Philips 50PUS8517/12 Debloat List

Notes & quirks on this firmware

• Android TV 11 (ro.build.version.release=11), firmware RTT2.211108.001, board PH1M_EA_9970A
• ro.product.model=2021_22_Philips_UHD_Android_TV ro.product.device=PH1M_EA_9970A
• Fingerprint: Philips/PH1M_EA_9970A/PH1M_EA_9970A:11/RTT2.211108.001/21029:user/release-keys
• TP Vision build: MediaTek MT5895, 2.67 GB RAM, armeabi-v7a userspace. Vendor packages use the
• org.droidtv prefix.
• Match on the BOARD and firmware, not on the model number. Philips/TP Vision ships one
• org.droidtv layer across a whole generation, so the x PUS85xx / PUS88xx sets of 2021-2022 on
• board PH1M_EA_9970A should carry the same package set; a different board or a Google TV
• (Android 12+) Philips will not. Check your own `pm list packages` against this file first —
• `tools/tv.py --probe` prints model, Android version and build.
• NOT the same platform as examples/philips-smart-4k-google-tv, which is a Skyworth-built
• India-market set on Google TV 14 with com.smartdevice/com.sky/com.tianci packages and no
• org.droidtv at all. Nothing from that file applies here and nothing here applies there.
• Everything is DISABLED, not uninstalled. Comment out any line you want to keep.
• Stock launcher: com.google.android.tvlauncher (pass it to restore.sh, as above).
• Live TV app: org.droidtv.playtv
• App labels below were resolved by pulling each APK and reading it with `aapt dump badging`,
• not guessed from the package name. That matters on this firmware: org.droidtv.nettvapp4546 is
• MEGOGO, nettvapp4655 is Deezer, nettvapp5512 is Fishing TV, and org.droidtv.candeebug is
• labelled "DAC" and is the analytics uploader, not a debug tool.
• QUIRKS ON THIS FIRMWARE, in the order they will bite you:
• 1. pm disable-user persists to /data/system/users/0/package-restrictions.xml with a delay
• (~10 s on AOSP). If you run disable.sh and power-cycle the TV straight away, the changes
• are silently lost: packages come back as enabled=0 (DEFAULT). Wait, or run
• `pm list packages -d` and confirm your packages are listed before pulling the plug.
• 2. `adb reboot` over TCP hangs holding the dropped connection. Kill the adb server right
• after sending it and wait on port 5555 coming back instead of on the adb command.
• 3. The Android 11 shell cannot disable individual components of another app
• ("SecurityException: Shell cannot change component state") and cannot `pm hide`
• ("Neither user 2000 nor current process has android.permission.MANAGE_USERS"). So if a
• package refuses to stay disabled, the shell has no second lever. `pm suspend` does work
• and survives a reboot, but it does not stop services: an app that starts itself from its
• own boot receiver keeps running with suspended=true.
• 4. com.google.android.apps.mediashell (Chromecast built-in) could NOT be kept disabled on this
• firmware and is deliberately absent from the list below. It returns to enabled=1 about 35 s
• after every boot, reproducibly. pm disable-user, pm suspend, per-component disable and
• pm hide were all tried; the first is undone, the second does not stop it, the last two are
• refused by the platform. Left enabled, at a cost of about 37 MB.
• 5. A DISABLED PACKAGE CAN LEAVE A PERSISTENT PROCESS RETRYING FOREVER, AND pidof WILL NOT
• SHOW IT. org.droidtv.xtv is the persistent TV engine and has five outbound service
• bindings. Two of them, eum and aurora, were on the first version of this list, and
• BOTH leave xtv in a permanent retry loop when disabled. Neither is in the list above now;
• both are in the DO NOT DISABLE block below.
• org.droidtv.eum -- bound via CONTEXTUALHELP. About five retries per second, logged as
• xtvContextInterface ... connectCotextService - Service failed to connec i is null
• 51 of them in 10 seconds of an idle TV. This one also BREAKS A VISIBLE FUNCTION, and not
• the one you would guess: the Philips sources drawer opens and vanishes within a second,
• so HDMI inputs stop being reachable from the remote's SOURCES key. SourcesService logged
• a correct, complete source list every time and the drawer still died.
• org.droidtv.aurora -- about five retries per second too, logged as
• xtvService: xtvService handleConnectAurora
• at a steady 201 ms interval, 59 in 30 seconds of an idle TV. This one breaks nothing
• visible and costs little: xtv sits at 0.2% of one core while doing it. It is log noise
• plus a retry that can never succeed. Re-enabling it is free in practice -- aurora's
• process does not start at idle even when the package is enabled, so xtv only needed the
• package to be resolvable.
• `pm enable` on either took its counter to zero immediately, with no reboot.
• THE METHOD, which is the transferable part: a live PID proves the process exists, not
• that it is healthy. xtv held the same PID through all eleven batches while both loops
• ran. And do not grep for a string you already know -- that is how I missed the second
• loop for a day after fixing the first. Clear the log, wait 30 seconds on an idle TV,
• then group EVERYTHING by frequency and look at what repeats:
• adb shell logcat -c; sleep 30; adb shell logcat -d > idle.txt
• On this set that takes an idle log from 130 lines per 30 s down to 50, and the 50 that
• remain are the firmware's own heartbeats.
• 6. The replacement launcher takes over on its own. Projectivy declares
• android.intent.category.HOME, so the moment com.google.android.tvlauncher is disabled the
• system resolves HOME to it — immediately, without a reboot and without enabling any
• accessibility service. Check with
• `cmd package resolve-activity --brief -a android.intent.action.MAIN -c android.intent.category.HOME`
• BEFORE disabling the stock launcher, and make sure your replacement is listed by
• `cmd package query-activities` for that intent. If it is not, you will be left without a
• home screen (see issue #90).
• TP Vision telemetry. org.droidtv.gel runs inside the candeebug process
• (android:process="org.droidtv.candeebug"), so disabling candeebug alone leaves a process with
• ───────────────────────────────────────────────────────────────────────────────────────────── DO NOT DISABLE on this platform. Checked by walking every intent resolver table and every service binding, not by reading package names: Ambilight and calibration org.droidtv.ambiair, ambihuecommon, alhuewizard, audiocalibration, calmancalibrator, bluetoothaudiosink Tuner, channels, inputs org.droidtv.xtv, xtvsystem, playtv, channels, epgdata, dvbi, teletext, ci, hdmiservice, euinstallertc, euinstallersat, hbbtvopapp, tv.tptvservice, icp, sunrise, com.mediatek.tvinput, tvinputservice.arbitratorservice, tv.agent, com.android.providers.tv Shell, power, remote, keys org.droidtv.tvsystemui, tvpowermanagerservice, GlobalKey, lkb (front-panel keypad), settings, menudataapp, homeintentresolver, com.google.android.tv.remote.service Firmware update org.droidtv.oad, com.easy.oad ("RC FW Update", remote firmware), devicesetup, networkwizard, nettvregistration Browser engine com.vewd.core.browserui plus org.droidtv.nettvbrowser — HbbTV and the app gallery run on it USB media org.droidtv.contentexplorer ("Media") Contextual help org.droidtv.eum -- see quirk 5. It looks like a pure e-manual and it is not: persistent xtv binds its EUMService and retries forever without it, which costs you the sources drawer Ambient screen org.droidtv.aurora -- see quirk 5. Disabling it leaves xtv retrying every 201 ms forever. Nothing visible breaks, but re-enabling costs nothing: its process never starts at idle org.droidtv.philipstvdrawer is safe to disable but was kept on purpose: it is the Philips side menu with shortcuts to the TV guide, channel list, sources, Media and quick settings — not ads.