format_, video_codec_, audio_codec_, subtitle_sidecar_fmt_ and
subtitles_codec_ were left indeterminate by the constructor; reading
them before the corresponding Enable* call was UB. Initialize to the
kFormatCount/kCodecCount invalid sentinels.
- Footage can store custom proxy parameters (width/height/crf/preset/
extension/audio) that override the global settings; they are
serialized with the project and used by every generation entry point
via Footage::GetEffectiveProxyParams()
- Proxies now include the source audio streams (AAC) unless disabled;
the proxy filename records the audio flag and offline audio rendering
decodes from the proxy when present
- ProxyTask resolves ffmpeg from the new FFmpegPath config key first,
then PATH, then common install locations (e.g. Homebrew on macOS),
instead of relying on PATH only; the error message points at the
preferences when no executable is found
- ProxyTask::BuildArguments() is extracted for testability
- Add ProxyIncludeAudio and FFmpegPath config defaults
- Add regression tests for the filename audio marker, config-backed
params, ffmpeg resolution, argument building, and custom-param
persistence
Ubuntu (FFmpeg 7.0) build failure: the public FB_PIX_FMT_* values were
hardcoded to AVPixelFormat enum values, but those shift between FFmpeg
releases (new formats are inserted mid-enum: X2RGB10 in 7.0,
RGBF16/GRAYF16 later), so the static_asserts in internal.h failed and
RGBF16LE/GRAYF16LE did not exist at all.
The FB pixel format values are now fixed identifiers translated inside
the library by pixel format name (stable across releases):
- fb::PixFmtToAV/PixFmtFromAV resolve the static table by name;
formats a decoder produces that have no static identifier (hardware
downloads like nv12/p010le/p210le) receive process-local dynamic ids
>= 1000 so they keep round-tripping through the API (scaler etc.)
instead of collapsing to NONE and crashing sws.
- All crossing points translated: scaler, frame get/set format,
decoder/probe stream info, encoder config and write_video_frame,
fb_pix_fmt_* utilities, fb_find_best_pix_fmt_of_list (which also
reinterpret_cast the caller's list; now translated element-wise and
properly NONE-terminated).
- nv12 and p010le get static identifiers as the common hardware
download formats.
- Hardware frame detection no longer compares pixel formats (breaks
across format spaces); it uses the existing fb_frame_is_hw().
macOS test failures (FFmpegBridgeEncoder.WritePngVideoAndProbeBack,
WritePcmAudioAndProbeBack): config.filename held a dangling pointer to
a temporary QByteArray (UB; happened to work on Linux). Keep the
QByteArray alive for the encoder's lifetime.
Windows (MSYS2 UCRT64) build failure: timecodefunctions.h used int64_t
without including <cstdint> (previously pulled in transitively).
Full gtest suite: 584 passed.
Replace every direct FFmpeg call in the editor and the render worker
with the pure C ffmpeg_bridge API, wrapped in thin C++ adapters that
preserve the original interfaces:
- avframeptr.h: olive::AVFrame adapter around FBFrame handles
- ffmpegutils: format conversion helpers on FB_* constants; the int
overload is renamed GetCompatibleBridgePixelFormat to avoid a silent
overload-resolution trap with the PixelFormat enum
- ffmpegdecoder/ffmpegencoder: rewritten as handle-based adapters over
FBDecoder/FBProbe/FBEncoder/FBScaler/FBResampler
- audioprocessor: FBAudioGraph push/pull adapter
- pluginrenderer/OliveClip: sws/pixdesc usage converted to FBScaler and
fb_pix_fmt_* queries
- AudioParams/channel layouts are plain uint64_t masks everywhere
Build integration: the root project no longer links FFMPEG directly;
only ffmpeg_bridge does. Binaries resolve the bridge library at runtime
via @loader_path inside the macOS app bundle (copied there post-build)
and via $ORIGIN/../ffmpeg_bridge/bin on Linux; on Windows the DLL is
installed next to the executables, so packages on all three platforms
ship the bridge library.
ffmpeg_bridge gains the extra API the adapters need:
fb_frame_make_writable, fb_decoder_get_format_duration,
fb_resampler_convert_frame, FB_PIX_FMT_YUV440P, SRT validation in
fb_probe_read_subtitle_stream, packed/planar fixes in
fb_encoder_write_audio, and a component-size fix in
fb_pix_fmt_component_size.
- Extend ProxyManager::ProxyParams with crf and preset, and update
ProxyTask/ProxyParamsEqual to use them.
- Add global config keys ProxyWidth, ProxyHeight, ProxyCRF, ProxyPreset.
- Add a Proxy Settings group to Preferences -> Disk with width/height/CRF/preset
controls.
- Read proxy settings from config when generating proxies from the timeline.
- Add Tools -> Proxy Settings... menu entry that opens Preferences on the Disk tab.
- Add ConfigDialogBase::SetCurrentTab() so PreferencesDialog can start on a
specific tab.
- Pass through input texture when the LUT processor is not ready yet,
preventing black frames while the processor is being generated.
- Serialize processor generation with a single in-flight task to avoid
concurrent OCIO lock contention that could freeze the UI.
- Invalidate cached frames after the async processor is set so the viewer
refreshes automatically without requiring the playhead to be moved.
- Add OAK_DISABLE_HWACCEL environment variable to force software decoding.
- Add FFmpegDecoderHW regression test for H.264 4:2:2 10-bit decoding.
This commit fixes crashes (SIGSEGV in CImg blur_bilateral) and black-frame
corruption artifacts during playback and scrubbing on macOS Apple Silicon.
Root cause analysis:
1. macOS uses Tile-Based Deferred Rendering (TBDR). glFlush() does not
guarantee tile memory writeback, causing glReadPixels to read incomplete
tiles (black/corrupted frames) and cache them to disk.
2. Olive uses multiple shared OpenGL contexts (RenderProcessor contexts vs.
thread-local PluginRenderer context). glFinish() only waits for the
current context, not the shared context that produced the texture. CPU
readback in PluginRenderer could read partially-rendered tiles.
3. OlivePluginInstance and OliveClipInstance are not thread-safe. Concurrent
RenderProcessors could corrupt internal QMap/images_ and params_ via
setInputTexture/renderAction races.
Fixes:
- OpenGLRenderer::Flush() on macOS now uses glFinish() unconditionally.
- OpenGLRenderer::DownloadFromTexture() and OpenGLRenderer::Blit() insert
glFinish() before readback/detach to ensure tile writeback completes.
- PluginRenderer::RenderPlugin() now acquires a per-instance mutex to
serialize concurrent OFX render calls.
- Before CPU readback in PluginRenderer, flush the renderer that originally
produced each input texture, ensuring cross-context synchronization.
- RenderProcessor::ProcessVideoFootage() flushes after BlitColorManaged.
- Add black-frame detection in ProcessVideoCacheJob() to auto-purge TBDR-
corrupted cache files.
- Add diagnostic qDebug() logging in viewer, decoder, renderer, and plugin
paths to aid future debugging.
Refactor FFmpeg frame processing and improve focus management in panels. Add memory sanitization for debug builds and update KDDockWidgets integration.
Simplify the frame reading process and metadata extraction in `ffmpegdecoder.cpp` by reducing redundant code and improving variable initialization. Add a helper function in `ffmpegdecoder.h` to map FFmpeg field orders to Olive interlacing types, enhancing code clarity. Adjust the condition for interleaved write in `ffmpegencoder.cpp` for better handling.
For some formats, particularly mxf, FFmpeg calculates the duration
of the stream incorrectly. Here we try to catch that and force Olive
to use it's much slower, but correct fallback method.
Mike-Solar
孙羽