2026-03-12 14:41:33 +00:00
# Building WEFT OS
## Prerequisites
2026-10-09 08:16:59 +00:00
Linux (x86_64 or aarch64). WEFT OS is built and validated on Linux only; runtime components require Linux kernel interfaces.
2026-03-12 14:41:33 +00:00
2026-03-12 19:48:17 +00:00
System packages (openSUSE):
2026-03-12 14:41:33 +00:00
```sh
2026-03-12 19:48:17 +00:00
sudo zypper install -y \
libwayland-devel libxkbcommon-devel \
2026-03-12 19:52:34 +00:00
libglvnd-devel libgbm-devel libdrm-devel \
libinput-devel seatd-devel libudev-devel systemd-devel \
2026-03-12 14:41:33 +00:00
pkg-config clang cmake python3
```
Rust toolchain: pinned in `rust-toolchain.toml` . Run `rustup show` to confirm the active toolchain matches.
## Workspace crates (non-Servo)
```sh
cargo build --workspace \
--exclude weft-servo-shell \
--exclude weft-app-shell
```
These crates do not require Servo and build in under two minutes.
## weft-compositor, weft-servo-shell, weft-app-shell (Linux)
```sh
cargo build -p weft-compositor
cargo build -p weft-servo-shell
cargo build -p weft-app-shell
```
fix(shells): paint and present Servo frames and report readiness truthfully
With servo-embed, weft-servo-shell and weft-app-shell did not compile
against the selected Servo revision: the keyboard event shape, key and
mouse action variants, frame readback, shutdown, URL type and window
handle APIs had changed. The frame path never called WebView::paint; the
software path rebuilt its softbuffer surface every frame and read pixels
without painting, and the EGL path presented nothing. weft-app-shell
printed READY right after creating its WebView, before any frame, and its
feature-disabled build printed READY and parked. Servo also disables CSS
Grid by default, which the system UI uses.
Both hosts now build against the selected API. Each frame is painted with
WebView::paint and presented: the software path reads the painted frame
back, presents the context and copies it into a persistent softbuffer
surface; the EGL path presents its window context. Nothing is presented
before Servo's first repaint request. Shutdown drops the WebView and
Servo, whose drop completes the shutdown, and exits the event loop. Both
hosts enable layout_grid_enabled.
weft-app-shell prints READY on the first frame presented after the
document has loaded and Servo reports its rendering settled: on load
completion it requests WebView::take_screenshot, which completes only once
every frame has fired load, render-blocking resources, images and web
fonts are done and the rendering is up to date. Feature-disabled builds of
both hosts exit with an error instead of reporting readiness. weft-appd
does not consume the app shell's READY yet.
tests/frame/check_frame.py runs weft-compositor nested on Xvfb with the
winit backend, starts a host showing tests/frame/reference.html and checks
the presented pixels (four colour quadrants and text). For the app host
the page must be on screen within two seconds of READY; --slow-style
delays a render-blocking stylesheet that reveals the page, so premature
readiness fails.
Validation (Ubuntu 24.04, Mesa 25.2.8 llvmpipe, software path):
- check_frame.py --host system, --host app and --host app --slow-style 3:
5 of 5 runs pass each
- negative fixtures fail: page with red and green swapped; a fake host
printing READY without rendering; a local build printing READY on its
first presented frame (--slow-style 4, 3 of 3)
- cargo clippy -D warnings with and without servo-embed; cargo fmt --check
- unverified: EGL path, DRM, resize, output scale, idle wakeup, input
2026-10-09 01:02:26 +00:00
Without `--features servo-embed` , the servo-shell and app-shell binaries compile for type and lint checks but exit with an error at startup; they never report readiness. This is the default and the CI baseline.
2026-03-12 14:41:33 +00:00
## Servo embedding (optional, slow)
```sh
cargo build -p weft-servo-shell --features servo-embed
cargo build -p weft-app-shell --features servo-embed
```
build: select compiling Servo and Stylo fork revisions
Both Servo hosts tracked the marcoallegretti/servo `servo-weft` branch,
which the lockfile resolved to 8e7dc40. That revision's servo-layout does
not compile (E0308 and E0716 in build_backdrop_filter). WEFT also resolved
upstream Stylo dca3934, where backdrop-filter is gated behind
layout.unimplemented, because the Servo fork's own [patch] for Stylo does
not apply to dependents.
Pin Servo by revision to f0bb1aa (8e7dc40 plus the build fix) and add a
root [patch."https://github.com/servo/stylo"] selecting the
marcoallegretti/stylo fork at f1ba496 for all eight Stylo crates Servo
uses. Cargo.lock changes only the 66 source lines of those packages.
Record both pins, their reachable branches, purpose and removal
conditions in SERVO_PIN.md, and restate its limitations against executed
evidence. Update the Nix outputHashes: the Stylo hash is recomputed for
f1ba496 with the method that reproduces the previous dca3934 value; the
Servo hash is not yet known for f0bb1aa and is left as lib.fakeHash, so
Nix builds report the expected value instead of using a stale one.
Validation:
- cargo metadata --locked: ok; no servo/stylo sources remain in the lock
- cargo check -p weft-servo-shell --features servo-embed --locked:
all of Servo, including servo-layout, compiles; the remaining errors are
in the shell's own embedder code
- cargo fmt --check; portable workspace tests: 78 passed
- Nix fetchgit could not run here (archive downloads and the sandbox's
TLS to the proxy fail), so the Servo hash remains unverified
2026-10-09 00:13:08 +00:00
This fetches and compiles the Servo and Stylo forks at the revisions recorded in `Cargo.lock` (see `crates/weft-servo-shell/SERVO_PIN.md` ). Expect 30– 60 minutes on a clean build. Servo's dependencies include SpiderMonkey (C++), which requires `clang` and `python3` .
2026-03-12 14:41:33 +00:00
fix(shells): paint and present Servo frames and report readiness truthfully
With servo-embed, weft-servo-shell and weft-app-shell did not compile
against the selected Servo revision: the keyboard event shape, key and
mouse action variants, frame readback, shutdown, URL type and window
handle APIs had changed. The frame path never called WebView::paint; the
software path rebuilt its softbuffer surface every frame and read pixels
without painting, and the EGL path presented nothing. weft-app-shell
printed READY right after creating its WebView, before any frame, and its
feature-disabled build printed READY and parked. Servo also disables CSS
Grid by default, which the system UI uses.
Both hosts now build against the selected API. Each frame is painted with
WebView::paint and presented: the software path reads the painted frame
back, presents the context and copies it into a persistent softbuffer
surface; the EGL path presents its window context. Nothing is presented
before Servo's first repaint request. Shutdown drops the WebView and
Servo, whose drop completes the shutdown, and exits the event loop. Both
hosts enable layout_grid_enabled.
weft-app-shell prints READY on the first frame presented after the
document has loaded and Servo reports its rendering settled: on load
completion it requests WebView::take_screenshot, which completes only once
every frame has fired load, render-blocking resources, images and web
fonts are done and the rendering is up to date. Feature-disabled builds of
both hosts exit with an error instead of reporting readiness. weft-appd
does not consume the app shell's READY yet.
tests/frame/check_frame.py runs weft-compositor nested on Xvfb with the
winit backend, starts a host showing tests/frame/reference.html and checks
the presented pixels (four colour quadrants and text). For the app host
the page must be on screen within two seconds of READY; --slow-style
delays a render-blocking stylesheet that reveals the page, so premature
readiness fails.
Validation (Ubuntu 24.04, Mesa 25.2.8 llvmpipe, software path):
- check_frame.py --host system, --host app and --host app --slow-style 3:
5 of 5 runs pass each
- negative fixtures fail: page with red and green swapped; a fake host
printing READY without rendering; a local build printing READY on its
first presented frame (--slow-style 4, 3 of 3)
- cargo clippy -D warnings with and without servo-embed; cargo fmt --check
- unverified: EGL path, DRM, resize, output scale, idle wakeup, input
2026-10-09 01:02:26 +00:00
## Frame check
`tests/frame/check_frame.py` runs `weft-compositor` nested on Xvfb with the
winit backend, starts a Servo host showing `tests/frame/reference.html` , and
checks the presented pixels. It needs `Xvfb` , `xwd` , ImageMagick's `convert`
and a Mesa OpenGL driver (llvmpipe is sufficient).
```sh
cargo build -p weft-compositor
cargo build -p weft-servo-shell -p weft-app-shell \
--features weft-servo-shell/servo-embed,weft-app-shell/servo-embed
python3 tests/frame/check_frame.py --host system
python3 tests/frame/check_frame.py --host app
python3 tests/frame/check_frame.py --host app --slow-style 3
```
`--host app` also requires the page on screen within two seconds of
`weft-app-shell` printing `READY` ; `--slow-style` delays a stylesheet that
2026-10-09 09:39:37 +00:00
reveals the page so premature readiness becomes visible. The stylesheet is a
file next to the page that the harness keeps from being opened with a file
lease, so the delay must stay below `/proc/sys/fs/lease-break-time` .
fix(shells): paint and present Servo frames and report readiness truthfully
With servo-embed, weft-servo-shell and weft-app-shell did not compile
against the selected Servo revision: the keyboard event shape, key and
mouse action variants, frame readback, shutdown, URL type and window
handle APIs had changed. The frame path never called WebView::paint; the
software path rebuilt its softbuffer surface every frame and read pixels
without painting, and the EGL path presented nothing. weft-app-shell
printed READY right after creating its WebView, before any frame, and its
feature-disabled build printed READY and parked. Servo also disables CSS
Grid by default, which the system UI uses.
Both hosts now build against the selected API. Each frame is painted with
WebView::paint and presented: the software path reads the painted frame
back, presents the context and copies it into a persistent softbuffer
surface; the EGL path presents its window context. Nothing is presented
before Servo's first repaint request. Shutdown drops the WebView and
Servo, whose drop completes the shutdown, and exits the event loop. Both
hosts enable layout_grid_enabled.
weft-app-shell prints READY on the first frame presented after the
document has loaded and Servo reports its rendering settled: on load
completion it requests WebView::take_screenshot, which completes only once
every frame has fired load, render-blocking resources, images and web
fonts are done and the rendering is up to date. Feature-disabled builds of
both hosts exit with an error instead of reporting readiness. weft-appd
does not consume the app shell's READY yet.
tests/frame/check_frame.py runs weft-compositor nested on Xvfb with the
winit backend, starts a host showing tests/frame/reference.html and checks
the presented pixels (four colour quadrants and text). For the app host
the page must be on screen within two seconds of READY; --slow-style
delays a render-blocking stylesheet that reveals the page, so premature
readiness fails.
Validation (Ubuntu 24.04, Mesa 25.2.8 llvmpipe, software path):
- check_frame.py --host system, --host app and --host app --slow-style 3:
5 of 5 runs pass each
- negative fixtures fail: page with red and green swapped; a fake host
printing READY without rendering; a local build printing READY on its
first presented frame (--slow-style 4, 3 of 3)
- cargo clippy -D warnings with and without servo-embed; cargo fmt --check
- unverified: EGL path, DRM, resize, output scale, idle wakeup, input
2026-10-09 01:02:26 +00:00
Logs and the last capture are written to `target/frame-check` .
test: verify native pointer and keyboard input through weft-compositor
Nothing checked that input reaches a page through the compositor; the
Servo hosts' input forwarding had only been compiled.
tests/frame/check_input.py runs weft-servo-shell with
tests/frame/input.html on the nested desktop and injects real X input
with xdotool into weft-compositor's window. The compositor's winit backend
routes it through its seat to the Wayland client, which has no X display.
The page must react in its presented pixels: clicks 6 px either side of
the boundary between its halves change only the half under the pointer,
and the a and Shift+a key presses arrive with the right key value and
modifier. The harness gives the compositor's X window input focus, as a
window manager would; the client's keyboard focus comes from the
compositor's click-to-focus.
The nested desktop setup shared with check_frame.py moves to
tests/frame/session.py. Xvfb now reports the display it picked through
-displayfd, so concurrent runs cannot share an X server, and a run clears
only the logs and capture it writes.
Validation:
- check_input.py: 3 of 3 runs pass; two concurrent runs pass
- negative pages fail: no script; left click handled by the right half;
an element extending the left half 20 px across the boundary; a key
handler ignoring Shift
- check_frame.py --host system, --host app, --host app --slow-style 3 pass
after the refactor
- unverified: key release and repeat, pointer motion and button release
events, wheel, focus loss, non-US layouts, non-ASCII text, scale changes,
weft-app-shell input
2026-10-09 01:18:53 +00:00
`tests/frame/check_input.py` uses the same nested desktop to send real X
pointer and keyboard input (with `xdotool` ) through `weft-compositor` to
`weft-servo-shell` showing `tests/frame/input.html` , and checks that the
page reacts to clicks on either side of a boundary and to the `a` and
`Shift+a` keys. Logs and the last capture are written to `target/input-check` .
```sh
python3 tests/frame/check_input.py
```
fix(notes): serialise saves and keep unseen notes from being replaced
The editor accepted typing before the stored notes had loaded, and a
load reply then adopted the stored revision for text edited from
nothing, so a save replaced notes the user never saw; the same happened
when typing during a Discard reload. The editor is now read-only until
it holds the stored text.
Two Notes windows saving at once shared one staging file and could each
pass the revision check, storing a mix of both texts. Each save now
holds a lock file while it checks the revision and replaces the notes,
writes its own staging file, syncs it before the rename, and removes
only that file on failure; a lock left by an ended session expires
after ten seconds.
A save larger than the 64 KiB app message limit after JSON encoding was
dropped on its way and left the page waiting forever with Save
disabled. The page now refuses such saves before sending, gives up on a
save after ten seconds without a reply, and the component answers every
request, including malformed ones, so no request goes unanswered.
check_notes.py now pauses weft-runtime with SIGSTOP to make text typed
during a save and keys typed during a Discard reload deterministic; the
previous package fails it by saving text typed before the reload. The
building guide lists the application checks, the pinned toolchain for
reproducing the demo binaries and the demo key path.
2026-10-09 08:55:50 +00:00
The application checks run `weft-appd` with `weft-runtime` (built with
`wasmtime-runtime` ) and `weft-app-shell` (built with `servo-embed` ) on the
same nested desktop and drive the demo apps with real keyboard input:
```sh
python3 tests/frame/check_counter.py # Counter round trip through the session bridge
python3 tests/frame/check_notes.py # Notes stores exactly what is typed, refuses stale saves
python3 tests/frame/check_app_data.py # Notes data from the earlier layout is moved and kept
2026-10-09 09:39:37 +00:00
python3 tests/frame/check_confinement.py # an app page reaches only its UI files and its session bridge
2026-10-09 16:00:38 +00:00
python3 tests/frame/check_activation.py # apps fill the work area above the taskbar; launch and ACTIVATE_APP raise and focus
2026-10-09 17:38:10 +00:00
python3 tests/frame/check_update.py # updates leave running sessions on their revision; failed updates, rollback and uninstall
fix(notes): serialise saves and keep unseen notes from being replaced
The editor accepted typing before the stored notes had loaded, and a
load reply then adopted the stored revision for text edited from
nothing, so a save replaced notes the user never saw; the same happened
when typing during a Discard reload. The editor is now read-only until
it holds the stored text.
Two Notes windows saving at once shared one staging file and could each
pass the revision check, storing a mix of both texts. Each save now
holds a lock file while it checks the revision and replaces the notes,
writes its own staging file, syncs it before the rename, and removes
only that file on failure; a lock left by an ended session expires
after ten seconds.
A save larger than the 64 KiB app message limit after JSON encoding was
dropped on its way and left the page waiting forever with Save
disabled. The page now refuses such saves before sending, gives up on a
save after ten seconds without a reply, and the component answers every
request, including malformed ones, so no request goes unanswered.
check_notes.py now pauses weft-runtime with SIGSTOP to make text typed
during a save and keys typed during a Discard reload deterministic; the
previous package fails it by saving text typed before the reload. The
building guide lists the application checks, the pinned toolchain for
reproducing the demo binaries and the demo key path.
2026-10-09 08:55:50 +00:00
```
2026-03-12 14:41:33 +00:00
## Demo apps (wasm32-wasip2)
fix(notes): serialise saves and keep unseen notes from being replaced
The editor accepted typing before the stored notes had loaded, and a
load reply then adopted the stored revision for text edited from
nothing, so a save replaced notes the user never saw; the same happened
when typing during a Discard reload. The editor is now read-only until
it holds the stored text.
Two Notes windows saving at once shared one staging file and could each
pass the revision check, storing a mix of both texts. Each save now
holds a lock file while it checks the revision and replaces the notes,
writes its own staging file, syncs it before the rename, and removes
only that file on failure; a lock left by an ended session expires
after ten seconds.
A save larger than the 64 KiB app message limit after JSON encoding was
dropped on its way and left the page waiting forever with Save
disabled. The page now refuses such saves before sending, gives up on a
save after ten seconds without a reply, and the component answers every
request, including malformed ones, so no request goes unanswered.
check_notes.py now pauses weft-runtime with SIGSTOP to make text typed
during a save and keys typed during a Discard reload deterministic; the
previous package fails it by saving text typed before the reload. The
building guide lists the application checks, the pinned toolchain for
reproducing the demo binaries and the demo key path.
2026-10-09 08:55:50 +00:00
Each demo is a standalone crate in `examples/` . Pre-built `app.wasm` binaries are committed; they are reproducible with the toolchain pinned in `rust-toolchain.toml` . To rebuild:
2026-03-12 14:41:33 +00:00
```sh
rustup target add wasm32-wasip2
cd examples/org.weft.demo.counter
cargo build --release
# output: target/wasm32-wasip2/release/app.wasm
fix(notes): store notes exactly and never overwrite newer text
The Notes demo exchanged hand-escaped strings: the page turned every
literal backslash-n into a line break on load and save, the component
escaped only some control characters, so other text produced invalid
replies, and every reply replaced the editor contents, discarding text
typed while a save was in flight. Saves wrote the file in place and
replaced whatever another Notes window had stored.
Requests and replies are now JSON objects. The component reports a
revision (a hash of the stored bytes) with each load; a save names the
revision it was edited from and is refused when the stored notes have
changed since, so a second window's text is never replaced silently. The
file is written beside notes.txt and renamed over it. A save
acknowledgement updates only the saved state, never the editor, and
Discard reloads the stored notes.
tests/frame/check_notes.py edits Notes with real keyboard input on the
nested desktop: text with a literal backslash sequence, quotes, a tab and
non-ASCII characters is stored byte for byte, text typed during a save
survives, and a save after another window changed the file is refused.
It fails against the previous package, which turned the seeded "\n"
into a line break. app.wasm is rebuilt and the package re-signed with the
demo key.
2026-10-09 08:47:01 +00:00
cp target/wasm32-wasip2/release/app.wasm app.wasm
rm -rf target
2026-03-12 14:41:33 +00:00
```
fix(notes): serialise saves and keep unseen notes from being replaced
The editor accepted typing before the stored notes had loaded, and a
load reply then adopted the stored revision for text edited from
nothing, so a save replaced notes the user never saw; the same happened
when typing during a Discard reload. The editor is now read-only until
it holds the stored text.
Two Notes windows saving at once shared one staging file and could each
pass the revision check, storing a mix of both texts. Each save now
holds a lock file while it checks the revision and replaces the notes,
writes its own staging file, syncs it before the rename, and removes
only that file on failure; a lock left by an ended session expires
after ten seconds.
A save larger than the 64 KiB app message limit after JSON encoding was
dropped on its way and left the page waiting forever with Save
disabled. The page now refuses such saves before sending, gives up on a
save after ten seconds without a reply, and the component answers every
request, including malformed ones, so no request goes unanswered.
check_notes.py now pauses weft-runtime with SIGSTOP to make text typed
during a save and keys typed during a Discard reload deterministic; the
previous package fails it by saving text typed before the reload. The
building guide lists the application checks, the pinned toolchain for
reproducing the demo binaries and the demo key path.
2026-10-09 08:55:50 +00:00
The signature covers every file in the package directory, so remove the crate's `target` directory before signing it again. The demo packages are signed with the test key in `examples/keys` , which is a fixture, not a signing authority:
```sh
weft-pack sign examples/org.weft.demo.notes --key examples/keys/weft-sign.key
```
fix(notes): store notes exactly and never overwrite newer text
The Notes demo exchanged hand-escaped strings: the page turned every
literal backslash-n into a line break on load and save, the component
escaped only some control characters, so other text produced invalid
replies, and every reply replaced the editor contents, discarding text
typed while a save was in flight. Saves wrote the file in place and
replaced whatever another Notes window had stored.
Requests and replies are now JSON objects. The component reports a
revision (a hash of the stored bytes) with each load; a save names the
revision it was edited from and is refused when the stored notes have
changed since, so a second window's text is never replaced silently. The
file is written beside notes.txt and renamed over it. A save
acknowledgement updates only the saved state, never the editor, and
Discard reloads the stored notes.
tests/frame/check_notes.py edits Notes with real keyboard input on the
nested desktop: text with a literal backslash sequence, quotes, a tab and
non-ASCII characters is stored byte for byte, text typed during a save
survives, and a save after another window changed the file is refused.
It fails against the previous package, which turned the seeded "\n"
into a line break. app.wasm is rebuilt and the package re-signed with the
demo key.
2026-10-09 08:47:01 +00:00
2026-03-12 14:41:33 +00:00
## weft-runtime with Wasmtime
```sh
cargo build -p weft-runtime --features wasmtime-runtime,net-fetch
```
2026-10-09 00:41:43 +00:00
Without `--features wasmtime-runtime` , the runtime compiles for type, lint and unit checks (the portable CI configuration) but exits with an error when asked to run a component; it never reports readiness.
2026-03-12 14:41:33 +00:00
## Signing packages
```sh
weft-pack generate-key ./keys
weft-pack sign ./examples/org.weft.demo.counter --key ./keys/weft-sign.key
weft-pack verify ./examples/org.weft.demo.counter --key ./keys/weft-sign.pub
```
fix(pack): install only trusted packages and bind each app ID to its owner
weft-pack install copied any package into the store: signatures were
never checked, no key was trusted for anything, any publisher could
install over another's app ID and reach its retained data, and the
signing inventory followed symbolic links, so a package could sign and
install files from outside itself.
Installation now requires a signature by a key in the trust store
($XDG_CONFIG_HOME/weft/trusted-keys and /etc/weft/trusted-keys, or
$WEFT_TRUSTED_KEYS alone). Unsigned or untrusted content installs only
with --dev, recorded as development content. The package is copied
into the store, checked and verified on that copy, and renamed into
place, so the verified bytes are the installed bytes.
The first installation of an app ID records its owner beside the app
data; later installations must have the same owner, so neither another
publisher nor a development build can take the ID over. The record
outlives uninstall while app data remains.
The content digest moves to weft_ipc_types::trust and refuses links,
special files and non-UTF-8 paths; it keeps the existing inventory
format, and both committed demo signatures still verify. Catalogs list
only directories named after their package ID. The security guide
described a signature over wapp.toml and app.wasm only; it now
describes the full inventory and the trust and ownership rules.
2026-10-09 13:49:49 +00:00
Installation requires a signature by a trusted key. To install the demos with the demo key, trust it for your user (the key is a test fixture; do not trust it on a system you depend on):
```sh
mkdir -p ~/.config/weft/trusted-keys
cp examples/keys/weft-sign.pub ~/.config/weft/trusted-keys/weft-demo.pub
weft-pack install examples/org.weft.demo.notes
```
2026-10-09 13:59:54 +00:00
Unsigned packages under development install with `weft-pack install <dir> --dev` , recorded as development content and marked `(development)` by `weft-pack list` . See `docs/security.md` for the trust store and app ID ownership.
fix(pack): install only trusted packages and bind each app ID to its owner
weft-pack install copied any package into the store: signatures were
never checked, no key was trusted for anything, any publisher could
install over another's app ID and reach its retained data, and the
signing inventory followed symbolic links, so a package could sign and
install files from outside itself.
Installation now requires a signature by a key in the trust store
($XDG_CONFIG_HOME/weft/trusted-keys and /etc/weft/trusted-keys, or
$WEFT_TRUSTED_KEYS alone). Unsigned or untrusted content installs only
with --dev, recorded as development content. The package is copied
into the store, checked and verified on that copy, and renamed into
place, so the verified bytes are the installed bytes.
The first installation of an app ID records its owner beside the app
data; later installations must have the same owner, so neither another
publisher nor a development build can take the ID over. The record
outlives uninstall while app data remains.
The content digest moves to weft_ipc_types::trust and refuses links,
special files and non-UTF-8 paths; it keeps the existing inventory
format, and both committed demo signatures still verify. Catalogs list
only directories named after their package ID. The security guide
described a signature over wapp.toml and app.wasm only; it now
describes the full inventory and the trust and ownership rules.
2026-10-09 13:49:49 +00:00
2026-03-12 14:41:33 +00:00
## NixOS VM (requires Nix with flakes)
Build the VM image:
```sh
bash infra/vm/build.sh
```
Run with QEMU:
```sh
bash infra/vm/run.sh
```
build: select compiling Servo and Stylo fork revisions
Both Servo hosts tracked the marcoallegretti/servo `servo-weft` branch,
which the lockfile resolved to 8e7dc40. That revision's servo-layout does
not compile (E0308 and E0716 in build_backdrop_filter). WEFT also resolved
upstream Stylo dca3934, where backdrop-filter is gated behind
layout.unimplemented, because the Servo fork's own [patch] for Stylo does
not apply to dependents.
Pin Servo by revision to f0bb1aa (8e7dc40 plus the build fix) and add a
root [patch."https://github.com/servo/stylo"] selecting the
marcoallegretti/stylo fork at f1ba496 for all eight Stylo crates Servo
uses. Cargo.lock changes only the 66 source lines of those packages.
Record both pins, their reachable branches, purpose and removal
conditions in SERVO_PIN.md, and restate its limitations against executed
evidence. Update the Nix outputHashes: the Stylo hash is recomputed for
f1ba496 with the method that reproduces the previous dca3934 value; the
Servo hash is not yet known for f0bb1aa and is left as lib.fakeHash, so
Nix builds report the expected value instead of using a stale one.
Validation:
- cargo metadata --locked: ok; no servo/stylo sources remain in the lock
- cargo check -p weft-servo-shell --features servo-embed --locked:
all of Servo, including servo-layout, compiles; the remaining errors are
in the shell's own embedder code
- cargo fmt --check; portable workspace tests: 78 passed
- Nix fetchgit could not run here (archive downloads and the sandbox's
TLS to the proxy fail), so the Servo hash remains unverified
2026-10-09 00:13:08 +00:00
See `infra/nixos/weft-packages.nix` for the package derivations. Its `outputHashes` hold one hash per git source in `Cargo.lock` (the Servo and Stylo forks); every package vendors them, so they must match the pinned revisions for any Nix package to build. The Nix shell packages do not yet enable `servo-embed` .
2026-03-12 14:41:33 +00:00
## CI
fix(runtime): close fetch and preopen gaps in grant enforcement
A component with a host-specific fetch grant could reach another origin
by setting its own Host header, and could inject request lines through
the method string, which ureq writes unvalidated. Fetch now accepts only
standard methods and refuses Host, framing and hop-by-hop headers. Fetch
grants accept canonical DNS names and dotted IPv4 addresses only, so a
grant can never be written in a form that silently never matches.
--preopen split host and guest at the first `::`, so a host path
containing `::` widened the preopened directory; it now splits from the
right, and appd refuses host paths that are not UTF-8 or contain `::`.
appd validated capabilities one at a time, creating the data directory
before a later capability was refused, and required HOME even for
packages without filesystem capabilities. It now checks every capability
first, resolves host directories only when needed, validates the app ID
with the rule weft-pack uses (now shared in weft_ipc_types::package), and
reports 400, 404, 403 or 500 according to the cause. user-dirs.dirs is
read as a shell would: the last assignment wins, $HOME must be followed
by a slash, escaped values are ignored and a relative XDG_CONFIG_HOME is
not used.
The grants probe now also overwrites, renames, removes and creates
directories in read-only data, and sends a forged Host header, an
injected method and a file: URL; it fails when directory permissions of
a read-only preopen are widened or the header check is removed. A new
CI job builds the runtime with its engine and fetch support and runs the
check. The mount test that sets WEFT_APP_STORE now holds the env lock.
Package images are still located separately from directory installs;
deriving grants from the verified image is part of resolving one package
descriptor.
2026-10-09 08:25:40 +00:00
Jobs run on every push to `main` and on pull requests:
2026-03-12 14:41:33 +00:00
2026-10-09 08:16:59 +00:00
- `portable` — fmt, clippy, tests for the crates that need no Linux system libraries (excludes Wayland crates)
2026-03-12 14:41:33 +00:00
- `linux-only` — clippy and tests for `weft-compositor` , `weft-servo-shell` , `weft-app-shell`
fix(runtime): close fetch and preopen gaps in grant enforcement
A component with a host-specific fetch grant could reach another origin
by setting its own Host header, and could inject request lines through
the method string, which ureq writes unvalidated. Fetch now accepts only
standard methods and refuses Host, framing and hop-by-hop headers. Fetch
grants accept canonical DNS names and dotted IPv4 addresses only, so a
grant can never be written in a form that silently never matches.
--preopen split host and guest at the first `::`, so a host path
containing `::` widened the preopened directory; it now splits from the
right, and appd refuses host paths that are not UTF-8 or contain `::`.
appd validated capabilities one at a time, creating the data directory
before a later capability was refused, and required HOME even for
packages without filesystem capabilities. It now checks every capability
first, resolves host directories only when needed, validates the app ID
with the rule weft-pack uses (now shared in weft_ipc_types::package), and
reports 400, 404, 403 or 500 according to the cause. user-dirs.dirs is
read as a shell would: the last assignment wins, $HOME must be followed
by a slash, escaped values are ignored and a relative XDG_CONFIG_HOME is
not used.
The grants probe now also overwrites, renames, removes and creates
directories in read-only data, and sends a forged Host header, an
injected method and a file: URL; it fails when directory permissions of
a read-only preopen are widened or the header check is removed. A new
CI job builds the runtime with its engine and fetch support and runs the
check. The mount test that sets WEFT_APP_STORE now holds the env lock.
Package images are still located separately from directory installs;
deriving grants from the verified image is part of resolving one package
descriptor.
2026-10-09 08:25:40 +00:00
- `runtime-grants` — clippy and a build of `weft-runtime` with `wasmtime-runtime,net-fetch` , then `tests/runtime/check_grants.py` , which runs a probe component under each kind of capability grant
2026-03-12 14:41:33 +00:00
- `servo-embed-linux` — `cargo check --features servo-embed` for servo-shell and app-shell
docs: add engineering contract, contributor toolkit and templates
Add AGENTS.md as the single engineering contract: sources of truth,
evidence and proof requirements, ownership, working modes, the reviewed
change loop, maintenance isolation, decision authority, dependency and
fork rules, the 0.0.x to 0.1.0 release progression, and branch and
publication rules.
Add .agents/ with scope, implement, verify, maintenance and release
procedures; reviewer, architecture debt scout, simplicity and refactoring
roles; skills for the compositor, Servo rendering, runtime capabilities,
appd sessions, packages and data, Linux deployment, integration
verification and documentation; and verify.py, which runs the existing
check profiles with --locked from any directory, stops on the first
failure and supports --dry-run.
check_toolkit.py validates local links and heading anchors, metadata,
catalog coverage and issue forms. A new contributor-toolkit CI job runs
both. The structure follows the Resina Design System toolkit at
1e768064f57a1e403ab271822383e732795905a6, adapted to WEFT.
Validation, against the committed lockfiles:
- verify.py toolkit: checker passes, 28 runner/checker tests pass
- verify.py portable, demos, packages: pass
- verify.py linux, runtime, servo-embed: fail on existing defects
(clippy collapsible_if in the winit backend, three clippy lints in
weft-runtime with wasmtime-runtime, and an E0308 in the selected Servo
revision's servo-layout); the runner reports the failing status
Add CONTRIBUTING.md, issue forms for defects and changes, maintenance
problems and decisions of direction, and a pull request template asking
for ownership, proof and the validation actually performed. README.md
links the contribution guide.
2026-10-08 23:40:08 +00:00
- `contributor-toolkit` — `python .agents/scripts/verify.py toolkit`