- New MulticamPanel: rows/cols angle grid with the current angle
highlighted, click-to-switch, 1-9 switch-and-split and cmd-1-9
switch-only shortcuts (focused-panel routed), deferred switch queue
during playback.
- src/oakui/multicam.rs: clip->connected-sequence resolution, multicam
state detection (selection then playhead fallbacks), per-angle frame
requests rendered through the process backend into an LRU cache.
- Timeline clip context menu Multi-Cam checkable item wired to
oaktimeline::multicam enable/disable with undo.
- Engine trait extended (real + mock); mock drives the real command
path with synthesized angle frames.
- WorkerPool thread pool deleted; RenderManager defaults to the
Processes backend (oak-worker children), Threads kept as a test-only
inline dispatcher; audio tickets stay in-process until S3.
- Onscreen path reads worker shm slots directly: BGRA8 slot format,
RenderedFrame::Shm wrapped into the display buffer (single disclosed
GPU-staging memcpy), scopes analyze BGRA8; the long-lived full-res /
thumbnail paths take the counted slot_to_vec copy and release.
- Playback pre-render window: forward 120 frames (configurable) fed to
the PreviewScheduler at Playback priority, interleaved across
workers, cached in shm slots until the playhead consumes them;
generation-based invalidation cancels and releases on edits.
- oaktask export and oak-cli run on private ProcessDispatchers (fixed
a pump-while-locked self-deadlock in the export loop); facade
get_frame handles ShmFrame payloads.
- Acceptance: preview path main_heap_frame_copies == 0 with spawned
workers, CLI transcode/render verified end to end.
- oaktimeline::multicam: clip_find_multicam (buffer/tex_in depth-1
lookup), multicam_enable/disable (rewire sequence<->clip through a
MultiCamNode), multicam_switch (split-preserving-links at the
playhead, each half owns an independent multicam copy, linked clips
switched together) as single undo commands with C++ labels.
- BlockSplitCommand now duplicates the clip's whole dependency graph
(copy_node_and_dependency_graph_minus_items) instead of just the
block core, matching the C++ BlockSplitCommand::prepare semantics;
undo detaches the copied subgraph, redo re-attaches identity-
preserving.
- oaknode: fix serializer dropping edges from the first-created node
(ptr=0 was not registered in id_map), restoring sequence_in edge
round-trips; multicam node and clip wiring serializer round-trip
tests.
Proxy: preview-path proxy substitution (global UseProxyMedia AND
per-footage enabled AND on-disk ready; export always uses originals),
proxy generate/delete/reveal/enable actions, ProxyDialog with global
and per-footage custom params, Tools menu + context-menu Proxy
submenus, progress in the status bar, OVE serialization of proxy
metadata and source_start_time.
Sync: timeline context-menu Synchronize by Source Time / by Waveform /
by Waveform (Adjust Speed) with ctrl-shift-w, cache-envelope
extraction with validity masks, reference/anchor selection and
single multi-undo application (replace-with-gap, speed adjust,
re-place) mirroring timelinewidget.cpp semantics.
FootageBehavior now keeps the probed stream list (video/audio, per-stream
duration in rationals) instead of dropping it, and the probe CLI walks
that inventory to report real durations, frame rates and stream counts
rather than the previous zero placeholders.
The static FFmpeg's external codec libs pull in -lz, which on this
toolchain resolves to a copy whose install name is @rpath/libz.1.dylib
(zlib-ng-compat); without an LC_RPATH entry all three binaries died in
dyld at startup. The app/cli/worker build scripts now emit
-Wl,-rpath,/usr/lib.
Also: FFMPEG_DIR moves into the committed .cargo/config.toml as a
workspace-relative [env] entry — ffmpeg-sys-next's build script cannot
read .env files, and without it the crate silently linked the shared
Homebrew FFmpeg while oakffmpeg-link emitted the static transitive
flags (mixed linkage). docs/build.md updated.
Module-internal object references are Rust types now (values, Arc,
Mutex); CHandle remains only at the oakengine C-ABI boundary:
- oakundo: the global stack holds UndoStack/UndoCommand values
directly (stack token is the static's address)
- oaktimeline: marker/workarea boxes carry Arc<Mutex<T>>; commands
share the same allocation through Arc clones (readers in oakengine
stubs and the app's graphops updated to lock)
- oaktask/oakstorage: sessions, write-through bindings and the
database backend pass ProjectArc; the Session drops its manual
release bookkeeping; nodeutil keeps the CHandle<->Arc boundary
conversion (release_project restored for the app)
- oakcodec: handle.rs deleted outright (no facade entry needed it);
texture/block placeholders are unit structs
- oakrender: copier's project handle is an identity u64; alive-count
machinery removed; handle.rs is make_owned/get/get_mut only
- oakplugin: the instance registry is gone (its unregister key never
matched, leaking weak entries); handle.rs is the RefBox boundary type
- oaknode/oakcommon: only dead guard/borrow helpers removed; external
payload handles (texture/processor) documented as the boundary
Flake hunts landed along the way: the audio recording test serializes
on the shared manager lock with a normalized state; the autocacher
cancel test uses a slow producer so cancellation is deterministic.
- oakengine is now cdylib-only (no rlib/staticlib consumers anywhere;
cargo tree verified) — the plugin/external C ABI layer; README and
docs updated
- cd.yml drops the dylib embedding/re-sign steps (the app no longer
links it)
- test race root-caused and fixed for good: the global undo stack lock
is now a re-entrant mutex (parking_lot) shared by every test that
drives the stack, including the previously unlocked node/render
families; the render-manager serial-ordering bug (an earlier repro
test initialized the global manager before the not-initialized test)
is fixed with a shared SERIAL guard and a manager shutdown
- 5 consecutive parallel runs clean; serial 209/209
- real.rs rewritten over module Rust APIs (Arc<Mutex<Project>> +
NodeId; the addref handle dance and renderer boxes are gone);
AppEngine trait and all panels untouched
- new app assembly layers: graphops (project/timeline/edit
primitives), effectchain (chain composition with undo groups),
renderops (montage build + ticket render + ExportTask export),
library via oakstorage directly
- module-side safe API additions: oakundo global value-semantic
push/undo/redo + from_closures, oakstorage project_arc_of
- deleted: src/oakui/ffi.rs, src/oakui/host_syms.rs, the dylib link
config in build.rs (only the gpui IOSurface framework link remains)
- the binary carries zero liboakengine references (otool/nm verified);
101 app tests green incl. the real-render and full-res e2e tests
- behavior improvements for free: sequences land in the project graph
(the facade scratch-project deviation is gone), footage drops take
one undo record, effect remove/reorder undo restores edges
- oak-cli: new engine.rs assembly layer maps every facade call to
module Rust APIs (oaknode graph/serializer, oaktimeline commands,
oakrender ticket arena, oaktask ExportTask, oakcommon config); the
ffi/optional/host layers and build.rs link config are gone
- oak-worker: the worker session + POSIX shm transport moved into the
crate (oakrender backend + serde_json control plane); no dylib
- both binaries carry zero liboakengine references (otool verified);
tests green (30 cli / 41 worker)
- oakundo::global: the process-global undo stack, grouping and a
command observer API; facade undo.rs becomes a thin forwarder
- oakstorage::writethrough: the binding table, snapshot thread, flush
and config resolution; it subscribes to oakundo's observer itself
- the remaining non-forwarding facade logic (TaskMeta, effect chains,
timeline composites, RendererBox, exporter path) is documented as
facade-owned with reasons
- facade exports unchanged; full suite stays green (the one
render_manager_not_initialized failure is pre-existing on the base
commit)
- full-res source jobs carried only the footage node box: dropping the
project mid-flight left the node dangling (crash in the worker's free
path) and the node was also freed twice (at renderer creation AND at
release). The request now carries an addref'd project copy and the
node is freed exactly once; regression test drops the project before
the worker runs
- the source clock ticked at the SEQUENCE length, so playing footage
with an empty sequence froze the source playhead at 0; the source
clock now loops at the selected footage's probed duration
- oak-cli/oak-worker: DeepSeek's refactor batch (clap migration, engine
FFI consumers); the stale exporter-family test flipped to the real
contract (mp4 is written)
- engine: render_audio smoke test on an empty sequence
The 'host-provided' oakcore_audioparams_* runtime imports dated from
the deleted C++ host; the facade is their only caller. The dylib now
defines and exports the six symbols itself (repr(C) AudioParams mirror,
liboakcore-compatible semantics), -Wl,-undefined,dynamic_lookup is gone,
and the Windows DLL undefined-symbol blocker is removed by construction
(Windows CI/packaging stays off until a real toolchain verifies links).
- viewers show the 480px proxy immediately and a background thread
fills the sequence-resolution frame (per-monitor in-flight job,
generation-based staleness, playback skips full-res)
- preferences dialog complete: cache dir (now consumed by
default_disk_cache_path), proxy policy/divider, snapshot interval
(write-through era autosave), default transition length, audio
in/out devices (new facade device-enumeration exports; audio init
from config — playback was never creating the audio instance),
language/theme/renderer backend, all persisted via config
- shortcut map (src/shortcuts.rs): space/J/K/L, I/O, S split, A/^A,
⌘Z/⌘⇧Z, ⌘N/⌘O/⌘S/⌘E/⌘Q, frame step, Home, track zoom; dispatch
shares the menu action path and stays silent over modals
- screenshots: preferences dialog zh/en captured and reviewed
- facade: oakengine_sequence_set_workarea_undoable (enable+range as one
undo record); marker add/remove/list already existed and are now
covered by it_timeline (10 new tests: markers, workarea, cross-track)
- fixes: stubs workarea_get honored NULL out-params (is_enabled always
failed); export tasks honor custom ranges (export_params_pod dropped
them; oaktask EncodingParams carries has_custom_range)
- app: markers on the ruler (gpui diamond markers), menu
sequence-add/remove marker, set/clear work area (selected-clip bounds
or playhead+1), ruler drag previews live and commits one undoable
record (C++ ruler semantics); export uses the work area when enabled
- cross-track clip moves were already wired; now covered end to end
- ProjectDataSource<RealEngine> reads the real bin tree (roots/
children via the facade folder/footage enumeration)
- AppEngine::import_footage implemented (facade project_import_footage,
last_error surfaced); double-click opens in the source viewer
- dragging files onto the browser imports them (also fixes a real gpui
bug: ExternalPaths arrives as the bare type, not Arc-wrapped, so
drops never fired)
- tests: real-engine import->browser listing, widget drop routing,
facade folder enumeration failure matrix
D3: oakdb+pg:// fully wired (shared sea-orm entities, BIGSERIAL DDL,
connect-probe instead of pool retry on dead servers); Storage/Backend=pg
+ Storage/PgUrl config; 13 OAK_TEST_PG_URL-gated PG tests (verified
against a Docker postgres:16), always-on clean-error tests otherwise.
D4: DaVinci-style project manager — list with derived stats, create/
rename/duplicate/delete (confirm)/import/export (native dialogs,
ove/otio/fcpxml), shown at startup and from the file menu; facade
oakengine_library_* exports (list/create/delete/rename/duplicate/
import/export + project_load_library that binds write-through);
save/save-as menu becomes 'export project file', open splits into
from-library/from-file; status bar shows library write state; storage
activates on app start and flushes on exit; spawn_modal reentrancy
fixed (window-callback path) with a doc note.
Also: the P1 audio test's environment probe was lost in the ffi purge;
restored on cpal (the output device is cpal now).
- facade storage session manager: project handles bind to the default
SQLite library on project_new/load; every undo push/group_end/jump
write-throughs via DatabaseBackend::save (diff journal); project_free
flushes and unbinds
- background snapshot thread (Storage/SnapshotIntervalSec, latest-wins,
newest 3 kept) with exit flush (oakengine_storage_flush)
- config-gated: storage only activates with an explicit
Storage/Backend=sqlite, so headless consumers and tests never touch
the real library; new exports: storage_flush/is_bound/last_error
- it_storage: kill -9 recovery, cross-session undo, snapshot pruning,
multi-project isolation, graceful degradation
- also fixes a real config test polluting the user config.ini and the
undo-stack test races
Single-lib cleanup: the per-crate src/bridge/ and src/ffi.rs layers are
gone (oakundo/oakcommon/oaknode/oaktimeline/oakcodec/oakaudio/
oakrender/oaktask/oakplugin/oakstorage); cross-crate calls are plain
Rust, CHandle marshalling shrinks to the oakengine boundary, and tests
call the Rust APIs directly (pure C-ABI wrapper tests removed where
the domain layer already covers the behavior).
exporter.h family implemented: oakengine_export_render (CLI contract),
oakengine_export_render_with_params (was a stub), last_error and
progress callback; synchronous path reuses task_create_export +
start_sync. Fixes on the way: oaktask video ticket self-deadlock,
audio params dropped on the export path, codec encoder AAC slicing and
H.264 time base. Real-mp4 tests cover both entry points, progress and
the illegal-argument matrix.
Also: oakstorage session maps null project handles to None (version-
info path), configstore test double literal 3.14 -> 3.15 (clippy PI
lint), oakaudio output callback scratch buffer + env-aware P1 test,
cli media round-trip test uses a generated 16-frame clip (no more
minute-long debug runs).
oakstorage (new workspace member): URI dispatch, pluggable backends
(ove-xml built in, otio/fcpxml via oakotio, C-vtable foreign
registration), the M10 C API surface, version info codes, last-error
and alive accounting; round-trip tests per backend.
oaknode serializer: persists the full timeline — sequence track lists,
track block lists, block ranges/media_in/speed/flags, clip footage
references, footage filename+streams, folder children — through
<custom> behavior hooks with two-phase reference resolution; loads the
C++ <olive><project><layout> containers (golden: tests/
project_with_footage.ove); round-trip is field-by-field and
byte-idempotent.
facade (API only extended): oakengine_sequence_as_node /
_sequence_node_count / _sequence_node_at / _sequence_remove_node;
remove validates ownership by project UUID (arena slot/generation
collide across projects). it_node covers enumeration/edits and the
NULL/illegal matrix.
app: NodeGraphDataSource enumerates the current sequence's graph —
footage / effect / clip / output cards at their context positions
(deterministic role-grid fallback), real edges plus synthesized
clip->sequence tex_in edges; connect/disconnect/remove/drag-release
all go through undoable facade commands, drag previews stay local.
- output callback reuses a scratch buffer instead of allocating per
call (real-time rule)
- P1 consumption test probes real callback delivery and skips on
headless/background sessions (CoreAudio starts the stream but never
runs it outside the GUI session), with a 30s poll for slow HAL
startup; restores the manager singleton state afterwards
- waveform: drop leftover DBG-WF debug prints
engine:
- oakrender eval footage hook decodes via oakcodec (JobSpec::Footage
carries filename/stream); ticket/ffi/manager wiring, real-media
decode test with programmatically generated MPEG-2
- oakrender bridge/codec.rs + node.rs: direct oakcodec/oaknode calls;
the crate's dlsym module is gone (project_deep_copy/sync_copy remain
documented always-fail stubs — never implemented in oaknode)
- oakaudio waveform/decoder path adjustments for the decode hook
app (gpui + gpui_widgets):
- menu bar scrubbing: hovering another top-level title while a menu is
open switches to it; popup width is content-aware (CJK-aware) instead
of fixed 160px
- density pass: window rem 16 -> 14px, menu rows 26 -> 22px, dock tabs
32 -> 26px, viewer transport tightened
- open/import/save-as use the native platform file dialogs
(prompt_for_paths / prompt_for_new_path; multi-select import);
MockEngine records imported footage for tests
- project explorer Tree/Icons toggle is localized (explorer.tree /
explorer.icons widget keys)
source monitor:
- new facade exports oakengine_renderer_create_for_node (render any
node, not just sequences) and oakengine_project_footage_at (fetch a
footage node by index); it_render covers the e2e render plus the
NULL/illegal matrix
- RealEngine renders the selected footage's real frames to the source
viewer (per-node renderer slot, frame cache invalidated on selection
change, synthetic fallback kept)
- known gap (documented): pixels stay transparent black until
oakrender's footage decode hook lands — the render surface itself is
real end to end
audio meter:
- oakaudio manager can now report per-channel linear peaks of the
buffered output (PreviewAudioDevice::peek_tail + levelmeter analysis
of the newest 8192 frames, packed/planar F32)
- new facade export oakengine_audio_output_levels (negative codes pass
through); RealEngine's AudioMeterDataSource reads it instead of
returning hardcoded silence
- tests: module peak readback + facade validation/readback matrix
- oakengine_sequence_move_clip implemented for real (oaktimeline
TrackMoveBlockCommand; fixes the graph-ownership/gap-anchor/ripple
trim bugs the stub was hiding); same-track via the frozen C ABI,
cross-track supported by the module command
- oaknode clip blocks now declare a tex_in texture input and set
effect_input to it, so timeline clips can host effect chains; facade
test covers effect insert/remove on a real clip
- oakffmpeg-link: FFMPEG_DIR is now mandatory with a clear panic (a
Homebrew upgrade left the system ffmpeg .pc pointing at a deleted
dav1d Cellar path, breaking links); reads a git-ignored workspace
.env for IDEs that cannot inject env vars (RustRover); links the C++
stdlib for C++ codec libs (svt-av1)
- oakengine re-exports oaknode so tests share one crate instance;
it_node uses the direct instance's value type where it calls the
module FFI (the --workspace dev-dependency feature split builds
oaknode twice)
- tooling/ffmpeg/build-ffmpeg.sh builds release/8.0 static+PIC into
.cache/ffmpeg: GPL/version3, every free-license external codec lib
probed via pkg-config (enabled when present), per-OS hardware
acceleration (VideoToolbox/AudioToolbox, VAAPI/VDPAU/libdrm,
D3D11VA/DXVA2/MediaFoundation, nvenc when ffnvcodec exists)
- tooling/install-deps.sh installs those libraries on Homebrew / MSYS2
UCRT64 / Debian-Ubuntu / Fedora / Arch; nothing in the build sudo's
- ffmpeg-next's own build feature is unusable (every crate-version to
FFmpeg-release pairing is broken upstream: 9.0.0->FF9 AVCodec fields,
8.1.0->FF8.1 new enum variants, 8.0.0->FF8 FF_PROFILE rename), so
ffmpeg-next 9 + FFmpeg 8.x headers via FFMPEG_DIR it is
- new links-crate oakffmpeg-link emits the static FFmpeg's transitive
link flags from its .pc files (cargo only propagates them from links
crates, and rustc prunes the flags unless the rlib is referenced —
hence the force_link statics)
- docs/build.md updated for the Rust workspace flow
facade (oakengine/oaknode, API only extended):
- oaknode_node_get_effect_input / oaknode_node_get_flags module exports
- oakengine_clip_as_node, oakengine_node_effect_count/at/insert/remove/
move/set_enabled, oakengine_node_identity/is_enabled/get_type_id,
oakengine_node_factory_id_at — chain edits wrap disconnect/reconnect
into single undo groups; insert/move/remove validated against the
factory and the host's effect input
- it_node covers enumeration/insert/remove/move/toggle/undo-redo and
the NULL/illegal-input matrix
app:
- RealEngine binds the selected clip's effect chain to the inspector's
effect stack (cards, enable, reorder, remove, expansion state), all
edits undoable through the facade
- AppEngine gains set_selected_clips / addable_effects / add_effect
(default no-ops keep the mock untouched); the timeline selection
drives the stack target; the inspector gets a small add-effect menu
- known module gap (documented): ClipBlockBehavior sets no effect_input
yet, so timeline clips cannot host effects until oaknode grows tex_in
- processor.rs: resample/channel-convert/time-stretch now runs an
in-process FFmpeg filter graph (abuffer -> atempo -> aformat ->
abuffersink) via ffmpeg-next
- waveform.rs: extraction decodes through oakcodec's FFmpegDecoder
(interleaved f32) instead of the fb_decoder/fb_audio_graph pair
- bridge/ffmpeg.rs and the null fb_* test stubs deleted;
oakcommon::ffmpegutils bridge constants become plain ints
- liboakengine.dylib no longer imports any fb_* symbol (nm -u clean);
the remaining runtime imports are the host-provided oakcore_* symbols
- real decode/resample tests added (generated PCM input, no network)
Mike-Solar