Downloads bypass stream selection and reach play() as ready-made local
file streams, so audio mode never affected them. Apply the mode at that
choke point instead: swap a local video stream for the separately
downloaded audio track, or play a muxed file with the video track
disabled (vid set per-load in MPVClient, never toggled live). The
transform is bidirectional so stored audio-only queue/history entries
restore full video when the mode is off, and setAudioMode now reloads
downloaded content at the current position in both directions.
On macOS 27 beta, mpv's coreaudio AO fails to initialize (channel layout
set rejected with paramErr -50). The existing ao=coreaudio,avfoundation
fallback restored audio, but exposed an mpv bug: init() registers a HAL
hotplug listener (AudioObjectAddPropertyListener on the system object,
with the raw ao pointer as user data) before the step that fails, and
the coreaudio_error path never unregisters it. mpv then frees the ao
and falls back to avfoundation, leaving a dangling listener per
playback start. The next audio device change (AirPods connect,
sleep/wake device republish) invokes the listener with the freed
pointer and crashes with a use-after-free SIGSEGV on the
HALC_ProxyNotification Call Listener Queue (observed in build 262;
faulting address was reused "texture_" shader string memory).
Gate the AO order by OS version: on macOS 27+ use
avfoundation,coreaudio so the failing coreaudio init never runs; older
macOS keeps the native coreaudio first with avfoundation as fallback.
Update the read-only defaults mirror in MPV options settings to match.
macOS 27 beta's AUHAL rejects mpv's channel-layout property with paramErr
(-50), so the coreaudio AO fails to initialize and playback is silent.
Change the macOS ao option from "coreaudio" to "coreaudio,avfoundation"
so mpv self-heals: older macOS keeps using coreaudio unchanged, while the
beta drops to avfoundation (AVSampleBufferAudioRenderer) instead of playing
no sound.
In separate-window mode, hide() keeps the ordered-out window's SwiftUI
hierarchy alive (nilling contentViewController crashes AVKit), and its
updateNSView unconditionally re-attached the shared MPVOGLView -
stealing it into a window CoreAnimation never composites. The layer's
silent skip-render fallback then consumed every frame flag, so video
stayed black (climbing vo drops) until app restart.
- Decline stealing the shared view from a visible container into a
non-visible one; owner-initiated transfers bypass the guard
- Only transfer to containers in visible windows or the window still
tracked by ExpandedPlayerWindowManager; park the view otherwise and
reclaim it when a container gains a window
- Log skip-render frame consumption (rate-limited) and track render
health counters (draws/skips/dropped draws)
- Add a watchdog to the playback stats task that detects a stalled
video output and re-attaches the view to a visible container
- Don't mount player layouts during zero-size layout passes
Switching videos during active PiP left the window at the old video's
size with white space around the new content, for two reasons:
- MPV reports the new video's width and height as separate property
events while the old values are kept across switches, so a transient
mixed size (new width with old height) could reach the PiP capture
path and AVKit sized the window from that frame. Coalesce the PiP
capture-size/aspect update with a short debounce so only settled
dimensions are used.
- On macOS nothing told AVKit about the new aspect: unlike iOS, the
bridge never flushed the sample buffer renderer on an aspect change
during active PiP, and AVKit doesn't resize the macOS PiP window on
its own. Flush the renderer, reset the format description, and
resize the PiP window to the new aspect ratio explicitly.
Since 1d9c5e3b performRender feeds swap timing to mpv via
mpv_render_context_report_swap, but only after presentRenderbuffer -
which is skipped while PiP is active. Once swap reports stop mid-stream,
display-vdrop's vsync estimation goes stale and mpv drops nearly every
frame (vo drop counter climbing ~19/s on a 24fps stream), turning PiP
into a slideshow.
The frame is genuinely displayed during PiP via the
AVSampleBufferDisplayLayer, so report the swap on that path too.
With Atmos "Continuous Audio Output" enabled, some HDMI routes report
32 output channels, which mpv's audiounit AO cannot open - playback
stays silent while other (AVPlayer-based) apps work fine.
Update MPVKit to 0.41.0-n8.1.2, which builds mpv's avfoundation AO
(AVSampleBufferAudioRenderer, previously macOS-only) for iOS/tvOS
(mpvkit/MPVKit#73), and prefer it on tvOS via ao=avfoundation,audiounit
so mpv still falls back to audiounit if it fails to initialize. iOS and
macOS audio outputs are unchanged.
The MPVKit pin must be exactVersion: SwiftPM sorts the pre-release
identifier below 0.41.0, so a range starting at 0.41.0 never resolves
to it. Note this tag also bumps FFmpeg from 8.0 to n8.1.1.
1d9c5e3b removed the main-thread EAGLContext bind from setupAsync so
the context is never left current on two threads. But libmpv's
mpv_render_context_create requires a current GL context on the calling
thread (it probes glGetString(GL_VERSION)), so render context creation
failed with MPV_ERROR_UNSUPPORTED (-18) and every video played with no
--vo ("No render context set") on real iOS/tvOS devices.
Bind the context around the createRenderContext call and unbind right
after. The bind is scoped to setup, before the display link starts and
before any frame is rendered, so the never-left-current invariant that
protects the A10X fence fix (#947/#949) still holds.
Verified on Apple TV 4K (3rd gen): createRenderContext succeeds and
video frames render.
Claude-Session: https://claude.ai/code/session_01UEXP5f6F4S1zLdY8fdVuLJ
The macOS CAOpenGLLayer is pull-based: after the shared render view is
re-parented into a reopened player window, a single forced repaint is
issued. draw() consumed the forceDraw flag before validating the
viewport, so when that repaint landed while the layer was still 0x0
(the expanded window builds its content a runloop late), the request
was swallowed. With mpv paused during load, nothing repainted again,
leaving the video permanently black with working audio.
- Validate viewport/framebuffer in draw() before consuming flags;
on failure keep the forced draw armed and retry (bounded, 5x50ms)
- Repaint from setFrameSize on the first non-zero layout while a
forced draw is still pending
- Drop the stale fbo=1 fallback; render only into the validated
live framebuffer binding
- clearToBlack: take the CGL lock and skip glClear when no
framebuffer is bound - clearing FBO 0 on this core-profile context
latched GL_INVALID_FRAMEBUFFER_OPERATION, which libmpv reported on
every load
- Add diagnostics for dropped draws and the first-frame render
Apple TV 4K 1st gen on tvOS 26 crashes to the home screen with a
"blocked by IOFence" gpuEvent at playback start (#949) and stutters
with a starved demuxer cache (#947). Defensive fixes in the shared
iOS/tvOS OpenGL path:
- Drain in-flight GPU work (glFinish) before destroying the
renderbuffer, so a pending presentRenderbuffer fence can't hold the
CAEAGLLayer IOSurface being torn down or reallocated
- Never leave the EAGLContext current on the main thread: framebuffer
create/destroy and the background glFinish now unbind on exit, and
setup no longer binds the context on main at all
- Disable retained backing on tvOS; it exists only for iOS PiP frame
capture and adds per-frame IOSurface fence pressure
- Report swaps to mpv after each present so display-vdrop (the tvOS
default video-sync) gets swap-timing feedback, matching macOS
Add verbose-gated diagnostics for remote triage: playback stats every
10s (cache fill/rate, dropped frames, hwdec, avsync) and a rate-limited
slow-frame warning in performRender.
Addresses #949 and #947
The player video surface is a single shared MPVOGLView that is re-parented
between the mini-bar container (main window) and the expanded-player sheet
(a separate sheet window). When presenting the sheet, the new container
attaches the shared view before the old mini-bar container tears down, so
the old container's viewWillMove(toSuperview: nil) was ripping the shared
view out of its new home via removeFromSuperview() — leaving it with no
window and a black surface. Symmetric, so it also blanked the mini-bar
preview after collapsing.
Only detach the shared view if it still belongs to the container
(superview === self). Also force a repaint on window reattach so a paused
frame redraws, and extend the onAppear resumeRendering to macOS.
MPVOpenGLLayer called fatalError() when CGLChoosePixelFormat failed for
both the 10-bit and 8-bit (kCGLPFAAccelerated) formats, killing the app at
launch on machines without a usable GPU such as virtual machines.
Add a non-accelerated software pixel format as a last resort and make the
whole OpenGL init path failable: createPixelFormat/createContext return nil
(with an error log) instead of aborting, MPVOpenGLLayer.init becomes init?,
and MPVOGLView holds an optional layer, throwing openGLSetupFailed from
setup(with:) when OpenGL is unavailable. That error is already handled by
MPVBackend.setupMPVAsync, so the app now launches cleanly and only logs that
video playback is unavailable.
Surface audio-delay (±10/±100 ms) and video-sync-mode controls behind an
"A/V Sync" navigation row in Advanced settings rather than inline, keeping the
Advanced page uncluttered. Audio delay applies live to the running MPV
instance; sync mode takes effect on next playback.
Lets the Apple TV switch its HDMI output to match the playing video's
frame rate and dynamic range via AVDisplayManager.preferredDisplayCriteria,
driven from MPV's container-fps and video-params/gamma. Two opt-in toggles
(default off) live under Playback → Display on tvOS; both are no-ops on
other platforms. Anchor an AVKit class symbol so the linker keeps AVKit
linked — Swift only autolinks AVFoundation here, and without AVKit the
UIWindow.avDisplayManager category isn't loaded at runtime.
Subscribe to mpv log messages and capture END_FILE error code/string so
load failures bubble up specific causes (HTTP 404/403, DNS failure,
demuxer errors) instead of a generic 10s timeout.
