31115ab29a
The mark box used the mark's em box (half the base size, square), while painting placed the shaped mark by its tight rect: advance wide, ascent plus descent tall. Text bounds, selection and export therefore disagreed with the canvas; over marks could paint above the bounds. Shape each span's mark once in HorizontalEmphasisMark and place that measured box for paint, text bounds and position data alike. Position data keeps an em box, now centred on the painted mark, because SVG export draws the mark at a font size equal to the box width. Tests that built synthetic placements now pass the mark ink explicitly; their assertions are unchanged. Vertical marks had no mismatch: paint and position data centre on the same point. AI-assisted-by: claude-opus-5-5
Penpot WASM render
This is the canvas-based WebAssembly render engine for Penpot.
Rust & Emscripten
This project is a Rust crate that targets Emscripten (wasm32-unknown-emscripten).
We use wasm32-unknown-emscripten compilation target:
- It compiles Rust code into WASM
- It generates the JavaScript code (“glue”) to load and run the WASM code
Skia
We use Skia, an Open Source 2D graphics library. In particular, the render engine uses Skia via custom binaries of the rust-skia crate.
How to build
With the Penpot Development Environment running, create a new tab in the tmux.
cd penpot/render-wasm
./build
You can also use ./watch to run the build on every change.
The build script compiles the project and copies the .js and .wasm files to the app that uses each target.
Render targets
The same Rust source produces two artifacts, which differ only in compiler options:
| Target | Tuned for | Cargo profile | Consumed by |
|---|---|---|---|
frontend |
speed | release (-O3) |
frontend/resources/public/js |
export |
size | size (-Oz) |
exporter/resources/wasm |
./build # both targets, frontend first
./build frontend # workspace / viewer renderer
./build export # headless exporter renderer
./watch still follows a single target (frontend unless you pass one),
since watching both would rebuild twice on every keystroke.
Each target keeps its own CARGO_TARGET_DIR (target/<target>), so switching
between them does not invalidate the other's cache. Set BUILD_MODE=release
(or NODE_ENV=production) for an optimized build; the default is debug.
Serialize builds in one checkout
All targets in one checkout share one flock lock stored at
render-wasm/.render-wasm-build.lock. The lock covers dependency setup, the
Cargo build, artifact copy, and target cleanup. If another build or cleanup
already holds the lock, the new process prints a waiting message and starts
only after the first process exits.
The watch command takes the lock for each build, then releases it while it waits for source changes. The lock file stays in the checkout after a build, but the operating system releases its lock when the process exits, including after an error or signal.
The lock applies to one checkout and only to commands that use these scripts.
A manual Cargo build can still write to the same target without taking the
lock. Set RENDER_WASM_LOCK_FILE to the same path in each process when you
make different checkouts share a CARGO_TARGET_DIR. The supported Linux build
environment must provide flock from util-linux; flock --version checks this
dependency.
Use the target cleanup commands instead of running cargo clean on the shared
target/ directory:
./clean frontend # remove target/frontend only
./clean export # remove target/export only
Each target writes its own generated shared.js (the enum discriminants the
CLJS side compiles against) next to the code that imports it — respectively
frontend/src/app/render_wasm/api/shared.js and
exporter/src/app/wasm/shared.js. Neither build writes to the other's paths.
Edit your local frontend/resources/public/js/config.js to add the following flags:
enable-feature-render-wasmto enable this render engine.enable-render-wasm-dpr(optional), to enable using the device pixel ratio.
How to test
We currently have two types of tests:
- Unit tests
cd penpot/render-wasm
./test