MCP tokens now use a separate JWT issuer claim (`urn:penpot:mcp-token`) instead of `access-token`, preventing them from being validated as API access tokens.
Fixes#10960
AI-assisted-by: qwen3.7-plus
Production crash where @(get bounds id) threw
"No protocol method IDeref.-deref defined for type null"
when a shape ID had no corresponding entry in the bounds map
during layout calculations.
Added defensive nil guards (when-let / when) to all unprotected
bounds dereference sites:
- flex_layout/bounds.cljc: layout-content-points (parent + child)
and layout-content-bounds
- grid_layout/bounds.cljc: layout-content-points and
layout-content-bounds
- min_size_layout.cljc: child-min-width grid branch (3 sites) and
child-min-height grid branch
Added 7 new tests in geom_bounds_layout_nil_test.cljc covering all
nil-bounds edge cases for flex, grid, and min-size layout paths.
Registered in runner.cljc.
Closes#10843
AI-assisted-by: qwen3.7-plus
* 🐛 Fix main menu is covered by the toolbar
* ♻️ Refactor SCSS
* 🐛 Fix adjust z-index of workspace context menu
* ♻️ Refactor SCSS
* 🐛 Fix adjust z-index of tokens context menu
* ♻️ Refactor SCSS
* 🐛 Fix adjust z-index of old context menu
* 📎 PR improvements
* 📎 Update serena documentation about creating-prs workflow
* 🐛 Handle unrecognized JSON escape sequences as malformed-json
When clojure.data.json's read-escaped-char encounters an unrecognized
escape sequence (e.g. a backslash followed by '}', or other case
fall-throughs in the parser) in a JSON request body, it throws a bare
IllegalArgumentException. Previously this fell through to the generic
RuntimeException branch in wrap-parse-request's handle-error, which
unwrapped and recurred without matching, eventually reaching the
internal-error handler and producing HTTP 500 + an error report — even
though the root cause was malformed client input, not a server bug.
The fix converts any IllegalArgumentException raised in the JSON parse
path into a `:validation`/`:malformed-json` error by raising a new
ex-info (which is caught by the top-level error handler in
`app.http/router-handler`). The result is an HTTP 400 response with a
descriptive hint, and no error report is generated. This addresses
~10% of all error reports received.
The new IAE branch is placed before the RuntimeException branch in
the cond (since IllegalArgumentException IS-A RuntimeException) and
uses the throw-style (ex/raise) to match the existing
RequestTooBigException / EOFException branches. A comment above the
handle-error cond documents why raising is intentional and is caught
by the top-level app.http error handler, not by the per-route
wrap-errors middleware.
Test suite changes:
- Extend the existing `DummyRequest` defrecord in
`http_middleware_test.clj` from 2 fields to 12 fields, implementing
every IRequest method, and add a private `make-dummy-request`
constructor that accepts an options map with every key optional and
sensible `:or` defaults. Future fields added to DummyRequest won't
break existing call sites as long as the `:or` defaults are kept in
sync.
- Remove the now-redundant `JsonRequest` defrecord and migrate all 11
`->DummyRequest` call sites to `make-dummy-request`.
- Add 6 new deftest cases:
- parse-request-illegal-argument-exception: malformed JSON body
(containing `\}`) is converted to `:malformed-json`.
- parse-request-request-too-big-exception: RequestTooBigException
is converted to `:request-body-too-large`.
- parse-request-eof-exception: java.io.EOFException is converted
to `:malformed-json`.
- parse-request-runtime-exception-with-cause: a wrapped
RuntimeException recurses on ex-cause and dispatches to the
matching specific branch.
- parse-request-runtime-exception-without-cause: a bare
RuntimeException falls through to errors/handle, returning 500
with :type :server-error :code :unexpected.
- parse-request-non-runtime-throwable: java.io.IOException (a
non-RuntimeException Throwable) is handled by the dedicated
handle-exception method, returning 500 with :code :io-exception.
Together, the new tests cover all 6 branches of wrap-parse-request's
handle-error cond.
Refs #10804.
AI-assisted-by: minimax-m3
When the user types a search term that filters out the currently-active
font, the picker showed no selection at all and Enter would close without
applying any font. Fix:
- Compute effective-selected (render-only, no state mutation) as the
first filtered font when the current font is absent from the results.
The row renderer and recent-fonts list use this for the tick mark, so
the top match is visually pre-selected while the user types.
- Enter key applies first-result (via on-select then on-close) when the
current selection is not in the filtered list; otherwise closes as
before. No font is live-applied on every keystroke.
- on-key-down deps extended to on-select/on-close; effect deps include
on-key-down so the global listener is always current.
Avoids the live-apply side-effect that caused the revert of #9512.
Fixes#3204.
Signed-off-by: Filip Sajdak <filip.sajdak@siili.com>
Co-authored-by: Andrey Antukh <niwi@niwi.nz>
* 🎉 Improve reset to default flow
* 🎉 Add diff modal
* 🐛 Fix measurement shortcuts
* 🎉 Separate open-sections for each tab
* 🐛 Align button link icon
* 🎉 Reset only works after press save
* 🐛 Fix little things
* 🎉 Make the import export row to be fixed
* ♻️ Fix CI
* 🐛 Cancel shortcut is not appearing on disabled tab
* 🐛 Fix tests
* 🐛 Fix loop on personalized
* 🐛 Fix show measurements shortcut
The timeout for tool calls (which is trictly relevant for plugin tasks only)
is now configurable via env. var PENPOT_MCP_TOOL_TIMEOUT_S.
The default was raised from 30 to 120, because 30 seconds was not enough for
some calls, especially in larger Penpot files. #10953
In multi-user mode, rejecting a second plugin WebSocket connection for an
already-registered user token performed the full removeConnection cleanup
for the newcomer. Since the token-keyed cleanup is keyed by token rather
than by socket, this deleted the clientsByToken entry and the Redis
request-channel subscription of the established, healthy connection. That
connection then remained open and heartbeating but was unroutable, so every
subsequent MCP tool call for the user failed although a valid plugin
connection existed.
removeConnection now performs the token-keyed cleanup only if the removed
connection actually owns the token registration, so rejecting a duplicate
releases only the resources the newcomer itself registered.
AI-assisted-by: claude-fable-5
In multi-user mode, plugin task requests are published to a Redis channel
keyed by user token. When no MCP server instance held a plugin connection
for that token (e.g. after the user navigated away from the workspace),
the publish reached zero subscribers and the request was silently dropped,
so every tool call stalled until the 30-second task timeout instead of
failing with a meaningful error.
RedisBridge.sendTaskRequest now returns the PUBLISH receiver count and
releases its response-channel subscription when the request reached no
receiver (or publishing failed), since no response can arrive. PluginBridge
uses the count to reject the pending task immediately with the multi-user
connection error message; publish failures likewise reject the task instead
of surfacing as an unhandled rejection followed by a timeout. The pending-
task settlement logic shared with the timeout handler is extracted into a
rejectPendingTask helper.
AI-assisted-by: claude-fable-5
