Skip to content

The truly free, open-source, local-first, cross-platform CapCut replacement. Scriptable through a JSON-RPC API & MCP.

ci: no web build, since Concat ships to desktops and phones only

Jareer0714fb8eae5d
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 7ac73be7..1b8ae482 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -213,38 +213,6 @@ jobs:
src/nextest-*.log
if-no-files-found: ignore
- # The engine crates that carry no native library build for the web. This
- # job is what keeps that true: a std::thread or a blocking wait that
- # creeps into concat-core or concat-project fails here, not when someone
- # next tries the target. See src/README.md for the list.
- web:
- name: Engine on wasm32
- runs-on: ubuntu-latest
- timeout-minutes: 30
- steps:
- - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
- with:
- ref: ${{ inputs.ref }}
- - name: The pinned toolchain
- id: toolchain
- run: echo "channel=$(sed -n 's/^channel = "\(.*\)"/\1/p' src/rust-toolchain.toml)" >> "$GITHUB_OUTPUT"
- - uses: dtolnay/rust-toolchain@e2a55d2ffb04f378e9626c28d38b36d230d1e12f # master (v1), 2026-10-09
- with:
- toolchain: ${{ steps.toolchain.outputs.channel }}
- components: clippy
- targets: wasm32-unknown-unknown
- - uses: Swatinem/rust-cache@6323deb102c322ba6fcbdcafc7e3dddab59af2b6 # v2.9.2
- with:
- workspaces: src
- key: wasm32
- cache-on-failure: true
- - run: >
- cargo clippy
- -p concat-core -p concat-project -p concat-effects -p concat-render -p concat-text
- --features concat-render/gpu --target wasm32-unknown-unknown
- -- -D warnings
- working-directory: src
-
# The window and the engine built where they ship. Linux is the job
# above; macOS and Windows used to be compiled only by a release tag,
# which is where a broken platform path was found. Windows runs the
diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md
index 91655b27..21510038 100644
--- a/ARCHITECTURE.md
+++ b/ARCHITECTURE.md
@@ -40,7 +40,7 @@ manifests, not a wish.
```mermaid
graph BT
- core["concat-core · 2 897<br/>Rational time, arenas, Frame, Timeline,<br/>animation, speed curves, ShaderPass<br/>std only; builds for wasm"]
+ core["concat-core · 2 897<br/>Rational time, arenas, Frame, Timeline,<br/>animation, speed curves, ShaderPass<br/>std only"]
project["concat-project · 10 285<br/>the document: model, commands,<br/>placement rule, undo, concat.json"]
media["concat-media · 9 897<br/>FFmpeg: probe, decode, encode, audio mix,<br/>reader pool, prefetch, hardware decode"]
render["concat-render · 7 450<br/>FramePlan; the wgpu compositor;<br/>SSIM; the CPU oracle (tests only)"]
@@ -104,10 +104,6 @@ Three things the graph says that matter when adding code:
(it has its own `Session`), but for the Remote page, which embeds a
server, and for the register of open project folders the two share.
-The crates that carry no native library, `concat-core`, `concat-project`,
-`concat-effects`, `concat-render` and `concat-text`, build for
-`wasm32-unknown-unknown`, and CI keeps them building.
-
## 2. From the document to a pixel
A clip exists in three shapes on its way to the screen, and each crossing
@@ -544,7 +540,6 @@ flowchart LR
push["push / PR"] --> ci
subgraph ci["ci.yml"]
engine["engine (ubuntu): fmt, locales, models,<br/>clippy -D warnings, test -j1 with lavapipe<br/>(CONCAT_REQUIRE_GPU), perf --check --quick,<br/>grpc clippy + tests"]
- web["web: clippy for core, project, effects,<br/>render, text on wasm32"]
desktop["desktop: macos-15 cargo check;<br/>windows-2022 full tests, 120 min job timeout"]
end
tag["tag v*"] --> rel
diff --git a/src/README.md b/src/README.md
index 26672c68..7037650d 100644
--- a/src/README.md
+++ b/src/README.md
@@ -67,20 +67,6 @@ dependency pinned: `nix build` (then `./result/bin/concat`), `nix run`, or
`nix develop` for a shell with the toolchain and libraries in it. Linux
x86_64 and aarch64.
-The crates that carry no native library - `concat-core`, `concat-project`,
-`concat-effects`, `concat-render` (the wgpu compositor included, on WebGPU)
-and `concat-text` - also build for the web, and CI keeps them building:
-
-```sh
-cargo check -p concat-core -p concat-project -p concat-effects -p concat-render -p concat-text \
- --features concat-render/gpu --target wasm32-unknown-unknown
-```
-
-The rest is native by nature: `concat-media` links FFmpeg, `concat-host`
-talks to the audio device and the file system, `concat-speech` compiles in
-whisper.cpp and sherpa-onnx, and `concat` is a window. `concat-export` goes
-with them for now because its render loop reads through `concat-media`.
-
### Phones
The whole engine and the window build for Android (arm64) and iOS
diff --git a/src/crates/concat-render/src/gpu.rs b/src/crates/concat-render/src/gpu.rs
index 31593719..d8e9cd18 100644
--- a/src/crates/concat-render/src/gpu.rs
+++ b/src/crates/concat-render/src/gpu.rs
@@ -702,10 +702,8 @@ impl WgpuCompositor {
/// adapter where there is not, which draws the same pictures, slower -
/// the one road left since the CPU compositor went.
///
- /// Native only: it blocks on the adapter and device requests, and on the
- /// web there is no thread to block. A web caller awaits those requests
- /// itself and hands the result to [`WgpuCompositor::with_device`].
- #[cfg(not(target_arch = "wasm32"))]
+ /// It blocks on the adapter and device requests. A caller that already
+ /// holds a device hands it to [`WgpuCompositor::with_device`] instead.
pub fn new() -> Option<Self> {
Self::own_api_first(false)
}
@@ -714,7 +712,6 @@ impl WgpuCompositor {
/// Windows, lavapipe on Linux - or `None` where it has none, as a Mac
/// has not. For going on with an export whose GPU has died under it
/// (#223): slower, and nothing left to lose.
- #[cfg(not(target_arch = "wasm32"))]
pub fn software() -> Option<Self> {
Self::own_api_first(true)
}
@@ -722,7 +719,6 @@ impl WgpuCompositor {
/// [`WgpuCompositor::new`]'s order of asking: the platform's own API,
/// then every API wgpu was built with. `software_only` leaves the
/// hardware adapters out.
- #[cfg(not(target_arch = "wasm32"))]
fn own_api_first(software_only: bool) -> Option<Self> {
let own = if cfg!(target_vendor = "apple") {
wgpu::Backends::METAL
@@ -739,7 +735,6 @@ impl WgpuCompositor {
/// A compositor on the best adapter among `backends`, hardware before
/// software - or software alone - or `None` when none of them opens a
/// device.
- #[cfg(not(target_arch = "wasm32"))]
fn on_backends(backends: wgpu::Backends, software_only: bool) -> Option<Self> {
let instance = wgpu::Instance::new(wgpu::InstanceDescriptor {
backends,
diff --git a/src/rust-toolchain.toml b/src/rust-toolchain.toml
index 1d511a0f..b90c0b67 100644
--- a/src/rust-toolchain.toml
+++ b/src/rust-toolchain.toml
@@ -5,6 +5,3 @@
[toolchain]
channel = "1.93.1"
components = ["rustfmt", "clippy"]
-# The engine crates that carry no native library build for the web too, and
-# CI checks that they keep doing so; see README.md.
-targets = ["wasm32-unknown-unknown"]