H700 port 08 Testing status

08 — Testing Status

This document is the canonical H700 test report. It records what was tested, the result, and the device that supplied the evidence. It does not decide whether a test blocks Alpha, Beta, RC, or Release; those decisions live in 09-roadmap. Detailed per-core coverage lives in 10-core-game-matrix.

Results recorded through 2026-08-09. Artifact lineage: Alpha → h700-beta1 → h700-rc1…h700-rc8. h700-rc8 is GM (rebuild on the NextUI v6.14.0 commit to keep RetroAchievements working; see 09).

Status and evidence rules

Status	Meaning
✅ Pass	The stated test completed successfully
⚠️ Partial	Useful behavior passed, but the row has a known issue or incomplete coverage
⬜ Not run	No completed result is recorded yet
➖ Not applicable	The behavior is unsupported or guaranteed by design rather than a runtime test
❌ Fail	The stated test failed

The H700 devices share one platform implementation, SoC, stock-OS integration, and core runtime. A successful shared-runtime test on one representative H700 device counts for the family. Repeat it on another model only when panel geometry/rotation, lid, controls, connectors, firmware, or model-specific code differs, or when evidence suggests a model-specific regression. A named device is the evidence source, not a limit on support.

Test fleet

Device	Evidence	Hardware coverage
RG40XXV	Probed live	640×480, one stick, HDMI, Bluetooth audio; family baseline
RG34XXSP	User-tested	720×480, dual sticks, clamshell lid
RG28XX	User-tested	480×640 portrait panel and rotation
RGcubexx	Community-tested	720×720 general UI scaling only
TrimUI Brick TG5040	Probed live	Shared-code regression reference

Build and binary compatibility

ID	Test	Status	Evidence/device	Result and notes
BUILD-01	CI-style H700 staging build (make setup, then make h700)	✅	CI/release build	H700 binaries, tools, cores, libraries, boot shim, and installer assets stage successfully
BUILD-02	Full tg5040 regression build after shared-code changes	✅	TrimUI Brick / shared h700-rc8 artifact	Same candidate runs on tg5040. Users report LedControl settings do not stick on TrimUI in this version (#23); H700 LED path is separate (LED-01…), treat as a tg5040 regression to track rather than an H700 build failure
BUILD-03	Jammy settings.elf runtime linkage check	✅	Automated container check	H700 library path resolves and required gio/glib linkage is present
BUILD-04	Bundled curl static linkage and HTTPS	✅	arm64 Jammy container	Pinned musl curl 8.21.0 + OpenSSL 3.5.7 has no dynamic section; DNS and TLS reach the RetroAchievements API using the bundled CA file (HTTP 422 is the expected unauthenticated response)

Boot, install, and launcher

ID	Test	Status	Evidence/device	Result and notes
BOOT-01	Cold boot to NextUI and 10+ reboot soak	✅	RG40XXV	Stable boot with reasonable startup time
BOOT-02	Stock-card hijack with the required old style theme	✅	RG40XXV/RG34XXSP	/mnt/mmc/dmenu.bin launches NextUI from TF2
BOOT-03	Repeated launcher crash limit	➖	Design / accepted	Five-crash cutoff then poweroff is implemented; deliberately not required as an induced E2E release gate for h700-rc8
BOOT-04	Remove dmenu.bin to restore stock boot	➖	Design inspection	Stock dmenu_ln cannot launch NextUI when the single hijack file is absent; no separate uninstall workflow exists
BOOT-05	Stock MU style theme behavior	✅	On-device observation	MU style bypasses /mnt/mmc/dmenu.bin and boots the stock/MU frontend; NextUI cannot display an in-shim warning

Display and user interface

ID	Test	Status	Evidence/device	Result and notes
DISP-01	Launcher, settings, keyboard, and in-game UI rendering	✅	RG40XXV, RG34XXSP, RG28XX	Clean rendering at 640×480, 720×480, and rotated 480×640
DISP-02	Brightness 0–10 and color-temperature behavior/persistence	✅	RG40XXV	Controls act and persist; fresh default brightness is 4
DISP-03	DisplayCal RGB gains, persistence, sleep, and game launch	✅	RG40XXV, RG34XXSP	Gains act and persist across reboot, sleep, and game launch
DISP-04	Dead enhance controls and Fn-switch settings hidden	✅	RG40XXV/code inspection	Contrast/saturation/exposure and the nonexistent RG XX Fn switch are capability-gated
DISP-05	Shaders, overlay pipeline, and effects on Mali-G31	✅	RG40XXV	All shipped GLSL paths tested; this tests the pipeline, not panel-sized overlay content
DISP-06	Box art, game switcher, and alpha-blended assets	✅	RG34XXSP	Clean at 720×480 after the 64-bit libpng fix
DISP-07	Screenshots	✅	RG34XXSP	Capture succeeds at 720×480
DISP-08	HDMI video hotplug and scaling	✅	RG40XXV + TV	Menu and game output, plug/unplug in both directions, repeated cycles, autosave/resume, and 1280×720 framebuffer scaling pass

Input and controllers

ID	Test	Status	Evidence/device	Result and notes
INPUT-01	Local controls in launcher, games, and menus	✅	RG40XXV, RG34XXSP, RG28XX	Raw evdev input path works across tested models
INPUT-02	MENU short tap versus hold	✅	RG34XXSP	Short tap opens shortcuts; hold changes brightness despite the firmware compound event
INPUT-03	Per-device stick layout, analog movement, and L3/R3	✅	RG40XXV, RG34XXSP	The Input pak correctly visualizes the RG40XXV left stick and the RG34XXSP left/right sticks, including movement and their L3/R3 clicks
INPUT-04	Rumble	✅	RG40XXV	moto on/off path works
INPUT-05	Bluetooth controller discovery, pairing, and attachment	✅	RG40XXV + DualSense	Discovers, pairs, trusts, connects, and exposes SDL joystick plus evdev; five restart cycles produced no duplicate daemon/attach processes
INPUT-06	Bluetooth controller semantic mapping	⚠️	DualSense/shared tg50x0 path	Attachment passes, but raw button indices vary by controller and are not normalized across platforms

Audio

ID	Test	Status	Evidence/device	Result and notes
AUDIO-01	Internal volume UI, mute, and levels from 0–100%	✅	RG40XXV, RG34XXSP	Audible levels and reversed digital volume scale behave correctly
AUDIO-02	Game audio and underruns	✅	RG40XXV, tested RG34XXSP games	GBA/SNES and broader gameplay have clean audio with no observed underruns
AUDIO-03	Headphone insertion and speaker routing	✅	RG34XXSP	Hardware auto-mutes the speaker and routes to headphones; NextUI does not expose a separate HP icon/volume
AUDIO-04	Bluetooth A2DP pairing and playback	✅	RG40XXV + AirPods 4 ANC	Pairing is remembered; manual Connect routes audible SBC audio; returning AirPods to the case restores the internal speaker. Nearby iPhone/Mac devices may win auto-connect. One post-sleep auto-repair report is expected-not-supported behavior pending triage (#20)
AUDIO-05	HDMI audio routing and restoration	✅	RG40XXV + TV	HDMI route is audible; unplug restores the panel/internal route

Network and online features

ID	Test	Status	Evidence/device	Result and notes
NET-01	WiFi scan, connect, and forget	✅	RG40XXV	NextUI-owned wpa_supplicant path works
NET-02	WiFi after sleep and NTP synchronization	✅	RG40XXV	Network recovers after sleep and system time synchronizes
NET-03	RetroAchievements login and unlock	✅	RG34XXSP	Credential entry through the on-screen keyboard and achievement unlock pass
NET-04	Pak Store install	➖	Scope	Not exercised for the H700 port
NET-05	OTA update flow	➖	Scope	Not exercised for the H700 port

Power, sleep, and battery

ID	Test	Status	Evidence/device	Result and notes
POWER-01	Manual/automatic deep sleep and wake	✅	RG40XXV, RG34XXSP	Repeated sleep/wake, in-game resume, and wake without the former ALSA delay pass
POWER-02	Fast resume and power-off auto-resume	✅	RG40XXV, RG34XXSP	Resume returns to the running game as designed
POWER-03	RG34XXSP lid close/open	✅	RG34XXSP	Close sleeps and open wakes. POWER may wake light sleep with lid closed; deep sleep re-sleeps while the lid remains closed
POWER-04	Charging indication and sleep while charging	✅	RG34XXSP	Indicator works; shared product behavior permits light/screen-off sleep and suppresses deep sleep while charging
POWER-05	Battery percentage versus stock	✅	RG34XXSP	Fresh-boot percentage matches stock; after long sleep the top-bar icon and Battery pak briefly disagreed with each other
POWER-06	Overnight suspend drain versus stock	✅	RG34XXSP	Sleep-drain work completed and verified against stock OS

Games and performance

See 10-core-game-matrix for the game-by-game ledger.

ID	Test	Status	Evidence/device	Result and notes
GAME-01	Browse, launch, play, menu, save/load, and clean exit	✅	RG40XXV, RG34XXSP	Primary system pak/core flow works end to end
GAME-02	GB/GBC/GBA/FC/SFC gameplay smoke	✅	RG34XXSP	Tested games launch and play correctly
GAME-03	MD/PicoDrive launch, gameplay, and Auto CPU	✅	RG34XXSP	Former slowdown near 480 MHz is fixed
GAME-04	PS1/PCSX-ReARMed launch and performance	✅	RG40XXV, RG34XXSP	Clean core rebuild fixed the stale/corrupted packaged artifact
GAME-05	FBNeo happy path and missing-BIOS error path	⚠️	RG34XXSP	With required BIOS, launch/play pass. Without BIOS, the upstream core stops polling frontend input and MinArch cannot open menu/exit
GAME-06	Remaining EXTRAS core smoke	⚠️	RG34XXSP (maintainer)	Maintainer-tested Pass: A2600, mGBA, SMS, and Pico-8 (Fake-08). Remaining long-tail systems in 10 are untested by the maintainer; other users may have exercised them — no specific bug reports so far. Known incomplete coverage, not a merge blocker
PERF-01	GBA/SNES full speed with vsync	✅	RG40XXV, tested RG34XXSP games	Full speed with no observed audio underruns
PERF-02	Frame pacing, tearing, and input latency	✅	RG40XXV	No visible issue during gameplay; result is observational, not instrumented
PERF-03	Manual CPU governor changes during play	✅	RG34XXSP	Powersave/performance changes do not stutter audio
PERF-04	Automatic governor load/frequency selection	✅	RG34XXSP	MD Auto CPU regression is resolved
PERF-05	Auto governor full-range scaling	✅	Representative H700 device	With Auto selected, scaling_max_freq reaches the greatest advertised frequency (~1.5 GHz) and schedutil scales under load; no new performance, thermal, or audio regressions observed

Applications and device-specific hardware

ID	Test	Status	Evidence/device	Result and notes
APP-01	Files/NextCommander input, scaling, and first frame	✅	RG40XXV, RG34XXSP	Works at PPU 2
APP-02	Recently Played and game switcher	✅	RG40XXV, RG34XXSP	Both flows work; rendered assets are clean
DEV-01	RG40XXV 640×480/single-stick baseline	✅	RG40XXV	UI, input, game runtime, HDMI, and Bluetooth paths pass
DEV-02	RG34XXSP 720×480/lid/dual-stick delta	✅	RG34XXSP	UI scaling, lid, controls, apps, and games pass
DEV-03	RG28XX 480×640 rotation delta	✅	RG28XX	UI, game Aspect/Fullscreen scaling, bootlogo apply, and rotated preview pass
DEV-04	RGcubexx 720×720 delta	✅	Community / user confirmation	General UI scaling passes. #18 under-fill fixed (MAIN_ROW_COUNT 10 on cube/HDMI); user confirmed the 10-row layout on cube hardware
DEV-05	Bootlogo 640×480 apply/backup	✅	RG40XXV	Carousel, apply, backup, and original entry pass; restore uses the same apply path
DEV-06	Bootlogo 720×480 preview/apply/backup/restore	✅	RG34XXSP	All 23 presets render; first backup and restore pass
DEV-07	Bootlogo 480×640 apply and preview orientation	✅	RG28XX	Applied logo and rotated previews are upright

Recovery and robustness

ID	Test	Status	Evidence/device	Result and notes
ROBUST-01	Stock launcher.sh stop SIGUSR1 does not terminate the NextUI boot shim	➖	Design / accepted	Shim uses trap '' USR1; not exercised end to end and not required as a Release blocker for h700-rc8
ROBUST-02	Dirty FAT card recovery via fsck.fat -a	➖	Design / accepted	Boot fsck.fat -a / repair path exists; not induced end to end and not required as a Release blocker for h700-rc8

Capability boundaries

ID	Test	Status	Evidence/device	Result and notes
CAP-01	FN1/FN2/HOME assignable pak actions	➖	RG XX hardware/platform API	RG XX devices have no dedicated buttons; Settings hides the assignment UI
CAP-02	Keep awake over USB data connection	➖	H700 platform API	PLAT_isUSBConnected() is a stub; charging-state power behavior is covered by POWER-04
LED-01	RGB LEDs driven over the UART5 MCU	✅	RG40XXV	Verified end to end: raw frames light the L stick ring, and NextUI drives it with the seeded 0x440044 default after update. Second bank drives nothing on the V, confirming dev_num_leds = 1. /dev/ttyS5 is free (console and getty are on ttyS0)
LED-02	Colour, brightness, all four effects	✅	RG40XXV	User-verified in LedControl
LED-03	Ambient Mode narrowed to Off/On	✅	RG40XXV	User-verified
LED-04	LedControl on a model without RGB	✅	RG34XXSP, RG28XX	Shows the "no RGB lights" screen, exits on B, and does not open /dev/ttyS5
LED-05	Speed control	✅	RG40XXV	Rainbow/Rainbow Multi take a continuous 0-255 speed byte and vary smoothly across the whole 0-4900 range. Breathe has only the three firmware rates (modes 2/3/4), so its Speed row quantises into thirds — a known cosmetic wart, not a defect
LED-06	tg5040 regression after PLAT_getNumLeds	✅	TrimUI Brick	Per-LED mapping still correct — each LED responded to its own mode change. This is the path where a wrong count would write brightness 0 to the global max_scale and kill every LED
LED-07	Ambient Mode UI when lights ≤2 / 0	✅	RG40XXV, RG34XXSP	Ambient Mode appears on RG40XXV (≤2 lights, Off/On). Hidden on RG34XXSP (0 lights / no RGB)
LED-01b	Two-bank models, bank order	⚠️	External users (RG40XX H / RG CubeXX class)	Only the right LED responds; the left cannot be changed. Likely bank-order / H700_LED_RIGHT_FIRST mapping. Single-bank RG40XXV path (LED-01) remains Pass. Accepted known bug for the merge PR — maintainer has no dual-bank device; upstream decides merge-with-bug vs wait-for-fix

Test execution protocol

• Candidate builds must record BUILD-01, BUILD-02, BUILD-03, and BUILD-04 against the actual artifact being promoted.
• Changes under workspace/all/ require both H700 and tg5040 builds.
• A new device needs a short shared-runtime smoke plus tests for its hardware/model delta; it does not need the full matrix repeated without a specific risk.
• Re-run stock integration probes after an Anbernic firmware update because boot, model detection, WiFi, Bluetooth, and sysfs paths are firmware-coupled.
• When shared code adds a button or PLAT_* API, verify all platform definitions and run a full build; single-pak builds can miss translation units.