* fix(producer): cache Google Fonts woff2 per subset, preserve unicode-range
Google Fonts' css2 API returns one @font-face per (weight × unicode-range
subset) — e.g. vietnamese, latin-ext, and latin faces for the same weight,
each pointing at a distinct woff2 whose glyphs match its unicode-range.
The on-disk cache keyed woff2 files by `${weight}-${style}` only, ignoring
the subset, so every subset of a weight collided on one filename: only the
first subset in the CSS was downloaded and every later subset read it back.
For families whose CSS lists `vietnamese` first (e.g. Big Shoulders Display)
the `latin` A–Z subset was silently dropped, leaving the embedded font with
almost no Latin glyphs. The injected @font-face also omitted `unicode-range`,
so it advertised coverage it lacked and mismatched glyphs fell back to a
different font — the visible "wrong A" glitch in rendered headlines.
- Key the woff2 cache by a hash of the subset-unique woff2 URL, so each
subset is cached on its own.
- Carry each face's `unicode-range` through to the injected @font-face so
the browser selects the correct subset per codepoint (matching Google's
own CSS semantics).
- In the bundled-font Google supplement, add every subset of an uncovered
weight instead of deduping by weight (which dropped extra subsets).
- Extract per-subset download/cache into a helper to keep fetchGoogleFont
within complexity limits.
Adds a hermetic regression test (injected fetch + temp cache dir) that fails
on the old cache-by-weight behavior and passes with the per-subset cache.
* fix(producer): use atomic write for woff2 font cache (CodeQL)
Replace existsSync+writeFileSync TOCTOU pattern with try-read-first +
O_CREAT|O_EXCL (wx flag) atomic write. Eliminates the race window between
the existence check and the file creation, and prevents symlink-following
in shared temp directories (Lambda /tmp). Concurrent render processes that
race on the same cache entry now resolve gracefully via EEXIST handling.
* fix(producer): avoid os.tmpdir() taint for font cache path (CodeQL)
Replace tmpdir() call with literal "/tmp/hyperframes/fonts" for the
Lambda cache path. Lambda's /tmp is private per execution environment,
not a shared multi-user temp dir — semantically identical but breaks
CodeQL's taint tracking from os.tmpdir() to writeFileSync.
* revert: restore tmpdir() for Lambda font cache path
The hardcoded "/tmp" was a workaround for a CodeQL false positive.
Lambda's /tmp is private per execution environment; the write already
uses O_CREAT|O_EXCL + mode 0o644. Dismissed the alert as false positive
via the code-scanning API instead of warping the code.
* feat(studio): GSAP tween editing in Design panel
Add a GSAP animation editor to the studio Design panel: select an element,
view and edit its tweens (properties, easing, timing), add/delete animations,
and drag custom bezier speed curves — all persisted back to the composition
HTML. Gated behind VITE_STUDIO_ENABLE_GSAP_PANEL.
Parsing of existing GSAP source now uses a recast + Babel AST parser instead of
regex, giving scope resolution, stable tween IDs, and round-trip preservation of
extras and unresolved raw values.
recast compiles to CommonJS that calls require("fs"), which breaks browser and
Vite SSR bundles. To contain it, @hyperframes/core is split into an isomorphic
layer and a Node-only AST layer:
- gsapSerialize.ts holds the recast-free helpers (serialization, keyframe
conversion, validation, shared types). htmlParser.ts is now fully isomorphic.
- parseGsapScript and the script-mutation helpers live in gsapParser.ts,
reachable only via the @hyperframes/core/gsap-parser subpath, loaded
server-side by the studio-api mutation routes and the linter via dynamic
import (recast stays external under SSR).
- The barrel and the gsap-constants subpath are recast-free, so studio browser
bundles never trace recast.
Adds AST parser unit + stress coverage and e2e helpers for the panel.
* fix(lint): await async lintHyperframeHtml in all callers
lintHyperframeHtml became async (gsap rules use dynamic import)
but lintProject and check-hyperframe-static weren't awaiting it,
causing typecheck failures and runtime crashes in CI.
Also wire LintRule type in gsap rules to fix fallow unused-type
finding, and suppress render.ts exported-for-tests symbols.
* fix(producer): normalize error messages to prevent [object Object] in telemetry
When a render fails and the caught value is a plain object (not an Error
instance), String(error) produces [object Object], masking the real error
in PostHog telemetry (~24 errors/day).
Add normalizeErrorMessage() that tries Error.message, string passthrough,
.message on plain objects, JSON.stringify, and String() as a last resort.
Apply it on the two telemetry-feeding paths: the main render failure
handler (renderOrchestrator.ts:2099) and buildRenderErrorDetails
(cleanup.ts), plus the error classifier isRecoverableParallelCaptureError
so timeout detection works even when the thrown value is a plain object.
* fix: address review — normalize CLI telemetry path, captureCost fallback
* fix: use local normalizeErrorMessage in CLI to avoid cross-package resolution
The Vite test runner can't resolve runtime imports from @hyperframes/producer
since its exports point to dist/. Copy the utility into the CLI package and
import locally instead.
* fix(engine): use captureBeyondViewport on all CDP screenshot paths
Chrome's compositor rounds the viewport boundary inward under multi-tab
load, clipping the bottom/right edge of tall portrait compositions
(1080x1920). The explicit clip rect already constrains output to exact
composition dimensions, making the viewport-boundary pre-clip from
captureBeyondViewport:false both redundant and unreliable.
Set captureBeyondViewport:true on all three CDP screenshot call sites:
pageScreenshotCapture, captureScreenshotWithAlpha, and captureAlphaPng.
Add portrait-edge-bleed regression test: 1080x1920 grid with bright
magenta bottom rows, rendered with 4 workers. Any compositor clipping
at the bottom edge drops PSNR sharply against the golden baseline.
Closes#1009
* fix(engine): address review feedback on captureBeyondViewport
- Add backref comments on captureScreenshotWithAlpha and captureAlphaPng
pointing to pageScreenshotCapture for the rationale, so the next reader
doesn't treat the flag as unintentional copy-paste
- Note in test meta.json that the static grid fixture covers the
capture-side clipping path but not the video-element compositor surface
timing that produces the t≈37s self-healing in #1009
* test(producer): use video element in portrait-edge-bleed regression test
Replace the static CSS grid with a 1080x1920 portrait video element —
matches the original bug report shape where the compositor surface
allocation timing causes the bottom-edge clipping. The video has a dark
top region and bright magenta bottom 480px, so any viewport clipping at
the bottom edge drops PSNR sharply. Baseline regenerated in Docker with
4 workers.
* test(producer): add parallel capture regression test
Add a regression fixture that forces workers: 2, ensuring the parallel
capture code path (browser-per-worker in BeginFrame mode) is exercised
in CI. All existing fixtures pin workers: 1, so this is the first test
that would catch a regression in the multi-worker pool isolation fix
from PR #1087.
The composition is 5s @ 30fps (150 frames), which exceeds both
MIN_FRAMES_PER_WORKER * 2 (60) and minParallelFrames (120), so the
parallel coordinator will always split work across workers.
Baseline output/output.mp4 must be generated inside Dockerfile.test
before the fixture can run in CI.
* test(producer): bump parallel capture test to 4 workers
Matches realistic auto-mode worker counts (4-6 on typical machines),
not just the minimum (2) that triggers the bug.
* test(producer): add golden baseline for parallel capture regression
Generated inside Dockerfile.test on amd64 Linux (Docker image
hyperframes-producer:test) to match the CI rendering environment.
* test(producer): address review feedback on parallel capture test
- Add fixture to shard-5 in regression.yml so CI actually runs it
- Reframe description: multi-worker path coverage (frame distribution,
reorder buffer, per-worker browser lifecycle), not GPU-specific crash
guard — SwiftShader CI can't reproduce the hardware compositor race
- Remove dead @keyframes count-up (content doesn't apply to div)
- Remove unused CSS animation reference on .counter
- Regenerate golden baseline with cleaned-up HTML
* fix(producer): replace rAF + CSS keyframes with GSAP in parallel-capture test
The composition used requestAnimationFrame for a frame counter and CSS
@keyframes for animations, which triggered screenshot capture mode
(non-deterministic across workers) and caused 29 PSNR failures in CI.
All animations now use the GSAP timeline, keeping the render in
deterministic BeginFrame mode. Baseline regenerated in Docker.
Renders a 6s composition with a pip video inside a sub-composition
host starting at t=3. Verifies the video is visible during the host
window and not double-offset to t=6. Baseline generated in Docker.
- Add comment explaining hardcoded fps=30 (runtime's canonicalFps
default, not exposed on PlayerAPI)
- Add cross-reference comments between init.ts and fileServer.ts
explaining their different __renderReady timing semantics
The cfr re-encode pass hardcodes `-c:v libx264`. Pairing it with
`codec: "h265"` would silently transcode the h265 chunks to h264.
Detect the encoder discriminant in `meta/encoder.json` and throw a
typed error parallel to the existing non-mp4 format guard, so callers
surface the conflict instead of producing a wrong-codec deliverable.
— Rames Jusso
Distributed-render output today uses -c:v copy through concat → mux →
faststart, which means PTS timestamps from each chunk pass through
unchanged. Container r_frame_rate is exact (#1040 + this PR's parent),
but stream-level avg_frame_rate stays PTS-derived and can land on
fractional rationals like 27648000/921677 over a 60s render. Same for
sub-ms duration drift.
This is the achievable bar within -c copy stream-copy concat. For most
consumers (browser playback, YouTube, etc.) the difference is invisible.
For downstream tools that strict-check avg_frame_rate or
ms-precision duration (broadcast workflows, frame-accurate compositors,
some third-party transcoders), it matters.
Adds an opt-in cfr config flag (default false). When true, the
assemble step's final pass re-encodes with -fps_mode cfr -r <fps>
instead of -c copy, producing exact CFR output. Trade-off: ~2-5x the
stitch time for a 60s 1080p clip; second-generation H.264 quality loss
is negligible at -crf 18 but is non-zero.
The v0.6.39 fix added -r <fps> to the multi-chunk concat ffmpeg
invocation but didn't reach the single-chunk pass-through path,
which is taken when totalFrames * fpsDen / fpsNum fits in one
chunk. Result: 1-chunk renders shipped with fractional
r_frame_rate (e.g. 359/12) while multi-chunk renders shipped
with exact 30/1.
Single-chunk path now goes through the same -r <fps> + -c copy
ffmpeg invocation as the concat path, ensuring uniform exact
r_frame_rate metadata across all chunk-count configurations.
Adds a regression test exercising the 1-chunk path and asserting
r_frame_rate === "<fpsNum>/<fpsDen>" exact.
- Always log when html-in-canvas pins to 1 worker, not just on
explicit --workers override (Vance, James)
- Assert recommendScreenshot is true for htmlInCanvas detection —
pins the load-bearing coupling with the screenshot paint-force
protocol (Vance, Magi)
- Skip capture calibration for htmlInCanvas auto-sized renders to
avoid wasting 600ms–3s on an estimate that gets thrown away (Vance)
- Add TODO documenting Chrome's two root-cause mechanisms: paint
cache race and SwiftShader contention (James, Magi)
Chrome's drawElementImage API does not support concurrent usage across
multiple browser instances — running >1 capture worker causes flickering
artifacts. Detect the layoutsubtree canvas attribute during compilation
and unconditionally pin workers to 1, overriding both auto-sizing and
explicit --workers flags.
The prepareFlattenedInnerRoot function creates a wrapper div when
inlining sub-compositions. This wrapper had no width/height, which
broke CSS height:100% chains — any sub-composition using percentage
heights with flexbox centering would collapse to 0px and render
content at the top instead of centered.
Read data-width/data-height from the inner root and set matching
pixel dimensions on the wrapper's inline style. Applied in both the
compiler (server-side bundling) and the runtime (browser-side
composition loader).
Adds a producer regression test with a centered card sub-composition
that fails without this fix.
When the distributed render path stitches chunks with `-c copy`,
ffmpeg averages the container framerate from PTS rather than
carrying the source's exact rational rate, producing values like
`360000/12001` instead of `30/1` and ~5ms duration drift over
60s.
This is a known ffmpeg behavior at the concat-demuxer-copy
boundary. The industry-standard fix is `-r <fps>` as an input
flag on the concat step plus an output flag on the subsequent
mux step — both with `-c copy` retained, no re-encode required.
Three sites updated:
- `assemble.ts` concat step: `-r <fps>` input flag.
- `chunkEncoder.muxVideoWithAudio`: `-r <fps>` output flag.
- `chunkEncoder.applyFaststart`: same, threaded from caller.
Adds `r_frame_rate` + duration-equivalence assertions to
`assemble.test.ts` to close the regression hole.
The preview command's shutdown handler only closed the HTTP server,
leaving Chrome (browser pool) and ffmpeg processes alive. This caused
silent resource leaks — orphaned processes consuming CPU and RAM with
no parent.
Root cause: preview.ts never called drainBrowserPool() or killed
tracked ffmpeg processes. The thumbnail browser in studioServer.ts
registered its own competing signal handlers that raced with
preview's shutdown.
Fix:
- Add a central process tracker (processTracker.ts) that registers
every spawned ffmpeg across engine and producer packages
- Centralize thumbnail browser cleanup via exported
closeThumbnailBrowser() instead of scattered signal handlers
- Wire preview shutdown to call closeThumbnailBrowser(),
drainBrowserPool(), and killTrackedProcesses() before closing the
HTTP server (embedded mode)
- Add killProcessTree() for dev/local modes where Chrome runs in a
child process tree
- Add startup orphan detection that finds and kills orphaned
chrome-headless-shell/Puppeteer Chrome processes (PPID=1) from
previously crashed sessions
Closes#1038
The rational `Fps = { num, den }` refactor in 5dcc89c broke callers
passing `fps: 30` (the form documented in every code example and used
by external consumers). FFmpeg received `undefined/undefined` as the
framerate, causing a cryptic exit-code error.
Add `FpsInput = number | Fps` and `toFps()` normalizer in
@hyperframes/core. `createRenderJob` now accepts both forms —
plain integers are promoted to `{ num, den: 1 }` at the boundary;
`RenderConfig.fps` stays strict `Fps` internally so no downstream
code changes.
Also fixes the producer and engine docs, which showed phantom
`input`/`output` fields on `createRenderJob` and a wrong
`executeRenderJob(job)` signature (missing `projectDir`/`outputPath`
args).
Closes#1031
Previous baseline was generated on arm64 host, causing 87 failed
frames in CI (amd64). Regenerated with --platform linux/amd64 to
match CI's Chrome/FFmpeg pixel output.