Two different gpui keystroke tests flaked on Windows CI with the same
signature: a synthetic keystroke occasionally never reaches the action
(secondary-z lost while secondary-shift-z delivered; then a plain 's'
lost). Both passed every other run — a gpui test-harness delivery
flake, not an oak regression. A single retry pass absorbs it; a real
regression fails both passes.
oakstorage: the sqlite URI parse tests used /tmp/lib.db, which is not
absolute on Windows, so parse_target's is_absolute check rejected it.
Pick the absolute path per platform (C:/tmp/lib.db on Windows).
ci (Linux): wrap the test step in a 1500 s watchdog — a deadlocked
test prints nothing and never fails; on timeout the watchdog dumps
every test/worker process's thread stacks with gdb and then kills the
suite. (One such hang already ate a run; the previous green run needed
~4 min.)
ci+cd: Swatinem/rust-cache gains cache-on-failure everywhere, so a
red run still saves its compile cache (the actions/cache FFmpeg cache
already saves in its post phase regardless of outcome).
Windows: tooling/package/bundle-dylibs-windows.sh collects the MSYS2
runtime DLLs (libstdc++, libgcc, OpenColorIO, ...) with ntldd -R,
iterated to a fixpoint over freshly copied DLLs; a packager resources
glob places them next to the executables in the NSIS installer.
macOS: tooling/package/bundle-dylibs-macos.sh copies every non-system
dylib otool reports into Contents/Frameworks, rewrites the install
names to @executable_path/../Frameworks to a fixpoint, and ad-hoc
re-signs every modified Mach-O (rewriting invalidates the seal).
The CD package version no longer comes from the git tag: the root
Cargo.toml gains [workspace.package] version = "0.5.0", the oak
package inherits it (version.workspace = true — which cargo-packager
also picks up), and the Linux container packaging parses that field.
oakcodec: gate find_ffmpeg_searches_path to unix (chmod 0755 + shebang
fixture) and make find_ffmpeg_missing_returns_empty assert absoluteness
instead of a '/' prefix so the tests compile and pass on Windows.
ci (Windows): export RUSTFLAGS=-C link-args=-lmsvcrt in the build and
test steps. mingw-w64 (Nov 2025) forwards _assert to __msvcrt_assert
inside libmingwex.a, and rustc's link order leaves -lmingwex last, so
binaries that pull _assert.o (oakcommon's real_ocio test) fail to link;
a trailing -lmsvcrt re-scans the CRT import lib afterwards.
ci (Linux): copier_test dies inside ld.so before printing anything.
Replace the LD_DEBUG probe with stronger forensics: exported dynsyms
(interposition suspects), strace tail, valgrind tail, and siginfo
(si_code/si_addr) from the gdb run.
oakaudio: drop cpal's `asio` feature. asio-sys needs the proprietary
Steinberg ASIO SDK at link time (undefined ASIOGetSamplePosition etc.
on the GNU toolchain); WASAPI remains the Windows backend.
ci: the failure-only gdb step passed test args without --args, so gdb
treated --nocapture as a core file. Also collect loader-stage evidence
for the copier_test dl_main SIGSEGV: IRELATIVE reloc count, LD_DEBUG
tail, full backtrace and registers.
- localtime_s/gmtime_s are MinGW header inlines, not symbols — link
_localtime64_s/_gmtime64_s
- copier_test also segfaults only on the Linux runner; add it to the
on-failure gdb backtrace
- oakffmpeg-link forwards pkg-config --static --libs verbatim; FFmpeg's
.pc files can list -ldl via external deps, and MinGW has no libdl
- suites_test segfaults on the Linux runner too; run both plugin test
binaries under gdb on test failure
The crate's build.rs unconditionally adds the MSVC + Windows SDK
include dirs on Windows (for MSVC hosts); on the GNU toolchain that
breaks the bridge compile with MSVC-only headers. The runner's job
hook re-exports INCLUDE/LIB per step, so the in-step unset did not
help — patch the extracted build.rs instead (both the env-var failure
modes are now documented in the step comment).
- the warp runner's job hook re-exports MSVC INCLUDE/LIB per step, so
the GITHUB_ENV clear did not stick — unset in the Build/Test steps
themselves
- node_e2e_test segfaults only on the Linux runner; rerun the binary
under gdb on failure to capture the native stack
Each distro package builds inside that distro's container so declared
dependencies always resolve to native names: hand-rolled deb via
dpkg-shlibdeps + dpkg-deb, rpm via rpmbuild's auto-requires, Arch via
makepkg (non-root builder user). git/curl install before checkout
(container jobs). AppImage keeps cargo-packager on the Ubuntu runner.
The release gates on all four package jobs plus macOS/Windows.
- the Windows runner image exports MSVC's INCLUDE/LIB; cc-rs was
appending the MSVC SDK headers to MinGW compiles (vcruntime.h not
found)
- oak-worker handshake test helper advertised the input pool's total
byte size as per-slot data bytes (macOS tolerated the oversized
attach; Linux correctly rejects it)
- hw/sw decode comparison tolerance 0.05 -> 0.08 (VideoToolbox's
YUV->RGB legitimately differs by ~1 LSB of intermediate depth)
- the vendored OCIO source needs MSVC-only constructs (wide-path
ifstream); MSYS2's mingw build of the exact 2.5.2 the bridge targets
is the sane Windows path — DLLs get packaged next to the binaries
- oak-worker handshake test prints the error response on failure
(CI-only attach failure needs the message)
- rust-toolchain sets CARGO_HOME to the Windows userprofile path while
the msys2 shell's HOME is elsewhere — the yaml-cpp patch targeted an
empty directory and the assertion ls failed
- examples/screenshot.rs uses the macOS-only VisualTestAppContext; its
items are now cfg-gated with a non-macOS stub main so workspace test
builds pass on Linux/Windows
On a fresh runner registry/src has no hash subdir yet, so the unpack
glob never expanded and the step exited 2; derive it from the cache
dir and assert the patched file exists at the end.
- cargo fetch does not extract sources; the yaml-cpp <cstdint> patch
now untars the .crate into the registry src dir first (the glob
found nothing and the step failed with exit 2)
- oakaudio: the watchdog-wrapped audio test called Self::... from a
free-function test module (compile error in lib test)
- Linux: libxkbcommon-x11-dev for the gpui X11 client link
- Windows: patch <cstdint> into the vendored yaml-cpp (a cached cmake
configure ignores CXXFLAGS; the patch is idempotent and runs after
cargo fetch)
- macOS: the hw-decode test skips its VideoToolbox engagement
assertions on hosts where VT cannot initialize (headless/virtualized
runners) instead of failing
- display color management: the display ICC (system or custom) is
applied to viewer frames at present time (F32 in place, or in place
on the BGRA staging copy with the R/B swizzle baked into the OCIO
chain); preferences get a Color section (mode + custom ICC file); on
macOS the Metal layer is tagged with the display colorspace when
self-managing so ColorSync passes pixels through (no double
correction); frame caches track the transform generation so a mode
or profile change drops stale pixels
dpkg-shlibdeps over the three shipped binaries resolves every NEEDED
library to exact build-distro package names (FFmpeg/OCIO are static so
only base-OS packages appear) and rewrites the deb's Depends. Distros
with divergent package names (openKylin) get their own build instead
of a wrong-name dependency list.
Static FFmpeg + static OCIO leave only base-OS libraries; the audit
step prints objdump NEEDED for each packaged binary so any accidental
dynamic dependency (and any distro-specific package-name surface) is
visible in the build log.
- OCIO_RS_LINK=static everywhere: the vendored OCIO is linked into the
binaries statically — the package carries no OCIO dependency
- Linux: libasound2-dev (alsa-sys), libpulse-dev, libsndfile1-dev —
the full audio dev set
- Windows: -include cstdint for the vendored yaml-cpp (pre-GCC-13
transitive includes)
- Windows: install MSYS2's own Rust (x86_64-pc-windows-gnu host); the
rustup MSVC toolchain is not on the msys2 shell's PATH and the MSVC
linker rejects the Unix-style link args the build scripts emit
- Linux: libpipewire-0.3-dev + libspa-0.2-dev for libspa-sys (gpui's
Linux screen-capture/audio stack)
The distro OCIO is too old for the bridge's API floor where it matters
(Ubuntu 24.04 ships 2.1; the bridge uses 2.4+ APIs), and version drift
across platforms is a support hazard — enable ocio-rs' bundled feature
and drop the OCIO_INSTALL_DIR/system-package wiring from CI and CD so
Linux, macOS and Windows all build the same vendored OCIO. cmake/make/
diffutils added where the runners lack them (Windows FFmpeg build needs
make + cmp).
- serializer resolves node types through the factory's dynamic
(runtime-registered OpenFX) entries, so a project carrying plugin
nodes loads again (was: "unknown node type"); covered by a new
CI-gated round-trip test driving the real fixture plugin
- built-in effect nodes expose their inputs as inspector parameters
like the C++ parameter editor: localized input names from the
behavior, combo option tables via the new
NodeBehavior::input_combo_strings (16 nodes, string-for-string from
the C++ set_combo_box_strings), connection/data inputs excluded
- effect library: live drag-and-drop — onto the inspector's effect
stack (lands at the indicator position) and onto the node editor
canvas (creates the node at the drop point); double-click still
appends to the selected clip
- inspector parameter controls are no longer recreated per render
(gpui stack view caches them per effect), so sliders drag and
checkboxes click; the view observes the engine and silently re-syncs
values (undo/redo land on the widgets)
- timeline: left-press selects clips (plain/keep-multi/Ctrl-Cmd
toggle); clip moves clamp the shared delta so no clip of a linked
group lands before frame 0 instead of failing with "invalid move
target"
- oakplugin: createInstance-rejected instances skip the destroyInstance
notification (the plugin never owned them); vendor-suite fetchSuite
misses moved behind OAK_OFX_TRACE; the worker logs the discovered/
registered plugin counts
- CI: the OFX probe step also runs the serialization round-trip test
- gpui submodule: params view caching, clip click-select, library
drag payload, graph_position_at
- new minimal C OFX plugin fixture (ci_test_plugin.c) compiled into a
real .ofx.bundle by build_fixture.sh; the CI step points
OFX_PLUGIN_PATH at it and asserts the scan_probe example discovers
AND registers it (Linux/macOS)
- host bundle binary search now also covers the OFX-standard Win64
platform directory
- cd.yml: restore the Windows NSIS packaging job (obsolete oakengine
cdylib prebuild dropped) and repair the job indentation that had
silently detached the macos/release jobs from the jobs: map;
releases now gate on all three platforms
- save() writes <config>/config.toml (atomic temp+rename), flat keys
at the top level and group/sub keys as [group] tables, values as
native TOML int/float/bool/string (non-finite doubles degrade to
strings and restore via the declared type)
- load() prefers config.toml; a legacy config.ini (C++ or pre-TOML
Rust builds) is read once and immediately re-persisted as TOML; the
INI file is left in place; a corrupt TOML is reported, never
silently discarded
- cd.yml: drop a stale oakengine comment (the crate is retired)
- 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
CI matrix: warp-ubuntu-latest-x64-8x / warp-macos-15-arm64-6x /
warp-windows-latest-x64-16x (the macOS steps were still in place, only
the matrix entry was missing). CD: macOS DMG job re-enabled on the warp
Apple Silicon runner; linux and release jobs back on warp-ubuntu;
Windows NSIS stays disabled until the engine links there.
- every job builds -p oakengine first (the dylib is a build-dep-only
artifact otherwise and cli/worker link-search the profile dir)
- 512px icon (tauri-icns only maps 512@1x/1024@2x); dylib embedding
derives the path from the binary's otool reference; fpm invoked from
the gem bin dir; GITHUB_ENV blocks batched (SC2129); actionlint.yaml
whitelists the warp runner labels
- Windows job disabled with a reference block until the engine links
there
- real run: Oak-macOS-arm64.dmg produced, app launches from the volume
(known gap recorded: the dmg still dynamically links Homebrew codec
libs; not self-contained yet)
- CI: push/PR to main only; ubuntu/macos-15-arm64/windows-ucrt64 matrix;
deps via tooling/install-deps.sh; project FFmpeg built once and cached
under .cache/ffmpeg keyed on the build script; rust-cache for target/
- CD: tag v* packages with cargo-packager — deb/AppImage/pacman on
Linux (rpm converted from the deb with fpm), NSIS on Windows, and a
DMG (Apple Silicon) whose .app embeds liboakengine.dylib via
install_name_tool; tag pushes publish a GitHub release with all
packages attached
- root Cargo.toml gains [package.metadata.packager]; the app icon is
generated from Oak_Icon.svg with rsvg-convert in CI
- root build.rs now also links the app on Linux (link-search + rpath +
--export-dynamic); Windows remains blocked on the DLL undefined-symbol
problem (oakcore_* host imports), documented in build.rs
Compiler hygiene (all platforms):
- Silence warnings across the tree: missing override, -Wreorder ctor
init, -Wshadow, -Wsign-compare, missing switch cases, unused
functions/captures, Qt 6.11 deprecations (QMouseEvent/QDropEvent
accessors, qAsConst, Q_FOREACH over non-shared containers,
AA_UseHighDpiPixmaps) and the .bak/ backup tree removal.
- Fix regressions from the cleanup: missing clip decls in
capi/timeline.cpp, plugin.cpp rename fallout, panel setFocus
ambiguity, QGraphicsItem::pos vs event->position(), duplicate
k_push_button case, boolean test variable.
Windows:
- Qt portability: CommandLineParser is_set/add_option, QTimeZone
systemTimeZone (QTimeZone::LocalTime is 6.11-only), k_progress_* enum.
- Linking: stop adding oakengine to OLIVE_LIBRARIES (import lib plus
oakengine-obj caused multiple definitions); add OAKENGINE_STATIC so
internal consumers no longer reference __imp_* stubs.
- oakengine.ver: export olive::Renderer typeinfo so liboakgl.so can be
dlopened (Linux), DynamicRenderer no longer dlcloses backend libraries
(crash in RenderManager's dtor calling into unmapped memory).
- OTIO runtime: copy DLLs next to every binary on Windows instead of
relying on PATH (0xc0000135 in gtest discovery).
- Headless GL: the runner only has GDI OpenGL 1.1, killing every render
worker. Deploy Mesa llvmpipe as opengl32sw.dll (Qt's software-GL
channel) with QT_OPENGL=software, and let QT_OPENGL override the
AA_UseDesktopOpenGL default. ExportTask fails fast after 8 consecutive
undelivered frames instead of segfaulting or grinding forever;
FFmpegEncoder::write_frame tolerates null frames.
- Tests: GetTempPathA+PID temp dirs, GetLongPathNameA for 8.3 names,
forward-slash normalization when comparing project filenames.
Linux:
- Install libshaderc-dev so oakvulkan compiles GLSL (Vulkan tests).
- Accept UNORM floor-or-round (63/64) in the blit ping-pong test.
- Skip MainWindow construction test on the offscreen QPA (cannot paint
QOpenGLWidget).
Also: oak_cli_transcode gets a 300s ctest timeout, worker logs GL
context version and LoadGraph/render_frame stages, and
docs/plans/eliminate-event-bridge-issues.md (English translation).
Install libvulkan-dev + mesa-vulkan-drivers (lavapipe = software Vulkan,
no GPU needed) and stop passing CMAKE_DISABLE_FIND_PACKAGE_Vulkan so the
Vulkan backend and its tests are built and exercised.
- make_oakengine_gtest embeds oakengine-obj; test code must see
OAKENGINE_API as dllexport too (OAKENGINE_BUILD), otherwise MinGW
resolves nothing against __imp_* stubs
- the render worker intentionally drops QT_QPA_PLATFORM=offscreen and
needs a real display for its GL context; run the Linux ctest and the
filtered gtest steps under xvfb-run
- liboakengine.dll does not export engine-internal C++ symbols
(olive::Renderer, FileFunctions, VideoParams, Texture), so oakgl.dll
failed to link on Windows. Embed oakengine-obj instead (same pattern
as the engine tests); MinGW's name-based RTTI keeps cross-DLL casts
working.
- ci.yml: move to warp runners
- Linux: Qt6 metatype static_assert requires complete types; declare
OakEngineBlock with Q_DECLARE_OPAQUE_POINTER in the public header
(outside extern "C")
- Windows: gtest discovery runs the freshly linked test exe at build
time and it cannot find the DLLs (0xc0000135); copy oakcore/oakengine/
ffmpeg_bridge DLLs next to the test binaries on WIN32
- macOS: Imath header mixing between otio-install's stale bundled Imath
and Homebrew's Imath broke ImathBox.h (V2h undeclared); bump the OTIO
cache to v2 and drop bundled Imath headers so the system one is the
only set in play
- OAK_ENABLE_DYNAMIC_RENDER_BACKEND now defaults ON; it was previously
undefined (OFF) unless passed, so CI configure failed on the
unconditional oakgl target references in app/worker/tests
- add git to the MSYS2 package list in ci.yml and cd.yml; the OTIO
build step runs git clone in the msys2 shell, where git is not
installed by default
- app/ no longer includes engine C++ headers nor holds engine C++ types:
engine access goes through the oakengine C ABI plus C++ wrappers
(oakutil/oaknode.h, oakutil/oakvideo.h) and app-local mirror types
(tooltypes, trackreferencehandle, timelinecommonapp, keyframetypes,
subtitleapp, serializedlayoutinfoapp, nodevaluehandle, sliderdisplaytypeapp)
- engine: new C ABI functions for block/track/clip/transition navigation
and predicates, links, caches, waveform/playback, disk folder,
sequence_track_list, node_free, footage_is_valid, block_get_track,
get_brush; loadotio/saveotio ported to the current engine API
- OTIO is now a required dependency: CI and CD build it on every
platform, FindOpenTimelineIO fixed for OTIO 0.16/0.19 (the old deps
include requirement silently disabled OTIO everywhere), runtime
libraries are bundled into packages and copied next to macOS binaries
(oak_copy_otio_runtime)
- fix ProjectViewModel drag&drop mime read/write size mismatch (segfault)
- unify color label naming (k_olive -> "Oak") in the app-side mirror
- docs: OTIO required, FFmpeg minimum corrected to 6.0 (en/zh)
- gtest suite: 1925 passed, 0 failed
Physical split: app/{audio,cli,codec,common,config,node,pluginSupport,
render,task,timeline,undo,tool,shaders} plus coreengine, version and
ui/icons+colorcoding move to a new top-level engine/ tree, built as
liboakengine.so (shared). The render backends (oakgl/oakvulkan) move
with it and link the engine library instead of embedding a static
render-core subset (libolive-rendercore is gone).
- oak-render-worker now links liboakengine instead of the whole
libolive-editor object set: 336MB -> 2.9MB, no Qt Widgets UI
- the editor links liboakengine for the engine and keeps only UI
objects in libolive-editor
- install/packaging: GNUInstallDirs libdir on Linux, bundle copy on
macOS, oakengine.dll staged for NSIS, AppImage validation entry
- fix backend lookup for the new layout: DynamicRenderer searched
../app but backends now live in engine/; a stale pre-split liboakgl
in the build tree got dlopened instead, re-initialized and later
destroyed the interposed engine statics (full-suite segfault at
DialogSequenceParameterTab, found via gdb watchpoint)
liboakcore is now a shared library that exposes only a C ABI:
- every value class (Rational, TimeRange, Color, Bezier, AudioParams,
SampleBuffer) and the free-function groups (StringUtils, fraction
utils, Timecode) is wrapped in an opaque-handle C API under
core/include/olive/core/oakcore/ (init/copy/free + self-first
functions), implemented in core/src/capi/
- consumers keep the original C++ API unchanged through same-name
wrapper classes that hold the handle and forward across the C
boundary; original implementations moved to core/src/oliveimpl
(namespace olive::core::internal) and are hidden from export
- TimeRangeList/TimeRangeListFrameIterator are reimplemented inline
over the wrapper (iterators/containers don't cross C ABI)
- generic Value container stays internal (unused by consumers) and is
no longer part of the public umbrella header
- hidden visibility + OAKCORE_BUILD export macro; nm shows zero
olive::* symbols exported
- install into the platform's standard libdir (GNUInstallDirs);
Windows DLLs next to the executables, macOS into the app bundle
- TimelineWorkArea::in/out/length now return by value: the wrapped
TimeRange getters return values, and forwarding them through const
references dangled (found via RenderWorkerFootageTest crash)
- tests: 9 new pure C ABI test executables (oakcore_*_test) covering
every public C function; 4 stale legacy core tests removed (they
targeted a long-renamed API and were never built due to a malformed
option() that also kept OLIVECORE_BUILD_TESTS off)
- CI/CD: oakcore.dll staged for NSIS, liboakcore.so added to the
AppImage validation list, build-tree DLL copies on Windows
- oak-render-worker now builds from worker/ (own CMakeLists.txt) as a
peer of app/; RenderWorkerPool resolves the new build-tree location
- deduplicated the Linux install() rules for the worker
- worker-spawning tests resolve build/worker instead of build/app
- cd.yml: Windows staging copies the worker from its new output path
- Runners: ubuntu-24.04, macos-15, windows-latest (free tier) instead
of WarpBuild 8x/12x instances
- openfx-misc parallelism adapts to nproc (GitHub runners have fewer
vCPUs than the WarpBuild instances)
- Schedule a daily 03:17 UTC run so ccache entries never hit GitHub's
7-day cache expiry and system dependencies stay current
WarpBuild's cache service is billed; GitHub's cache is free (10 GB
repo limit with LRU eviction, ample for ccache entries). All three
cache sites — ccache for the matrix and dynamic-backend jobs, plus the
macOS OpenTimelineIO build — now use actions/cache@v4, which also
removes the Windows-specific fallback step.
Point CCACHE_DIR and the cache action at a fixed workspace directory
instead; checkout cleans untracked files before the restore, so the
ordering is safe.
- Restore ccache with WarpBuilds/cache@v1 on all three OS and the
dynamic-backend job (per-OS keys with branch fallback), and wire it
up through CMAKE_C/CXX_COMPILER_LAUNCHER; builds print ccache -s so
hit rates are visible in CI logs
- Cache the macOS OpenTimelineIO build (keyed on v0.16.0) and skip the
rebuild on cache hit
- Install ccache via apt/brew/MSYS2 as needed; CCACHE_MAXSIZE=1G
- Runner sizes 16x/32x -> 8x (macOS 12x kept); openfx-misc make -j8 to
match
Mike-Solar