when Android's TvInputManager exposes visible physical passthrough inputs. The code uses public input types/metadata, custom labels, hidden-input settings, and connection states; it contains no model/manufacturer/IP checks. CEC children whose physical parent is already listed are omitted. Tuners and virtual/app-only inputs do not create the button. Discovery runs on the existing worker, and input-change events are coalesced. The listener and pending refresh callback are removed on pause; stale results are discarded. No new Android permissions are requested.
On the onn, the system exposed no physical input entries and UIAutomator confirmed the header contains Settings with no input button. The existing header spacing remains intact. On the integrated TV, UIAutomator and screenshots confirmed the input icon, focus colors, D-pad Up to Settings then Left to inputs, and a picker with HDMI 1–4 plus Cancel. The connected CEC device is not duplicated as a fifth port.
The implementation discovers an exported/accessible standard channel-view handler through its MIME filter, excluding generic content browsers. Where no such handler is available, a private input activity uses Android's TvView and the selected passthrough URI. The TV did not expose a native channel-view activity to the active user during testing, so the TvView fallback was exercised without enabling/reinstalling any system app. The viewer requests media audio focus while visible, resets its input session and abandons audio focus on pause, and returns to Luma on Back.
Selecting HDMI 1 created a real TV-input session for Luma's UID with the expected physical input ID. The video-available callback hid the loading text. Selecting HDMI 2 created the corresponding port session and displayed “No signal · Press Back to return.” HDMI 1 was then selected again; its video-available callback fired, and Back returned to the launcher with artwork restored. TV-input dumps after returning confirmed Luma's session was released. The pre-existing provider session was retained. Live HDMI pixels were not available in the ADB screen capture; verification used input session IDs, video callback state, and the UI hierarchy. Audio was not listened to remotely.
The enhanced wake regression passed on both devices, including interrupted-load recovery, pause-only image reuse, partial-grid recovery, memory trimming, screensaver dismissal, nonblank bounded backgrounds, five stop/start cycles, and an assertion that the TV-input listener is removed when paused. This change makes no rendering-performance claim and does not change image resolution or the app grid layout.
The platform interfaces are documented in [TvInputManager](https://developer.android.com/reference/android/media/tv/TvInputManager), [TvInputInfo](https://developer.android.com/reference/android/media/tv/TvInputInfo), and [TvView](https://developer.android.com/reference/android/media/tv/TvView).
Final 1.1.10 package/version checks passed on both devices, with the input button visible only on the TV. After settling with Android settings foreground, both UIs and the onn Home helper consumed zero CPU ticks over five seconds. The diagnostic APKs and temporary on-device files were removed. The onn's original Home/accessibility service list was restored after instrumentation; Home returned from Android settings, and Back stayed in Luma. The TV's existing Home override remains off. The previous launcher remains installed/enabled on both devices, and both devices were returned to Luma. Raw local checks are in ignored `output/performance/input-switching/`.
## Warm Home returns and Plex artwork reuse — October 19, 2026
Version 1.1.11 keeps two reductions in unnecessary work, with the same defaults and rendering implementation on both devices:
- Reuse the app grid when its ordered packages, labels, launch components, APK source paths, column count, and tile widths are unchanged. Rediscover installed apps and rebind listeners to the new models. Layout changes rebuild the grid; icon completions check the current binding generation. A paused focus animation is normalized before reuse.
- Reuse an already displayed Plex image and composed scene for the same artwork path, including Plex/Plezy selections sharing that artwork. Keep the previous image while a different one loads; a failed replacement retains that image. Coalesce identical pending requests, cancel obsolete delayed callbacks, reject stale completions, and retry the focused artwork after resume. Changing the Plex connection or selecting another app clears the previous source appropriately.
Images and image caches are still released on stop. Retained views already existed while backgrounded before this change; this avoids replacing them on return rather than retaining additional bitmap memory. Image resolution, layout, animation timing, and the optional Memory saver default are unchanged. There are no device/model/manufacturer/IP branches.
`WarmReturnInstrumentation` invokes eight complete pause/stop/start/resume cycles, waiting for all app icons and the scene to be present. The first two cycles are excluded from the medians below. Allocation is the target process's ART Java allocated-byte counter during resume-to-image-ready, including concurrent target-process work. This is a controlled lifecycle workload, not physical app-switch latency, peak RAM, or a GPU allocation measurement.
| Device | 1.1.10 Java bytes per return | 1.1.11 Java bytes per return | Reduction | Image-ready median, before → after |
| --- | ---: | ---: | ---: | ---: |
| onn, 21 app tiles | 344,776 | 197,320 | 42.8% | 182 → 181 ms |
| Integrated TV, 14 app tiles | 295,816 | 197,512 | 33.2% | 167 → 165 ms |
All tiles were reused in every candidate cycle. Image-ready times are effectively unchanged; the demonstrated benefit is less Java allocation and view construction. A prior candidate run gave 184 ms on the onn and 165 ms on the TV, consistent with that interpretation.
Four repeated selections of the same featured Plex app produced eight extra scene compositions on 1.1.10 on each device. The candidate produces zero, retaining both the same artwork bitmap and the same finished background frame. `PlexArtworkInstrumentation` uses an authenticated loopback-only HTTP fixture running inside the test process, with the bundled wallpaper as fixture artwork. It requires no home Plex server or network connection. Original encrypted credential preferences and the media-row setting are restored in finally; no real credential values are printed. Earlier host-forwarded fixture attempts were interrupted by the TV's intermittent ADB connection, so the final fixture runs on-device.
`GridReuseInstrumentation` verifies label changes retain views, rebound focus callbacks retain selection, interrupted focus lifts finish at the correct scale, and hiding/reordering/changing columns produces the correct grid. Layout preferences are restored after testing. The complete wake suite passes with default rendering and Memory saver on both devices, including partial/interrupted loads, pause-only image reuse, memory trimming, screensaver dismissal, five stop/start cycles, runtime renderer switching, and removal of TV-input listeners while paused.
An exact-pixel background damage experiment was rejected: the bounding rectangle of changed RGB pixels still covered 93.1–97.6% of the screen across tested accent pairs, while scanning the bounded 960 × 540 image cost roughly 12–14 ms after warmup. No clipping or image-quality approximation was added to the launcher. The already rejected custom GPU shader from the earlier round remains absent.
The final signed-build comparison reinstalls 1.1.10, runs the native 42-key normal/rapid navigation workload, installs 1.1.11, and repeats it. The device runs the script locally so ADB connection interruptions do not interrupt the measurements:
| Device / pace | 1.1.10 median / p90 | 1.1.11 median / p90 | Missed deadlines, before → after |
| --- | ---: | ---: | ---: |
| onn, normal | 10.44 / 13.35 ms | 10.25 / 13.61 ms | 14/740 → 10/741 |
| onn, rapid | 11.01 / 14.84 ms | 11.36 / 15.00 ms | 1/256 → 1/255 |
| TV, normal | 23.87 / 24.72 ms | 23.85 / 24.72 ms | 35/803 → 32/807 |
| TV, rapid | 24.06 / 24.78 ms | 24.05 / 24.83 ms | 27/217 → 27/219 |
These show essentially unchanged navigation frame times rather than a consistent frame-rate gain. Earlier candidate samples fluctuated: TV rapid p90 was 24.82 ms in one run and 40.91 ms in another; the final consecutive baseline/release comparison returned to about 24.8 ms. onn candidate rapid p90 ranged from 14.37 to 15.22 ms versus the initial baseline's 14.49 ms. All captures had zero dropped FrameMetrics callbacks. FrameMetrics totals include graphics-pipeline waits; these samples neither establish a hardware ceiling nor compare identical installed-app grids across devices. The changes are retained for the demonstrated reduction in allocations and redundant scene composition, without claiming to solve the TV's remaining graphics-pipeline timing.
The final 1.1.11 APK retains the existing signing certificate and was installed on both devices. The self-contained Plex fixture passes on both, including latest-request selection, failed-image retention, cancellation on pause, automatic focused-artwork retry on resume, rejection of obsolete selections, and image release/restoration across stop/start. Raw local reports are under ignored `output/performance/round4/`.
Final screenshots confirm wallpaper, app artwork, focus styling, and the original spacious layout on both devices. UI hierarchy checks show the input button on the TV only. With Android settings foreground after settling, each launcher process consumed zero CPU ticks over five seconds. Back from Luma stayed in Luma on both. The previous launcher remains installed and enabled. The enabled accessibility lists match the start of this round: Button Mapper only on the onn, and null on the TV. Both system Home resolvers point to Google TV; the onn's Luma Home override was already off at this round's initial capture and remains off. This round preserves that configuration rather than enabling a helper that was absent initially. Both devices were returned to Luma. Temporary diagnostics, reverse ports, and this round's on-device APKs/scripts/reports were removed after validation.
## Launcher integration removal and current-jank investigation — October 19, 2026
At the user's request, the previous alternate launcher was uninstalled from both devices. Luma's dedicated shortcuts, special launch handling, explicit fallback broadcast/preferences, accessibility package subscription/redirect, and explanatory UI strings for that integration were removed. Home-key handling, Google Home redirection, Android settings access, and the user-controlled override remain. No other installed launcher or accessibility service was removed. Earlier validation entries above describe historical configurations, not a requirement to keep another launcher installed.
The supplied benchmark reports use Android's current `Janky frames` counter from gfxinfo. That differs from the legacy counter and from FrameMetrics total-duration thresholds used in prior development notes. The reported TCL gap was reproduced after uninstalling the other launcher; removal alone did not fix it. A neighboring-tile comparison also differed in workload: the onn's Plex/Plezy pair shares an accent, while the TV's Plex/Spotify pair changes it. A second comparison therefore used Plex/Spotify on both devices, the same Alpine theme and dim level, real alternating D-pad keys at a scheduled 200 ms cadence, and no Plex network/media rows.
`JankInstrumentation` resets gfxinfo immediately before each 60-key sequence and reads its current jank counter after settling. It temporarily changes only the test order, media-row setting and accent, restoring those keys in finally. Memory-saver isolations change the scene at runtime without changing the saved preference. Each trial executes once on-device and saves its report locally; retrieval does not replay navigation. The first two apps are asserted in matched tests. Tile contents elsewhere still differ. The test does not measure input-to-photon latency.
Three-trial baseline medians after the uninstall were 0.76% current jank on the onn's original neighboring tiles and 8.50% on the TV's original neighbors. With the same Plex/Spotify pair, medians were 0.14% and 8.36%. The workload difference matters, but does not explain away the TV gap. Removing the focus shadow left TV median jank at 8.25%, so the shadow was retained. Hiding the entire background gave a 0% median on both; that diagnostic loses the intended visual design and was not shipped. A fixed accent gave TV samples of 1.67%, 7.83%, and 0%, so it is not a consistently faster mode claim.
Perfetto captures included FrameTimeline, main/RenderThread slices, scheduling, and graphics events. In the retained old-delay trace window, mean RenderThread dequeueBuffer duration was 7.36 ms on the TV versus 3.36 ms on the onn. Mean DrawFrames duration was 13.42 versus 8.30 ms. Mean main-thread doFrame was 6.23 versus 1.85 ms, but included 4.75 versus 0.89 ms waiting in postAndWait; recording View drawing itself averaged roughly 0.25–0.30 ms. These are wall durations including waits, not CPU execution times.
The TV's retained FrameTimeline sample contained 573 buffer-stuffing/late-present frames finishing on time and one app-deadline-missed frame. Its retained SurfaceFlinger frames finished on time. The onn also had buffer stuffing with on-time app completion. Those states are not the same as gfxinfo's current-jank percentage: [Perfetto describes buffer stuffing as queued frames that can remain smooth while adding latency](https://perfetto.dev/docs/data-sources/frametimeline). This evidence points toward buffering/render-pipeline waits, rather than proving excessive Java UI work or a hard GPU-compute ceiling. Neither device exposed GPU render-stage slices in these captures, so the GPU/driver/display contribution cannot be completely separated. The initial 32 MiB ring traces retain only part of the input window; TV process-name metadata was incomplete and its target PID was verified against the corresponding gfxinfo report. A larger 64 MiB configuration with periodic process stats is included in `tests/device/jank-trace.pbtxt` for reproduction.
The retained software change extends the accent quiet period from 160 to 250 ms. It avoids starting full-screen transitions for passing selections at a 200 ms cadence, while preserving immediate 160 ms tile focus and the existing 280 ms final background crossfade. The explicit tradeoff is 90 ms more delay before an animated app-color change begins after focus settles. Initial rendering, reduced motion, image resolution, blur/composition quality, and all device-independent defaults are unchanged. Slower navigation can still trigger expensive background transitions; this does not establish universally stutter-free rendering.
Testing also exposed a redundant Memory-saver animation: coalescing a move away and back could animate a color to itself. The final build cancels an obsolete color animation and returns immediately if the displayed color already matches. The scheduling regression now uses 200 ms passing selections and verifies that returning to the displayed color starts neither composition nor a color animator, in both rendering modes.
The trace also exposed repeated window relayouts at remote input: the old timer reset reapplied FLAG_KEEP_SCREEN_ON even when unchanged. The final code checks the actual window flag before changing it. This removes redundant window-attribute dispatch from the per-key path and preserves real flag changes for timeout settings, screensaver entry/exit, and pause/resume. `WindowFlagsInstrumentation` wraps the Window callback and verifies twenty interactions produce zero attribute-change callbacks; switching the timer Off/On produces one change each, and pause/resume releases/restores the flag. It restores the original timer preference and callback.
A same-version diagnostic control APK containing the integration removal but the old 160 ms renderer was used to capture the control trace and three Memory-saver-On trials without downgrading an installed package. It was not published. The Memory-saver-Off baseline medians use the original 1.1.11 renderer. The final release was installed over all temporary controls. Profiling captures are separate from the untraced median trials.
Final untraced matched-pair results (three-trial medians, current gfxinfo jank):
| Device / Memory saver | Old rendering current jank → 1.1.12 | Reported duration p90, old → new |
| --- | ---: | ---: |
| onn / Off | 0.14% → 0.30% | 11 → 10 ms |
| onn / On | 0.14% → 0.76% | 13 → 12 ms |
| TV / Off | 8.36% → 0.00% | 24 → 9 ms |
| TV / On | 6.94% → 0.00% | 23 → 12 ms |
TV Off trials were 0.15%, 0%, and 0%; On trials were 0%, 0%, and 0%. onn Off trials were 0.45%, 0.30%, and 0.30%; On trials were 1.21%, 0.30%, and 0.76%. The onn current-jank counter increased slightly despite a lower duration p90; this is not a universal jank improvement. A preceding onn crossover with only the longer delay gave 0.14%/0.14% on the old delay and 0.76%/1.21% on the candidate, so the small onn increase is explicitly retained in the evidence. The final code also includes the window-update guard. It is retained for the large repeatable TV improvement and elimination of redundant work, while preserving the onn's existing Memory saver Off setting. These are short, ordered trials rather than a randomized endurance study. Zero measured janky frames does not promise zero stutter in other workloads.
The final 64 MiB traces cover startup and the complete input sequence. Only the initial window relayout remains on either device; the per-input relayout slices disappear. Mean dequeueBuffer time in these captures was about 0.24 ms on the TV and 0.18 ms on the onn, compared with 7.36/3.36 ms in the retained control samples. TV mean DrawFrames fell to 5.12 ms and mean doFrame to 1.39 ms; most app FrameTimeline slices were on-time rather than buffer-stuffed. Startup still contributes a few late frames in this trace, unlike the untraced counters reset after settling. Trace overhead and different retained windows prevent treating these means as exact speedup ratios, but they corroborate removal of the repeated window/queue work. The final UI-only subtitle correction does not change these rendering paths.
The broader 42-key normal/rapid navigation regression also passed on both devices. Final FrameMetrics median/p90 totals were onn 10.28/13.50 ms normal and 10.67/13.37 ms rapid; TV 12.50/13.58 ms normal and 23.92/24.45 ms rapid with its saved Memory-saver setting On. These totals include pipeline waits and are not the current gfxinfo jank counter. In particular, some rapid TV frames remain longer even though the matched 200 ms workload's current jank is much lower. All samples had zero dropped callbacks.
Both devices passed the final 200 ms accent/coalescing regression, same-color cancellation, reduced motion, pending-work cancellation on pause, full wake recovery, pause-only image reuse, memory trimming, screensaver dismissal, five stop/start cycles, and runtime renderer switching. The original app/timer/media/accent/order preferences were restored. The final signed version is 1.1.12, versionCode 15, using the existing certificate. Full raw reports, SQL summaries, and traces remain in ignored `output/performance/round5/`.
Final UI checks verified the removed package is absent on both devices, Settings and Default launcher contain no obsolete integration text/actions, Back stays in Luma, and physical-input controls remain visible only on the TV. The installed APK hashes match the final signed payload. Wallpaper, app art, original spacing, and focus behavior remain intact. Enabled accessibility lists remain Button Mapper only on the onn and null on the TV; the existing Home override remains Off on both. Memory saver remains Off on the onn and On on the TV. The supplied benchmark documents in the repository root were left untouched and untracked.
After settling with Android settings foreground, both launcher processes consumed zero CPU ticks over five seconds. This round's temporary diagnostic APKs, scripts, on-device reports, locks, and trace files were removed after archiving the reports locally. Both devices were returned to Luma. Android's stock launcher and unrelated apps/settings remain unchanged.
## 1.1.13 app actions and network shortcut (2026-10-25)
Built on Windows with JDK 17 and Android SDK/build tools 36, retaining minSdk 26 and targetSdk 34. VersionName 1.1.13 / versionCode 16. The original signing key was missing; the user explicitly authorized a replacement identity and a reinstall. The new RSA certificate SHA-256 is `5982677675475eab76e8aaa52f939b7a563aef69e869b8c5e342ac9e921b12ad`. The key and randomly generated signing password remain under ignored `.tools/`; neither is published. Build scripts now pass signing passwords through environment variables and refuse automatic key replacement.
On the authorized onn 4K Plus (Android 14/API 34), `scripts/check-app-move.py --adb <adb> --device <serial>` failed against installed 1.1.12: Right after selecting Move in grid left the first three visible apps in the same order. It passes against 1.1.13. The cause was mutating `visibleApps` before `renderApps` captured the displayed order, so the reuse comparison incorrectly skipped rebuilding. The fix swaps a separate order list, and preserves the requested selection while replacing focused children.
The final `HomeActionsInstrumentation` passed menu entry with remote OK, all four directions, upper/left boundaries, focus remaining on the moving app, Back restoring order, OK saving it across stop/start, and preference restoration. It also passed offline presentation followed by live-network recovery, callback stop/restart, absence of Personalize, and header bounds/D-pad navigation with simulated physical-input availability. The test exits Android touch mode with a remote arrow before requesting focus, so prior tap-based checks do not consume the first OK merely to leave touch mode.
`GridReuseInstrumentation` passed label-only reuse, focus callbacks, interrupted focus animation, hiding, reordering, and column changes. `WakeRegressionInstrumentation` passed interrupted loading, pause-only recovery, memory trimming, screensaver dismissal, and five stop/start cycles. Normal/rapid `NavigationInstrumentation` passed with zero dropped callbacks; FrameMetrics p95 was 16.37 ms normal and 16.64 ms rapid. These are functional navigation samples on one device, not a performance improvement claim or a comparison with another GPU.
The real UI check `scripts/check-home-actions.py` opened the native TV network page. Canceling Android's uninstall confirmation preserved the disposable Luma test app; confirming removed it for the current Android user and the resumed Luma library refreshed. The test checks the current user explicitly because the device has a second Android profile. System apps retain app info and do not receive an uninstall action. No user app was uninstalled by these checks.
Physical Ethernet, captive-portal/limited-internet hardware states, and real physical-input switching were not exercised in this run. Ethernet has its own vector port icon and accessible label selected by Android's active-network transport. Disconnected presentation was injected through the UI update seam without disconnecting wireless ADB. Input availability was simulated only for header bounds and focus; no input hardware or second TV was modified.
The signing migration required a reinstall of the authorized test device. The visible five-column layout, Kaleido-first ordering, Scenic screensaver with clock, and previously downloaded nature video were restored through Luma UI. Existing versions on the other two devices require a one-time reinstall; signature and update integrity checks remain enabled. Future releases must retain the new key.
Final cleanup restored the prior enabled-accessibility list (Button Mapper plus Luma's own Home override) and notification permission. Home returned from native Android settings to Luma after its existing accessibility redirect. Android's native Home selection and other launchers were preserved. The diagnostic APK was removed from all profiles, temporary on-device captures were deleted, and the installed APK SHA-256 matched the release manifest. The final Home screenshot was visually checked: the Wi-Fi icon has separate space beside the gear and the five-column app layout is intact.
Before publication, the Wi-Fi/Wi-Fi-off vectors were replaced with standard Material paths to correct uneven band spacing. The Unicode Settings glyph was replaced with the matching filled Material Settings vector and the same ImageButton padding/tint/focus treatment as the network button. Both were visually checked in a fresh TV screenshot. HomeActionsInstrumentation passed again after the Settings view-type change, including header focus and non-overlap; clicking the new gear opened Luma settings. Source attribution and the Apache 2.0 license are retained in docs/ASSETS.md and docs/licenses/.