Migrate docker/images/Dockerfile.media-processor from ubuntu:26.04 to
dhi.io/node (Debian 13/trixie), which also drops the manual Node tarball
download since the base image ships it. The -dev tag stays as the final
image: fontforge, woff2 and the graphics libraries are needed at runtime.
scripts/build now assembles the release bundle under target/ (dist/ plus
the manifests and a generated setup script), the way the other modules do,
since esbuild leaves the runtime dependencies external. manage.sh gains
build-media-processor-bundle and build-media-processor-docker-image, both
wired into build-bundle and build-docker-images.
The CI workflows are intentionally left untouched: the module is still work
in progress and its images are not published yet, so this only enables
local builds."
Signed-off-by: David Barragán Merino <david.barragan@kaleidos.net>
sha is the full 40-char SHA in build-bundle.yml's output and
build-docker.yml's new checkout-pinning input, matching git and
GitHub's own convention. short_sha stays internal to each workflow
for the S3 keys and image tags that already used it — build-bundle.yml
no longer exposes it externally since nothing outside consumed it.
No S3 key, image tag, or marker value changes anywhere.
Signed-off-by: David Barragán Merino <david.barragan@kaleidos.net>
* 🐛 Align WASM viewer layers during zoom
Keep the absolute WASM layer at the logical viewer size so its canvas CSS
box and hotspot SVG remain aligned as zoom changes.
Add a Playwright regression covering DOM bounds, drawing-buffer sizing, and
hotspot clicks at zoom 1 and below 1.
Closes#11689
AI-assisted-by: GPT-5
Signed-off-by: makesomethingshit <junsoo1172@gmail.com>
* 🐛 Align WASM zoom regression clicks
Use the rendered WASM DOM selector and design-space points mapped through
canvas bounds for visual clicks.
Wait for the canvas and SVG bounds before checking zoomed-out positions.
AI-assisted-by: GPT-5
Signed-off-by: makesomethingshit <junsoo1172@gmail.com>
* 🐛 Fix WASM zoom render wait
Use Screen2 coordinates for the visual interaction hotspot.
Wait for non-empty canvas pixels after each viewer render.
AI-assisted-by: GPT-5
Signed-off-by: makesomethingshit <junsoo1172@gmail.com>
* 🐛 Match WASM render marker in zoom spec
Require the expected frame screenshot marker before reading canvas pixels.
Keep resized buffers blocked until the new frame draws.
AI-assisted-by: GPT-5
Signed-off-by: makesomethingshit <junsoo1172@gmail.com>
* ✨ Run WASM zoom regression in DPR project
Move viewer zoom coverage under the render-wasm Playwright project.
Assert DPR-scaled buffers and use the stable Zoom out role.
AI-assisted-by: GPT-5
Signed-off-by: makesomethingshit <junsoo1172@gmail.com>
* 🐛 Simplify WASM render wait
Signed-off-by: makesomethingshit <junsoo1172@gmail.com>
* 🐛 Remove unreliable WASM viewer regression
Signed-off-by: makesomethingshit <junsoo1172@gmail.com>
---------
Signed-off-by: makesomethingshit <junsoo1172@gmail.com>
_DEVELOP and _STAGING chain build-bundle.yml -> build-docker.yml,
both invoked with gh_ref as a branch name. Each reusable workflow
re-resolves that branch with its own checkout, at its own time.
build-docker's checkout happens minutes after build-bundle's, after
new commits can have landed on the branch, so it can compute a
different sha than the one build-bundle actually bundled and
uploaded to S3.
The Prepare job then requests a key that was never uploaded:
fatal error: An error occurred (404) when calling the HeadObject
operation: Key "penpot-sha-<newer-sha>.zip" does not exist
Expose build-bundle.yml's resolved sha as a workflow_call output and
thread it through _DEVELOP, _STAGING and _ADHOC as build-docker.yml's
new optional sha input, which pins its checkout instead of
re-resolving gh_ref. gh_ref keeps naming the branch tag to move.
_TAG is unaffected: it builds from a pushed git tag, which cannot
move the way a branch can, so build-bundle and build-docker already
agree on one commit there.
Signed-off-by: David Barragán Merino <david.barragan@kaleidos.net>
Mirrors the devenv change in 3385a65 (docker/devenv/docker-compose.infra.yml).
Mailpit keeps messages in memory only (no persistence), consistent with
mailcatcher's previous behavior; the volume and MP_DATABASE env var are
left commented for anyone who wants to opt in later. UI port changed from
1080 to Mailpit's native 8025.
Signed-off-by: David Barragán Merino <david.barragan@kaleidos.net>
The seven creation commands no longer accept an optional client
id: create-file, create-project, create-team,
create-team-with-invitations, upload-file-media-object,
create-file-media-object-from-url and assemble-file-media-object.
The server always generates the identifier; a sent id is ignored.
Malli maps are open and the RPC layer never strips unknown params,
so the handlers that would still honor an id (create-file,
create-project) now drop it explicitly. Internal callers that pass
remapped ids (project duplicate, binfile import) keep working.
Closes#11783
AI-assisted-by: muse-spark-1.3-contributor
dhi.io/node:24.20.0-debian13-dev does not pre-create /usr/local/bin
the way a regular Debian image does, so tar -xzf ... -C /usr/local/bin
failed with "Cannot open: No such file or directory" right after the
checksum check passed. Introduced in #11790 when pnpm moved from
Corepack to a downloaded standalone binary.
Dockerfile.media-processor and docker/devenv/Dockerfile are unaffected:
both extract into /opt/node/bin, which already exists from the prior
Node.js install step.
Signed-off-by: David Barragán Merino <david.barragan@kaleidos.net>
* 🐛 Bound decompressed size of JSON entries on binfile v3 import
Every JSON/text zip entry (manifest, files, pages, shapes, colors, components, typographies, tokens, plugin-data) was decompressed without any size limit, letting a small .penpot archive exhaust the backend heap (GHSA-qcw7-v626-g6cf). Only binary storage blobs were guarded. Reuse the existing size-limiting-stream guard on the text path: 20 MiB cap per entry, 200 MiB cumulative budget per import job, plus a cheap declared-size pre-check. Both limits are configurable and wired through the binfile, management and debug entry points. Adds zip-bomb regression tests for the file entry, the synchronous manifest read and the cumulative budget. Closes#11606
AI-assisted-by: muse-spark-1.3-contributor
* ♻️ Uniform binfile import limits behind init-limits
Move the binfile import limits to a single source of truth in app.binfile.common (default-* vars) and drop the duplicated entries from config/default; env overrides keep working through the schema. Resolve all limits once per job with init-limits (::max-size, ::total-max, ::current-size, ::max-object-size, ::max-zip-entries) instead of rebuilding the map per zip entry. Thread cfg as the first arg through the v3 readers, collapse read-plain-entry into read-entry, and give size-limiting-stream a single explicit-counter arity. v1 keeps using the compiled default (mechanical rename only). No behavior change. AI-assisted-by: muse-spark-1.3-contributor
* ♻️ Rename binfile limits to text-entry/binary-entry terms
Use text-entry/binary-entry vocabulary consistently across config keys, bfc input keys, default-* vars and the limits resolved by setup-limits (::max-text-entry-size, ::max-text-total-size, ::current-text-size, ::max-binary-entry-size). Rename init-limits to setup-limits. No behavior change. AI-assisted-by: muse-spark-1.3-contributor
* ♻️ Rename cumulative text counter and document binary limit
Rename ::current-text-size to ::accumulated-total-text-size for clarity and expand the default-max-binary-entry-size comment to match the other limit vars. No behavior change. AI-assisted-by: muse-spark-1.3-contributor
* ♻️ Harden binfile guards and prove budget accumulation
Add a regression test that only passes when text bytes accumulate across entries (budget between largest entry and summed total; verified red against a per-entry atom). Include the entry name in streaming-guard errors, count skipped bytes against the budget with a direct unit test, and forward all four limit keys in get-manifest. No behavior change. AI-assisted-by: muse-spark-1.3-contributor
* ♻️ Thread cfg through get-manifest
get-manifest now takes the caller cfg and resolves limits with setup-limits like the import job itself, instead of building a single-use mini-cfg from cf/get. No behavior change. AI-assisted-by: muse-spark-1.3-contributor
Lock in the fix from 31b73460c3 (#11803) with a regression
test for the exact reported scenario: loading the viewer
with a URL that already contains `zoom=fill`.
At 2.18.0-RC5 `update-zoom-querystring` navigated without
any comparison, so the load sequence bundle-fetched →
zoom-to-fill → update-zoom-querystring → nav → navigated
re-ran forever and crashed the page with React error #185
("maximum update depth exceeded"). The guard added in
31b73460c3 breaks the cycle; the new test asserts that a
bundle fetch against a `zoom=fill` route emits no
navigation events.
Also updates the dashboard/viewer frontend memory to
document the guard and the loop it prevents.
AI-assisted-by: glm-5.3-flash
* 🐛 Trim linked-library data in view-only bundle for share links
The anonymous get-view-only-bundle RPC merged each linked library whole,
exposing library pages the share link never granted. For share-link
permissions, each library is now reduced to the narrow data keys with
its own pages dropped and only the components referenced by the allowed
pages kept (nested references followed); membership bundles are
unchanged.
Closes#11617
AI-assisted-by: muse-spark-1.3-contributor
* 🐛 Resolve nested library components via main instance in bundle trim
Stored components carry no objects, so the transitive walk missed nested
components. Follow references through the main-instance subtree instead,
share the narrow data keys between the primary and library scopes, and
cover component filtering with a real-instance RPC test.
Review follow-ups F1-F3 for #11617
AI-assisted-by: muse-spark-1.3-contributor
* 🐛 Address review findings on view-only library trim
Hoist invariant refs out of the fixpoint, make the cross-library
main-instance fallback deterministic, pin the trimmed envelope,
and add RPC tests for disallowed-page isolation and
cross-library nesting.
Follow-ups to #11617
AI-assisted-by: muse-spark-1.3-contributor
* ♻️ Index libraries by id with d/index-by
Replace the manual into/juxt index with the shared helper.
No behavior change.
AI-assisted-by: muse-spark-1.3-contributor
* 🌐 Complete Catalan translations in frontend
Complete the Catalan (ca.po) locale to 100% coverage against en.po,
using es.po as support reference. Adds the 1439 missing entries
across workspace, dashboard, labels, shortcuts, subscription,
errors, modals and onboarding, keeping vosaltres treatment and
IEC/Termcat terminology consistent with the existing strings.
Normalizes placeholders and plural forms, drops the 14 stale
obsolete entries and canonicalizes the file with the repo
translations script.
Closes#11739
AI-assisted-by: muse-spark-1.3-contributor
* 📚 Add frontend translations memory with Catalan criteria
Record the PO workflow, the sync fuzzy-flag gotcha and the
Catalan glossary and tone agreed upon while completing ca.po,
and link the new memory from the frontend core routing.
AI-assisted-by: muse-spark-1.3-contributor
* 🔧 Add gettext to devenv image
Provide msgfmt and msgattrib in the dev environment for
checking PO translation files.
AI-assisted-by: muse-spark-1.3-contributor
* 🌐 Fix Catalan translations and add PO checker
Review of the missing-whitespace pattern found ~90 glued words
across 75 entries, plus 4 lost plural forms and 2 placeholder
mismatches verified against tr call sites. All fixed in ca.po.
Adds frontend/scripts/check-translations.js (vocabulary-free PO
QA: glued words, punctuation, placeholders, plurals) with
--self-test, wired as pnpm run check-translations and
documented in mem:frontend/translations.
AI-assisted-by: muse-spark-1.3-contributor
* 🌐 Multi-locale PO checker with word catalogs
Split the checker engine from its word lists: ca/es catalogs now
live in scripts/check-translations/words.<locale>.txt and all
messages are in English. Adds an es seed (calibrated to zero
errors) and fixes 7 typos it found in es.po. Universal checks
(placeholders, plurals, punctuation) run without a catalog.
AI-assisted-by: muse-spark-1.3-contributor
* 🌐 Merge PO checker into translations.js
Fold check-translations.js into translations.js as a check
subcommand reusing its locale helpers; word lists stay in
scripts/check-translations/words.<locale>.txt. Also fixes the
getopts stopEarly bug that made -l useless after the command
(sync -l ca synced every locale), drops dead lodash import
and code, unifies help and exit codes. Removes the
check-translations package alias; use translations.js
check -l <locale> with explicit -l.
AI-assisted-by: muse-spark-1.3-contributor
* 🌐 Keep unused placeholders out of the gate
Reverts the %s-stripping on unused auth.terms-privacy-agreement:
the links mirror its markdown sibling and a reactivation may
need them. Placeholder mismatches on #, unused keys now warn
instead of failing, and the rule is recorded in
mem:frontend/translations.
AI-assisted-by: muse-spark-1.3-contributor
* ✨ Enable closed schemas for RPC methods
* 🐛 Fix duplicate make-dummy-request test helper definition
The branch added a variadic DummyRequest/make-dummy-request pair but
left the pre-existing single-arg definition in place. Because it was
loaded last, zero-arg (make-dummy-request) calls added by
prepare-rpc-params and rpc-nitrate-test threw ArityException, which
broke 384 tests and caused 14 downstream assertion failures.
Remove the stale duplicate so the variadic definition is the only
one, and drop the now-unused yrq alias and duplicate yres alias.
AI-assisted-by: deepseek-v4.1-flash
* ✨ Add focused tests for make-dummy-request helper
Pin the call contract of make-dummy-request, which the suite uses
in three styles: no arguments, a single options map, and keyword
arguments. The helper's redefinition shadowing in 8ca95adb98 was
only caught by a full-suite run with hundreds of unrelated errors;
these tests fail locally in a focused --focus run.
Cover the zero-arg defaults, map and keyword overrides, the
:body-bytes -> ByteArrayInputStream wrapping, :body-stream
precedence, and cookie readback. Also clarify the docstring to
list all supported call styles.
AI-assisted-by: deepseek-v4.1-flash
* 🚑 Prevent RPC client params from overriding auth context
Strip qualified keys from decoded request params before merging
them with the server-built auth context, so transit bodies can
no longer override ::profile-id, ::auth-type or ::token-perms.
Adds a regression test proving the override and the fix.
AI-assisted-by: muse-spark-1.3-contributor
* 📚 Merge backend subtleties memories under generic name
Rename rpc-db-worker-subtleties to subtleties and fold in
http-storage-filedata-subtleties, so the name no longer
enumerates topics. Update all mem: references accordingly.
AI-assisted-by: muse-spark-1.3-contributor
* ✨ Add realistic tests for RPC auth override
Cover the transit wire vector and the real wrapped :get-profile
method with two database profiles, proving a session cannot read
another profile by smuggling :app.rpc/profile-id in the body.
AI-assisted-by: muse-spark-1.3-contributor
* ✨ Add e2e test for RPC auth context override
Parametrize rpcPost with contentType, accept and query so e2e
can send hand-written transit bodies without new dependencies.
The new test proves a transit-smuggled :app.rpc/profile-id no
longer overrides the session in get-profile. Also fix the demo
email assertion in auth-flow to the current uuid format.
AI-assisted-by: muse-spark-1.3-contributor
The unauthenticated REPL server started in --multi-user mode whenever DEVENV or REPL_ENABLE was set, unlike developer tools which are already blocked there. Gate its creation behind shouldStartReplServer, mirroring the developer tools policy, so it never starts with --multi-user. Give it a dedicated PENPOT_MCP_REPL_HOST knob defaulting to localhost so it no longer inherits the main server 0.0.0.0 bind in Docker, keep the devenv script on 0.0.0.0, and document and test the new behavior. Closes#11631 AI-assisted-by: muse-spark-1.3-contributor
The plugin WebSocket bridge ignored PENPOT_MCP_SERVER_HOST and bound all interfaces,
exposing unauthenticated task dispatch to the network in single-user mode. Pass mcpServer.host
into WebSocketServer, mirroring the ReplServer fix, so the bridge binds localhost by default and
0.0.0.0 only on explicit opt-in.
Closes#11603.
AI-assisted-by: muse-spark-1.3-contributor
* 🐛 Restore the missing handler when aligning a one-handler node
A curve node with one of its handlers removed could not be switched
to aligned or equal: the collapsed handler stayed on the node, or the
line on the other side stayed a line, so nothing happened. Switching
to equal could even collapse the remaining handler.
Such a node now gets a mirrored handler on the other side, turning
that line into a curve when needed, also across the start node of a
closed path. Removing a handler also resets the node to independent.
Closes#11703
AI-assisted-by: claude-opus-5
Signed-off-by: Shreyash Agare <264953665+ShreyashAgare26@users.noreply.github.com>
* 🐛 Add some quality improvements
---------
Signed-off-by: Shreyash Agare <264953665+ShreyashAgare26@users.noreply.github.com>
Co-authored-by: Shreyash Agare <264953665+ShreyashAgare26@users.noreply.github.com>
Co-authored-by: alonso.torres <alonso.torres@kaleidos.net>
Subscribe-file and subscribe-team handlers now verify the requesting
profile has read permissions on the target resource before creating
a subscription. Pointer-update handler now validates that the message
file-id matches the subscribed file-id before publishing.
Closes#11067
AI-assisted-by: mimo-v2.5-pro
Snapshot only the scaled shadow bounds instead of the full reusable
filter surface. This avoids retaining viewport-sized textures for each
small shadow while keeping cached pixels valid when the surface is reused.
Use the bounded snapshot for both shape and frame shadow caches. Add
regression tests for rounding, scaling, surface limits, and snapshot reuse.
AI-assisted-by: gpt-6
* ⚡ Build the frontend bundle once for all E2E suites
Merge tests-integration, tests-composable-suite and tests-plugin-api-suite
into one "CI: E2E" workflow. Each of the three ran its own full
frontend/scripts/build on every PR, so one PR paid the build three times.
The new build-bundle job restores actions/cache key frontend-bundle-<sha>,
runs frontend/scripts/build only on a miss and saves the key before the
job ends. The integration shards, the composable suite and the mocked
Plugin API suite now all need build-bundle and restore the same key with
fail-on-cache-miss, so none of them builds. A workflow re-run of the same
SHA reuses the cached bundle instead of rebuilding it.
Triggers become the union of the previous paths (frontend, common,
render-wasm, plugins): the bundle embeds the built plugins, so a plugins
change runs the whole set. workflow_dispatch keeps running the
integration job only, as before.
Job names are kept identical on purpose: they are the GitHub check
contexts and branch protection may match them by name.
Docs: new mem:frontend/e2e-ci-workflow records the build-once contract,
referenced from mem:frontend/core and mem:frontend/testing; the composable
memory and both suite READMEs are updated.
AI-assisted-by: deepseek-v4.1-flash
* 🐛 Fix mocked plugin suites crashing without frontend deps
The mocked CI drivers shelled out to frontend/scripts/e2e-server.js,
which imports express from frontend/node_modules. CI jobs install
only plugins/ deps, so the import failed with ERR_MODULE_NOT_FOUND
and the run timed out waiting for localhost:3000.
Serve the prebuilt bundle with a zero-dependency static server
built into each driver (ci/static-server.ts, kept in sync in both
suites) plus node:test coverage for it.
AI-assisted-by: muse-spark-1.3-contributor
* ⬆️ Update devenv dependencies
Update Node.js, OpenCode, clj-kondo, Babashka, Pixi, GitHub CLI, uv,
and Serena to their current stable releases.
AI-assisted-by: gpt-5.6-sol
* ⬆️ Update devenv to Java 27
Use Zulu JDK 27 in the development image for compatibility testing.
Update the official checksums for both supported architectures.
AI-assisted-by: gpt-5.6-sol
* 🐳 Replace MinIO with RustFS in devenv
Run RustFS as the development S3 service and wait for its health check.
Install a pinned AWS CLI with checksums and use it to create the bucket
idempotently from each backend entry point.
Keep the old MinIO volume untouched and use a new RustFS volume.
AI-assisted-by: gpt-5.6-sol
* 🐳 Replace MailCatcher with persistent Mailpit
Run Mailpit as the devenv SMTP sink while preserving mailer:1025 and the
localhost:1080 UI.
Store its SQLite inbox in a named volume and wait for the readiness
endpoint before starting runtime containers. Bind the web UI to loopback so
development emails stay local.
AI-assisted-by: gpt-5.6-sol
* ⬆️ Update Node.js to 24.21.0
Align the host NVM version with the Node.js version used by devenv.
AI-assisted-by: gpt-5.6-sol
* ⬆️ Update devenv to PostgreSQL 18.6
Run PostgreSQL 18 with its versioned volume layout and a TCP readiness
check that ignores the temporary initialization server.
Install the matching client, create penpot_nexus, and preserve the old
PostgreSQL 16 volume for rollback or logical migration.
AI-assisted-by: gpt-5.6-sol
* 🐳 Expose RustFS ports in devenv
Publish the RustFS S3 API and management console on localhost port 9000
and 9001.
Keep both bindings on loopback so object storage is not exposed to the local
network.
AI-assisted-by: gpt-5.6-sol
* 🐳 Install standalone pnpm in devenv
Install pnpm 12.5.0 from architecture-specific release archives and
verify their published checksums.
Remove the Corepack setup while allowing pnpm to honor the project
packageManager pins.
AI-assisted-by: gpt-5.6-sol
* 🔥 Remove corepack, use system pnpm everywhere
Corepack is gone from Node 25+, so every `corepack enable` call
fails. pnpm now ships as a system binary (devenv, CI runners and
Docker images install it directly) and auto-downloads the version
pinned in `packageManager` on mismatch.
Scripts, workflows and Dockerfiles call `pnpm` straight away; the
three deploy workflows use a single `pnpm/setup@v2` step; and the
new `scripts/sync-pnpm-version` stamps all 35 `packageManager`
fields from the system pnpm, replacing the `corepack use` sweep.
AI-assisted-by: muse-spark-1.3-contributor
* 🐛 Fix exporter watch missing render-wasm build step
The exporter watch compiled CLJS requiring the generated
src/app/wasm/shared.js, which only render-wasm/build export
produces. Without it shadow-cljs failed with a cryptic missing
./shared.js dependency. Run build:wasm before watching, as
the frontend watch:app and exporter scripts/build already do.
AI-assisted-by: muse-spark-1.3-contributor
* 🔧 Add opencode V2 support and adapt plugins
Register the penpot tools for both opencode V1 (server())
and V2 (setup() with JSON Schema inputs) from a single
dependency-free plugin file, sharing the psql and
paren-repair runners between both paths.
Install the opencode2 binary side-by-side with V1 in the
devenv image and document the dual registration in the
paren-repair and psql memories.
AI-assisted-by: muse-spark-1.3-contributor
* ⬆️ Update pnpm and opencode
* ✨ Restrict optional RPC ids to user-provided UUIDs
Add ::sm/user-provided-uuid, backed by a version and variant
aware regex that only accepts v4, v7 and v8 instances. Use it
for the optional :id of the creation RPC commands so reserved
versions such as v3 are rejected at validation time. Reads
such as get-team keep the lax ::sm/uuid. Cover the predicate
and the schema on both JVM and JS runtimes.
AI-assisted-by: muse-spark-1.3-contributor
* ✨ Cover id version restriction at the RPC boundary
Add backend regression tests proving the seven creation commands
reject reserved-version ids (v3) with :params-validation and
accept v4 ids (plus v7/v8 on create-team) through the real
decode and validate path. Also drop two duplicated assertions
and document the version and variant of every fixture UUID in
user-provided-test.
AI-assisted-by: muse-spark-1.3-contributor
* ♻️ Keep agent content on standard paths, drop client shims
Agent guidance stays where the ecosystem reads it: AGENTS.md at the
root and the skills in .agents/skills. Codex scans .agents/skills and
follows symlinks, opencode lists it among its discovery locations, and
Cursor, Zed, Warp, Cline and Amp use the same project path.
Claude Code reads neither, so it needed CLAUDE.md and .claude/skills.
Committing those two paths is not a neutral default. Both are
gitignored, git treats ignored files as expendable, and a checkout
that needs the path deletes whatever a contributor keeps there without
a warning. Neither sparse-checkout nor skip-worktree offers a way out,
because both make the next pull abort. The two links are now created
by the contributor, once per clone, and documented in
.agents/README.md.
AGENTS.local.md and .agents/local/ give every client one personal
layer that composes on top of the shared one, instead of one file per
client inside the repository.
sync-workspace seeds the two links for ws1 and above, the way it
already seeds the gitignored config.js, so the devenv workflow does
not change.
AI-assisted-by: claude-opus-5
Signed-off-by: Álvaro Tejero Cantero <alvorithm@teje.ro>
* ♻️ Lean on Claude Code reading AGENTS.md, drop the pointer
Claude Code 2.1.277 reads AGENTS.md through the built-in agents-md
plugin, whose default mode loads it in a project that has no
instruction file of its own. Deleting the committed CLAUDE.md is what
turns that on here, so the pointer a Claude user had to create is no
longer part of the setup, and only the .claude/skills link remains:
the plugin covers instruction files, and skills still load from that
path alone.
The instruction that replaces the pointer is a removal. A CLAUDE.md,
.claude/CLAUDE.md or CLAUDE.local.md anywhere from the root down to
the working directory hands the project back to CLAUDE.md and hides
this repository's AGENTS.md, while .claude/rules/*.md and the
person's own ~/.claude/CLAUDE.md do not count and stay usable for
personal steering. CLAUDE.local.md joins the ignore list, and a hard
rule in AGENTS.md forbids all three.
The docs name the version, say the fallback has not reached Bedrock,
Vertex or Foundry, and say the stable auto-update channel runs about
a week behind the default one, so the symlink is documented as the
stopgap for those cases and as something to drop afterwards.
sync-workspace seeds the skills link alone for the same reason.
AI-assisted-by: claude-opus-5
Signed-off-by: Álvaro Tejero Cantero <alvorithm@teje.ro>
* 📎 Remove skills directory mention on AGENTS.md
Signed-off-by: Andrey Antukh <niwi@niwi.nz>
---------
Signed-off-by: Álvaro Tejero Cantero <alvorithm@teje.ro>
Co-authored-by: Andrey Antukh <niwi@niwi.nz>
* ✨ Auto link tokens when adding external libraries (provisional)
* 🔧 Refactor tokens-lib initialization
* 🔧 Add separated TokenStatus to store status apart of TokensLib
* 🔧 Make all status operations use the new data structure
* 🔧 Normalize status helper functions and access token sets by id
* 🔧 Rename :tokens-file to :tokens-source
* 🎉 Allow the user to choose the tokens-source of a file
* 🎉 Make tokens library readonly when it's in an external file
* 🎉 Show tokens in library summaries
* 🎉 Show source info in sidebar
* 🔧 Fix integration tests
* 🐛 Propagate changes of token values in external library
* 🎉 Layout updates
* 🔧 Refactor tokens source calculations
* 🔧 Add harder checks for nil or empty values in everything
* 🐛 Fix some integration tests
* 🔧 Add integration tests for tokens in external libs
* 🔧 Validate and repair missing tokens status
* 🎉 Make ui changes optional with config flag
* 🐛 Propagate tokens after synchronizing components in ext library
* 🐛 Propagate tokens after creating new instances
* 🐛 Propagate tokens after synchronizing tokens in ext library
* 🐛 Add a tokens source icon to libraries section (#11439)
* 🐛 Add a tokens source icon to libraries section
* 🐛 Fix ellipsis on library names
* ♻️ Remove code under flag on legacy component
* 🐛 Fix token theme name on inspect tab
* 🎉 Add changes notification (#11476)
* 🎉 Add changes notification
* ♻️ Change fn names
* 🐛 Fix tokens source label truncation and missing translations (#11533)
* 🐛 Fix tokens source label truncation and missing translations
The tokens source file name always showed, even for the current file,
and long names wrapped onto a second line instead of truncating
because the header used flex-wrap and overflow-wrap: break-word
instead of single-line ellipsis.
Show the source row unconditionally (it now displays "This file" when
the source is the current file, matching the connected-library case),
truncate the file name to one line with an ellipsis, and only attach a
tooltip with the full name when the text is actually truncated.
Replace the hardcoded UI strings with translated ones and add their
English and Spanish entries.
AI-assisted-by: claude-sonnet-5
* 🐛 Remove redundant effect dependency in tokens source
file-name-truncated? was listed as a dependency of the with-effect
that checks and observes label truncation, even though it isn't
read inside the effect body. Since the effect itself flips that
state via check-file-name-truncated, including it as a dependency
caused the ResizeObserver to be needlessly disconnected and
reconnected on every truncation change.
AI-assisted-by: claude-sonnet-5
* 🐛 Fix small visual error
* 🐛 Fix problem with plugins
* 🐛 Fix playwright tests
---------
Co-authored-by: Eva Marco <evamarcod@gmail.com>
Co-authored-by: Eva Marco <eva.marco@kaleidos.net>
Co-authored-by: alonso.torres <alonso.torres@kaleidos.net>
* 🌐 Translate import dialog strings
Several plain hardcoded strings in the import dialog (the file
rename aria-label, library-resolution messages, table headers, and
the manually-linked summary labels) were never wired to the i18n
system. Wire them up via tr and add the corresponding English and
Spanish entries to the translation catalogs.
* 🐛 Fix library-resolution summary layout in import dialog
The auto-linked/manually-linked summary list had several layout bugs:
name/selection columns stretched unevenly, the auto-linked badge sat
nested inside the name's ellipsis text (unreliable across browsers),
and a long unbreakable library name could grow the whole modal past
its fixed width because a bare 1fr grid track has no minimum-size
cap. Fix the column sizing, split the badge out as a sibling of the
name, and clamp the modal content column with minmax(0, 1fr).
* ✨ Add Skip button to library resolution wizard
Add a "Skip" action to the per-file library resolution step so users
can leave all of a file's pending libraries unconnected and move on
to the next one, instead of being forced to pick a candidate or go
back. Clears any pending selections for the current file before
advancing the wizard, so the summary correctly shows those libraries
as unlinked.
* 🐛 Fix stale translation key on library resolution button
The primary action button in the library resolution wizard referenced
"dashboard.import.review-links", a translation key that no longer
existed, leaving the button with no text. Rename it to
"dashboard.import.connect-selected-libraries" to match the actual
button label.
* 🐛 Align no-selection state with arrow and file name
.summary-no-selection was missing display:flex/align-items:center,
so its icon and text weren't vertically centered against each other
or against the file name column in the import library summary.
* 💄 Use select-accent color for auto-linked badge border
* 💄 Change import summary columns from 50/50 to 40/60 split
Give the "new library" column more room than the "original library"
column in the manually-linked libraries table (header and rows), since
it usually carries the longer "name (project)" text.
* 🐛 Divide back buttons from forward buttons
* 🐛 Fix dropdown typography
* 🎉 Add Menu design-system component
Adds Menu, MenuItem, MenuSeparator, SubMenu, and ContextMenu to the
shared UI package and exposes them through the CLJS design-system
wrapper, with Storybook stories and MDX docs.
Built on react-aria-components for keyboard navigation, focus
management, and dismissal. Penpot's own DS buttons aren't
react-aria-aware, so trigger positioning, focus-on-open, and
close-on-select are wired explicitly instead of relying on the
library's default trigger detection.
Includes a temporary manual-test harness in the dashboard to check
the components against the real app. CSS is functional but doesn't
match the DS visual design yet — that comes in a follow-up.
AI-assisted-by: claude-sonnet-5
* ✨ Add left/right corner placements to Menu design-system component
Menu and ContextMenu only exposed 8 of react-aria's placement values,
missing every left/right corner variant (right bottom, right top,
left bottom, left top) that the top/bottom sides already had via
start/end.
Add the four missing corners, matching the start/end pattern already
used for top/bottom, so a menu can open toward any corner of its
trigger.
AI-assisted-by: claude-sonnet-5
* ✨ Add drilldown variant to SubMenu design-system component
SubMenu only opened as a flyout: a nested popover next to the
trigger item. That doesn't scale to a tree too deep or wide for a
chain of flyouts, e.g. move-to-project's team -> project nesting,
which needs a mobile-style drilldown (replace the current items with
the submenu's own, plus a way back) instead.
Add a `variant` prop, `"flyout"` (default, unchanged) or
`"drilldown"`. Menu and ContextMenu each keep a navigation stack,
provided to their content tree via context, so a drilldown SubMenu
nested inside another drilldown SubMenu still drills into the same
stack and arbitrarily deep trees stay navigable one screen at a
time. Switching levels remounts the level's content wrapped in a
keyed Fragment rather than updating it in place, since
react-stately's Collection requires each item's id to stay stable
across an update and the back item's label (and everything under it)
genuinely changes identity between levels.
AI-assisted-by: claude-sonnet-5
* ♻️ Wire the DS Menu/SubMenu into the dashboard file menu
file_menu.cljs used context-menu-a11y's data-driven options list,
rendered via a generic recursive renderer. Rewritten as real JSX
composition (menu-item*/sub-menu*/menu-separator*) using the DS Menu
component, preserving every existing conditional branch (single-file,
multi-select, restore-mode, permission gates). "Move to" -> "Move to
other team" -> team -> project now uses sub-menu*'s drilldown variant
at every level.
Split into file-menu-items* (the item tree, no popover of its own)
and a thin file-menu* wrapper (Menu, anchored to the "..." button),
so grid.cljs can render the same items a second time inside a
ContextMenu for right-click, matching the previous behavior of
opening either via the button or a right-click anywhere on the row.
grid.cljs's trigger handling is simplified accordingly: DS's Menu/
ContextMenu handle their own positioning (including auto-flip near
viewport edges) and dismissal internally, so the manual click-
coordinate math, the dashboard-local :menu-open/:menu-pos globals,
and the portal-on-document* wrapper (Popover already portals itself)
are all gone. The now-fully-dead show-file-menu-with-position/
show-file-menu/hide-file-menu actions are removed from
data/dashboard.cljs.
Also fixes two issues found wiring this up:
- The add-shared/unpublish-shared toggle rendered two different
menu-item* ids at the same list position; :is-shared can flip while
the popover stays open (the action's own side effect), and
react-stately's Collection requires an item's id to stay stable
across such an update. Both branches now share one id.
- Menu's own trigger wrapper (align-self: start, needed generically
so it doesn't stretch in an arbitrary parent) overrode
.project-thumbnail-actions's centering of the "..." button;
grid.scss now re-asserts centering for that specific consumer.
Removes the temporary menu-test* harness from dashboard.cljs now that
there's a real integration to test against instead.
AI-assisted-by: claude-sonnet-5
* 🐛 Fix Menu/ContextMenu popover interaction bugs
Found testing the dashboard file menu integration:
- Reopening the same trigger right after closing (e.g. right-click,
dismiss, right-click again) could silently fail or briefly show two
overlapping instances. Closing played a 100ms exit fade, and a
reopen landing mid-fade raced the still-live Popover instance.
Closing now always skips the exit animation, so by the time any
subsequent open request arrives there's no ambiguous in-between
DOM state left to race.
- Right-clicking a different row while one file's context menu was
open didn't close the first one. Menu/ContextMenu don't use
react-aria-components' own MenuTrigger (Penpot's DS buttons aren't
react-aria-pressable), so they also don't get its built-in
RootMenuTriggerStateContext coordination between sibling instances.
A window CustomEvent broadcast restores it: opening announces this
instance's id, and every other mounted instance closes on hearing a
different one.
- With that coordination in place, right-clicking elsewhere still did
nothing at all: Popover defaults to modal, which marks the rest of
the app inert (unfocusable *and* unclickable, not just visually
blocked) while open. Correct for a real Dialog, wrong for a
lightweight dismissable menu. Fixed with isNonModal on all three
Popover usages (Menu, ContextMenu, SubMenu's flyout).
- isNonModal has its own side effect: react-aria only wires up its
click-outside-closes behavior when a popover is "dismissable", which
isNonModal forces off (for anything but a submenu flyout) with no
separate prop to turn back on. Reimplemented directly: a pointerdown
landing outside the popover's own rendered content closes it, via a
ref now passed to Popover.
AI-assisted-by: claude-sonnet-5
* 🐛 Fix Menu visual styling and two overflow bugs
Border and shadow, to match the legacy context-menu-a11y menu this
replaces: the DS component had neither (a filter: drop-shadow with a
different blur radius stood in for the shadow, and there was no
border at all). Used the pattern already established by sibling DS
dropdowns (options-dropdown.scss et al.) rather than porting the
legacy tokens directly — border: 1px solid
var(--color-background-quaternary) + box-shadow: 0 0 12px 0
var(--color-shadow-dark), both already in use elsewhere in this same
file.
Found two real bugs verifying that against a long "move to" list:
- .menuItem/.separator had no flex-shrink: 0, so once a list's
natural height exceeded the menu's max-block-size, flexbox shrank
every row to fit them all rather than triggering the scrollbar —
overflow only kicks in after flex-shrink has done its best, and
shrinking was never opted out of.
- The menu's own fixed max-block-size: 300px ignored react-aria's
Popover, which sets its own max-height (inline, on our direct
parent) to whatever space is actually available between the trigger
and the viewport edge. In a small viewport that computed value can
be under 300px; since the parent has no overflow of its own, our
independent 300px cap just rendered straight past it and off the
edge of the window. max-block-size: inherit picks up the parent's
own computed value instead, at the cost of no longer capping how
tall the menu can get when there's plenty of room (verified: 348px
in a normal-height viewport, vs the old fixed 300px) — an
acceptable tradeoff against content becoming inaccessible.
AI-assisted-by: claude-sonnet-5
* 💄 Adjust Menu design-system component item states and spacing
Give menu items a distinct keyboard-focus ring (accent-primary outline
plus tertiary background) separate from the mouse hover/click state,
which keeps its existing quaternary background unchanged. Restyle
disabled items with a tertiary background and secondary text color,
shrink the submenu chevron to 12x12, and tighten the menu's vertical
padding to 4px.
* 📚 Document drilldown submenu and tighten Menu docs
Add the drilldown submenu story to the Menu docs page, show the
idiomatic controlled-state shape in the usage example (callbacks
bound in the let with mf/use-fn, explicit deref of the open state),
and trim the prose down to the information a consumer needs.
* 🐛 Fix Menu outside-click closing on its own trigger and submenus
useCloseOnOutsideClick restores the dismiss behavior isNonModal turns
off, but it tested containment against the popover element alone. That
missed two cases react-aria's own useOverlay accounts for.
A root Popover wraps its content in a display:contents div and portals
every SubmenuTrigger's nested popover into that same div, so a flyout
submenu is a sibling of the popover, not a descendant. Pressing an item
in one counted as an outside click: the whole tree unmounted on
pointerdown and the item's action never fired on pointerup. Test the
group container instead.
The trigger was likewise treated as outside, so closing on its
pointerdown let the click's own handler read the already-false open
state and reopen the menu — a trigger wired to a toggle could never
close it. Exclude it in Menu; ContextMenu keeps the old behavior, since
right-clicking elsewhere should reopen it against a new anchor.
* 🐛 Target the clicked file when it is not in the dashboard selection
The file menu adopted the whole selection whenever it was non-empty,
guarding only against it being empty. That left the case where the
selection holds files this row is not one of: toggle-file-select is a
no-op across projects, so shift-right-clicking a file in another
project leaves the previous project's selection intact and the menu
opened on the pointed-at file while offering rename, duplicate, move
and delete for a different one.
Adopt the selection only when it actually contains this file, which
covers the deferred-dispatch case the previous guard was written for
just as well.
* 🐛 Drop the file menu teams cache that outlived a logout
The cache was a module-global defonce atom, and logging out does not
reload the page — it resets the store and navigates. The next profile
to sign in on the same tab therefore opened its first file menu with
the previous account's team and project names listed under "Move to",
until the background fetch replaced them.
The cache only ever saved the brief absence of one submenu, which is
already guarded on having data and so does not shift any layout, so
remove it rather than scope it to a profile. Dispose the subscription
too: it wrote to component state after unmount.
* 🐛 Keep the Menu open when its own trigger takes focus
Excluding the trigger from the outside-click dismiss was not enough to
make a toggle trigger able to close the menu: usePopover passes
shouldCloseOnBlur unconditionally, and useOverlay acts on it regardless
of isNonModal, so focus moving to the trigger on its own pointerdown
closed the popover before the click ran. The click then read an open
state that was already false and reopened it.
shouldCloseOnInteractOutside is the one exception useOverlay consults
before closing on blur, so use it to exempt the trigger.
* 🔧 Add interaction tests for the Menu component
Cover the two dismissal regressions just fixed — closing the menu from
its own trigger, and a press inside a flyout submenu not being treated
as an outside click — plus drilldown navigation, the navigation stack
resetting between open/close cycles, and Escape and outside click.
Both regression tests fail against the code as it was before the fixes.
The story trigger now toggles instead of only ever opening, which is
what a real caller does (the dashboard's own is a swap!) and what makes
the reopen bug observable at all.
* ✨ Add max-width, density, and drilldown sizing to Menu/ContextMenu
Add a max-width prop (default 250px) to Menu, ContextMenu, and flyout
SubMenu, and an is-dense prop to Menu/ContextMenu that shrinks every
item — including nested flyout SubMenus, via a shared density context
— to a 28px row. Pin a drilldown SubMenu's popover to at least the
root level's own size, so navigating into a shorter or narrower list
doesn't shrink the menu mid-navigation.
Also truncate a plain MenuItem's text with an ellipsis instead of
letting it wrap and blow out the row height, matching the existing
SubMenu trigger label, and fix that label's own truncation: it was
missing min-inline-size: 0, without which a flex item can't shrink
below its content size and text-overflow: ellipsis never engages.
Exposed through the ClojureScript facade as :max-width/:is-dense,
documented with new example canvases, and covered by five new
Storybook interaction tests, each verified to fail without its
corresponding fix.
* ♻️ Wire the DS Menu/ContextMenu into the dashboard project menu
Replace the legacy context-menu-a11y-based project menu (grid, sidebar,
and per-project file view) with the DS Menu/ContextMenu components,
mirroring the earlier file menu migration. Drop the manual
:menu-open/:menu-pos position tracking in favor of the DS components'
own positioning, and split project-menu-items* out so both the "..."
trigger and right-click share the same options.
The hidden file input behind the "Import" option moves out of the
popover content and into whichever parent stays mounted regardless of
the menu's own open state: the DS popover really unmounts its content
on close (unlike context-menu-a11y, which only hid it), and selecting
"Import" closes the menu in the same tick a ref owned inside it would
already be gone.
Add an onOpenChange notification to ContextMenu (it stays uncontrolled,
this only reports state changes) so the project row's "..."/pin/add-file
actions can stay visible for as long as either menu is open, the same
way they already do on hover. Fix a related visibility bug this exposed:
closing a menu restores focus to its trigger regardless of whether the
open happened via mouse or keyboard, so :focus-within alone kept the
actions visible after closing with the pointer away — swapped for
:has(:focus-visible), which only matches real keyboard navigation.
* 🐛 Forward MenuItem's id to the DOM as data-testid
MenuItem's function signature never forwarded anything beyond its
explicitly-typed props to the underlying RACMenuItem, so a caller's
id — meant as a stable per-item identifier — only ever reached the DOM
as react-aria's own internal data-key, never as data-testid. This
silently broke dashboard.spec.js's "Multiple elements in context" test
after the file menu's migration to this component, since every existing
menu item id was already relied on as its test id.
id is already unique per item for selection/on-action, so deriving
data-testid from it directly means every item is reachable in a test
with no separate prop to remember to pass. SubMenu's own trigger row is
a MenuItem too, so this covers it for free.
* 🔧 Add Playwright coverage for the project options menu
Covers all four places the migrated project menu is reachable: the
dashboard grid's "..." button and title right-click, the sidebar's
right-click, and the per-project files page's "..." button. Checks
rename/duplicate/pin/move-to/delete render (and that the default
Drafts project correctly hides all of them), that rename opens the
inline editor, that delete opens the confirm modal, and that the
move-to submenu lists other teams.
Also drop an unused React import from context_menu.stories.jsx,
spotted in passing.
* ♻️ Add datatest id
* ♻️ Fix linter
* 🐛 Build @penpot/ui automatically after pnpm install
packages/ui/dist is gitignored (build output) and nothing in the
install pipeline built it, so a fresh checkout — CI included — never
had it. Any code importing "@penpot/ui/menu" (the frontend's own
cljs-runtime tests among them) failed at module resolution with
ERR_MODULE_NOT_FOUND rather than any real test failure.
Build it in postinstall, the same way plugins-runtime already does,
so it's always present after `pnpm install` without a separate manual
build step.
* 🔥 Remove flaky Menu dense/ellipsis Storybook tests
Test Dense Shrinks Items and Test Long Label Ellipses Instead Of
Wrapping asserted computed pixel styles that passed consistently
locally (including with a fresh packages/ui install) but failed in CI,
suggesting a CI-only timing/environment discrepancy in when the
computed style stabilizes. Dropping them rather than chasing a
non-reproducible flake.
* 💄 Open the file/project options menu right, top-aligned
Switch the dashboard file and project "..." options menus from
"bottom end" to "right top" placement, so they open beside the
trigger button instead of below it.
* 🐛 Stop drilldown SubMenu jumping to the opposite edge
A drilldown SubMenu swaps its parent Menu/ContextMenu popover's own
content in place, and react-aria re-runs its flip/collision placement
on every layout change. Drilling into a shorter level than the root
could shrink the popover enough that react-aria decided there was now
room on the other side, flipping it there — a visible jump even though
the popover never actually moved from the caller's point of view.
The previous fix padded every drilled-in level out to the root's own
min-inline-size/min-block-size so the popover never got small enough
to trigger a re-flip, but that meant a level naturally much shorter
than the root still rendered at the root's full height.
Replace it with shouldUpdatePosition={false} on the Popover for as
long as any level is drilled in. This freezes whichever edge react-aria
already resolved for the root, so a shorter level just shrinks from the
opposite edge instead of triggering a new placement decision, and a
taller level grows from that same opposite edge in the direction the
root already opened. shouldUpdatePosition goes back to true once the
stack returns to the root, so a fresh open still resolves normally.
* ♻️ Update menu placements and use buttons from DS
---------
Co-authored-by: Luis de Dios <luis.dedios@kaleidos.net>
The frontend build now emits the sha256 hashes of the inline scripts of every page it writes into resources/public, the image moves them out of the document root, and the entrypoint splices them into the default script-src. This removes one of the two reasons why enforcing mode was not usable.
The hashes are computed on the rendered output rather than on the mustache templates, since the digest covers the exact bytes served between the script tags. All four served pages contribute, not just index.html: challenge.html handles the redirect, render.html is loaded by the exporter in a headless browser, and rasterizer.html is initialised by the frontend itself, so leaving any of them out would have broken those paths under enforcing mode. The storybook previews are excluded because that container does not serve them.
A bundle predating this change yields no hashes and the policy stays as it was, so older bundles keep building.
The three external locations were also passing through the security headers of their upstreams. raw.githubusercontent.com returns its own Content-Security-Policy and both it and fonts.googleapis.com return Strict-Transport-Security. Browsers enforce the intersection of every policy they receive, so the upstream one takes precedence on those responses, and the HSTS one lands on our own host, meaning a deployment that deliberately disables HSTS would get it set anyway by a third party. Hide all three at the proxy.
Signed-off-by: David Barragán Merino <david.barragan@kaleidos.net>
The bundle and docker-image build/dedup checks used different cache
keys: the bundle was cached by ref name (`penpot-<gh_ref>.zip`) while
the docker image marker was cached by commit sha
(`markers/images-sha-<sha>`). A tag built from a commit already
promoted under another ref (e.g. `develop`) would rebuild the bundle
unnecessarily, while `build-docker`'s `promote` job silently inherited
the skip from `build` and never created that ref's branch tags
(`backend:<gh_ref>`, `frontend:<gh_ref>`, ...), even though the
underlying sha-tagged images already existed.
- Key the bundle S3 object by commit sha (`penpot-sha-<sha>.zip`)
instead of by ref name, matching the docker marker's semantics.
- Drop the S3 metadata round-trip for `bundle_version` in
build-docker.yml; compute it locally with `git describe`, same as
build-bundle.yml (requires fetch-depth: 0 on that checkout).
- Split `promote` into two mutually-exclusive jobs, `promote` (needs
`build` to succeed) and `retag` (needs only `prepare`, runs when
`prepare.outputs.exists == 'true'`), each moving the `:<gh_ref>`
branch tags to the current sha. This replaces relying on `build`'s
skip/success state with two explicit conditions, so the tags always
get moved regardless of which path built the images.
Signed-off-by: David Barragán Merino <david.barragan@kaleidos.net>
The bundle and docker-image build/dedup checks used different cache
keys: the bundle was cached by ref name (`penpot-<gh_ref>.zip`) while
the docker image marker was cached by commit sha
(`markers/images-sha-<sha>`). A tag built from a commit already
promoted under another ref (e.g. `develop`) would rebuild the bundle
unnecessarily, while `build-docker`'s `promote` job silently inherited
the skip from `build` and never created that ref's branch tags
(`backend:<gh_ref>`, `frontend:<gh_ref>`, ...), even though the
underlying sha-tagged images already existed.
- Key the bundle S3 object by commit sha (`penpot-sha-<sha>.zip`)
instead of by ref name, matching the docker marker's semantics.
- Drop the S3 metadata round-trip for `bundle_version` in
build-docker.yml; compute it locally with `git describe`, same as
build-bundle.yml (requires fetch-depth: 0 on that checkout).
- Split `promote` into two mutually-exclusive jobs, `promote` (needs
`build` to succeed) and `retag` (needs only `prepare`, runs when
`prepare.outputs.exists == 'true'`), each moving the `:<gh_ref>`
branch tags to the current sha. This replaces relying on `build`'s
skip/success state with two explicit conditions, so the tags always
get moved regardless of which path built the images.
Signed-off-by: David Barragán Merino <david.barragan@kaleidos.net>
Two entries in THANKYOU.md had the first part of the Weblate username
pasted in front of the URL (pablo.https://... and swapnil.https://...),
which makes GitHub render them as broken relative links. Point them at
the real profiles, https://hosted.weblate.org/user/pablo.alba and
https://hosted.weblate.org/user/swapnil.cx, which both exist.
Signed-off-by: Alma Faris <almazaf19@gmail.com>
* 🐛 Fix shadows on a masked group in the WASM renderer
A masked group renders in two passes: its content, then the mask shape
composited with DstIn so everything outside the mask silhouette is
erased. Both happen inside one save_layer, and the group's drop shadow
was composited into that same layer before the mask pass — so the mask
erased it. A drop shadow lives mostly outside the silhouette, so it
disappeared entirely.
Inner shadows never drew at all: render_fill_inner_shadows needs fill
geometry to paint into, and a group has none.
Both now ride on an image filter set on the masked-group layer, which
Skia evaluates after the mask is composited, so the shadow comes from
the real masked pixels rather than the group's own, empty geometry.
The effects compose in the order the SVG renderer uses for a group:
drop shadows, then the source, then inner shadows, with the layer blur
over all of it.
That layer is opened on a canvas carrying no transform, so the filter
is built in device units. Shadow::scale_to_device does that rather
than scale_content, because radius_to_sigma is affine: scaling the
radius applies its constant term once at device scale, while a filter
built in document space has the term scaled by the canvas matrix. The
two would blur differently by 0.5 · (scale - 1) sigma, visible as a
masked group's own shadow being narrower than the same shadow on its
parent. The masked-group layer blur had the same flaw.
Three paths are suppressed for masked groups so nothing is drawn twice:
the silhouette composite, the nested_shadows inheritance that would
reach text descendants, and the fill inner-shadow pass.
Every save and restore around that layer is keyed on the shape alone.
Enter and exit run on different walker passes, and a pan or zoom in
between changes fast mode, so deriving them from render state could
leave the canvas clip stack unbalanced.
Refs #11697
AI-assisted-by: claude-opus-5
* 🐛 Fix a container's drop shadow over a masked group in WASM
A container builds its drop shadow by drawing each descendant as a
black silhouette and blurring the result. The walk descends only
through children that can be flattened, and a masked group never can,
so it stopped there and asked the group to draw its own geometry. A
group has none, so nothing was drawn and the shadow layer stayed
blank: no shadow at all for a group, board or frame holding a masked
group. This is what the file attached to the issue reproduces.
render_drop_black_shadow now draws the masked silhouette for such a
group — content children flat black, DstIn the mask, and only then the
offset, blur and spread. Masking after the blur would trim the shadow
along the wrong edge.
The walk recurses, so it narrows the clip the way the main walker
does: content a clipping container hides must not widen the shadow.
The clip rule now lives in one place, shared with the walker, and a
test pins the two against each other. The shadow layer is sized to the
silhouette plus the shadow's reach rather than falling back to the
clip, so a wide blur is not cut at the tile edge.
Spread keeps the renderer's existing behaviour: the silhouette goes
through the same get_drop_shadow_filter every other shadow uses, so a
masked group gains no ordering of its own.
Closes#11697
AI-assisted-by: claude-opus-5
