H.264 High 10 on Android TV boxes without a 10-bit hardware decoder
(Amlogic S905X4 class: onn 4K Pro, Homatics Box R) started black for over
ten seconds and then played choppy: mpv tried MediaCodec first, hit
"Could not initialize video chain", and only then fell back to software
decode on the GL vo, where the Mali-G31 driver has no
GL_EXT_texture_norm16 and pays an integer-texture conversion pass plus
lanczos scaling it cannot afford at 1080p.
Decide the decoder before mpv creates it. An on_preloaded hook now carries
both per-file policies (Dolby Vision P5 reshaping and this one): when the
track's codec-profile is High 10 (published at demux by the fork, see the
mpv-build pin) and MediaCodecList advertises no hardware AVCProfileHigh10,
hwdec is held at `no` and the session goes to the GL vo directly. The hold
parks any hwdec write from Dart and restores it for the next file, as the
DV P5 hold already did.
On a GL vo whose driver lacks norm16 (probed once through a pbuffer EGL
context), scale/cscale/dscale drop to bilinear and dither to off for the
session, restored when the vo goes back to the plane. Only options still at
their mpv defaults are touched, so a user's mpv.conf wins. Paired with the
fork's rg8-backed 16-bit plane emulation, the box goes from 13 drops/s and
a 13.7 s first frame to 0.7 drops/s and 2.5 s. Devices with norm16 (Shield,
Pixel 7) take neither the tier nor the emulation.
The native pin moves to edde746/mpv-build@0601b034da, which also picks up
the earlier android/linux/windows patch-series squash.
close#2065
Linux and Windows each wrote the same eleven pre-initialize mpv options
(keep-open, idle, input disabling, OSC, ytdl, audio fallback, and the
audio-only set); ApplyCommonStartupOptions in mpv_player_common.h owns
them once, with the ytdl security rationale kept in one place. Rendering,
windowing, HDR, and log-level options stay per platform.
The Android JNI event thread forwarded every mpv event to Kotlin, where
MpvEvent modelled ten variants but MpvPlayerCore consumes four. The
native switch now forwards only START_FILE, FILE_LOADED, and
PLAYBACK_RESTART (END_FILE keeps its own path), and the unused variants
are gone.
Emoji and other supplementary-plane characters in titles, file names, and
subtitle paths were corrupted on the Android MPV path, and malformed bytes
from mpv logs could abort under CheckJNI. NewStringUTF/GetStringUTFChars
speak JNI's modified UTF-8, not the standard UTF-8 mpv produces and
consumes; the native lead-byte filter and the Kotlin surrogate scrubber
ran on the wrong side of that conversion and could not recover it.
Both directions now go through UTF-16 (NewString/GetStringChars) with a
small JNI-free transcoder that replaces malformed input with U+FFFD, the
same policy as shared/cpp/sanitize_utf8.h on desktop. The header is
covered by the Android host native test harness.
Plezy consumed mpv through four unrelated supply chains: an MPVKit fork
via SwiftPM for the Apple platforms, a libmpv-android fork's AAR for
Android, an in-CI from-source build for Linux, and an unpinned
sourceforge mpv-dev 7z for Windows. All four now consume the same
per-commit, content-addressed binaries from
https://github.com/edde746/mpv-build, pinned to one commit and built
from one set of pinned sources (mpv v0.41.0 on Apple/Android/Windows,
ffmpeg n8.0.1, our libass fork).
- Apple: the SwiftPM package moves from edde746/MPVKit to
edde746/mpv-build across the ios/macos/tvos projects;
scripts/set_mpvkit_revision.sh becomes set_native_revision.sh, writes
every pin site plus the new repo-root mpv-build.lock.json, and
tvos/scripts/wire_mpv.rb derives the package repo from the locks.
- Android: the mpv Kotlin API and JNI glue move in-app under
android/libmpv (repackaged com.edde746.plezy.libmpv, exports renamed,
shrinker rules covered), and the module downloads per-ABI native
tarballs (lib/*.so incl. libc++_shared.so + include/) driven entirely
by mpv-build.lock.json, with a PLEZY_LOCAL_MPV_DIR escape hatch for
locally built artifacts. The fork AAR and its Maven coordinates are
gone.
- Linux: CI downloads the prebuilt self-relocating libmpv prefix
(lock-driven, sha256-verified) instead of compiling mpv/ffmpeg/dav1d/
libplacebo/shaderc from source; linux/packaging/build-libmpv.sh and
its test are deleted and native-inputs.json shrinks to the simdutf
entry the CMake builds still fetch.
- Windows: both arches FetchContent the mpv-build dev zips with
URL_HASH enforcement, replacing the checksum-less sourceforge
download and its ARM64 7-Zip special case; guard scripts updated.
The lock plus the Apple pin sites all point at mpv-build commit
d7c3d559, whose manifest was verified asset-by-asset against the
published release digests (17/17 match). Verified locally: wire and
pin-script suites green, runtime-input checks green, all four Android
ABI tarballs downloaded/verified/extracted through the real Gradle
tasks, the Windows FetchContent block exercised end to end through
cmake, the Linux asset hash and layout checked against the workflow
contract, and xcodebuild resolved the flipped SwiftPM graph with
SwiftPM validating every binary checksum.
edde746