* ♻️ Simplify storage GC delays and add skip-delay task params
The touched GC no longer applies an extra deletion-delay when marking
storage objects as deleted. By the time a storage object is touched, its
referencing domain row has already passed its own deletion delay, and the
reference scan is the only safety check needed. Touched objects are now
marked with deleted_at = now, so the deleted GC removes them on the next
run.
For the tempfile bucket, upload chunks now set touched-at in the future
(1h, aligned with the upload-session-gc TTL) instead of relying on a
special-case deletion delay.
Task handlers now read their task props:
- storage-gc-touched accepts :skip-delay to process all touched objects
immediately, bypassing the min-age threshold.
- objects-gc accepts :chunk-size and :skip-delay to process recently
deleted rows without waiting for the deletion delay.
This allows running the deletion cascade immediately from the REPL via
run-task! with the skip-delay option.
AI-assisted-by: deepseek-v4-flash
* ✨ Add storage object status lifecycle, verified dedup, and deletion retry tracking
Storage object lifecycle hardening:
- Add status column ('valid' | 'pending') as write-ahead marker for
object creation. put-object! inserts in 'pending' state, writes blob,
then promotes to 'valid'. Failed writes remove the pending row.
- Add :storage-pending-gc task to reclaim orphaned pending rows (e.g.
after crash between blob write and promotion).
- Verify blob existence on every dedup hit via exists-object? (fs stat /
s3 headObject). Missing blobs mark the row as deleted and create fresh
object.
- Add deletion_attempts column (migration 0154) to track physical blob
deletion attempts. Restructure gc_deleted to use chunked processing
with per-chunk transactions (short lock duration). Failed deletions
are deferred to tomorrow (deleted_at = NOW() + 1 day) to prevent
infinite loops. After 7 attempts, give up and accept orphan.
- Change del-objects-in-bulk contract to return #{fail-ids} for precise
per-id tracking (fs and s3 backends updated).
- Use tmp/tempfile for fs atomic writes with cleanup queue registration
(crashed-JVM temp files swept ~60min later). Document ATOMIC_MOVE
POSIX-only assumption.
- Add linear backoff to s3 exists-object? retries (100ms/200ms/300ms).
- Wrap compensating delete in put-object! catch block to prevent
masking original error when connection is aborted.
- Fix assert messages in pending_gc.clj and gc_deleted.clj (pool
assertion said 'expected valid storage' instead of 'db pool').
- Add pending-objects-excluded-from-gc-deleted test. Use unique path in
put-object-write-failure-leaves-no-row test to avoid collisions.
AI-assisted-by: qwen3.7-plus
* 🐛 Fix review comments on gc-deleted and storage
- Fix process-chunk! returning nil causing (+ acc nil) crash
- Add FOR UPDATE SKIP LOCKED to sql:get-deleted-chunk to prevent
infinite loop when another worker holds locks
- Pass :cause to log messages in gc_deleted.clj and s3.clj
- Fix extra space in log hint string
- Remove unused ::blob-missing? reference from storage memory
- Rename test to match actual behavior (leaves pending row)
- Add test for gc-deleted giving up after max attempts
AI-assisted-by: qwen3.7-plus
* ✨ Improve path operations and edition
* 🐛 Fix floating-point equality issues in path editing
Replace exact equality checks with tolerance-based comparisons
in path editing functions to handle floating-point rounding
differences after transforms, rotations, or curve fitting.
Changes:
- distribute-content: Round coordinates to 0.1 precision before
grouping to ensure coincident nodes move together
- separate-node: Use gpt/close? instead of exact equality to
find nodes with floating-point imprecision
- collision-step: Use mth/close? for tolerance-based comparison
to detect paste collisions correctly
- resolve-edit-fills: Add cycle detection to prevent infinite
loops with corrupted parent chains
Made collision-step, available-offset-step, and resolve-edit-fills
public for better testability.
Added comprehensive tests for all fixes covering both exact and
floating-point coordinate scenarios.
AI-assisted-by: qwen3.7-plus
* 🐛 Fix path editor code review findings
Fix issues identified during code review of path editor enhancements:
- Fix unused binding lint warning in distribute-content that blocked CI
- Fix collision-step floor comparison to use round instead of floor,
correctly detecting collisions when coordinates drift slightly below
integer boundaries
- Fix resolve-edit-fills to recurse through empty parent groups when
searching for inherited fills in nested group hierarchies
- Fix expand-coincident-node-indices to use fuzzy comparison (gpt/close?)
instead of exact equality, handling floating-point divergence after
transforms or rotations
- Remove unreachable dead code in path-point* on-pointer-down handler
- Add tests for collision-step boundary cases, nested group fill
inheritance, and coincident node alignment/flipping
AI-assisted-by: mimo-v2.5-pro
---------
Co-authored-by: Andrey Antukh <niwi@niwi.nz>
* ✨ Auto-link libraries during import based on slugified name
When a Penpot file is exported without bundled libraries and then
imported into a different environment, external library links are
broken because library UUIDs differ across environments.
This feature adds a heuristic to auto-relink libraries by matching
slugified library names against shared files in the target team:
- Export: embed external library metadata (id, name, slug, used-by)
in the manifest when libraries are not included in the export.
- Import: resolve external libraries by slugifying shared file names
in the destination team and matching against manifest slugs.
- Single match: auto-link silently (creates file-library-rel row).
- Multiple matches: emit SSE event so the frontend shows a selection
dialog for the user to pick the correct library.
- No match: import continues without linking (current behavior).
Backend changes:
- Extended manifest schema with optional :external-libraries field
- Added slugify-name, get-files-names, get-shared-files-for-team,
find-shared-files-by-slug helpers in app.binfile.common
- Threaded team-id into import cfg from RPC layer
- Added resolve-external-libraries and auto-link-libraries in v3
- Emit :library-candidates SSE event for multi-match cases
Frontend changes:
- Worker captures library-candidates SSE events and forwards them
- Import dialog shows auto-link notification and multi-match
selection UI with select dropdowns
- Added link-files-to-library! RPC helper for user selections
- Added en/es translations for new UI strings
Closes#9263
Signed-off-by: Andrey Antukh <niwi@niwi.nz>
* ✨ Add UI for the auto-link plumbing
* ♻️ Consolidate auto-link libraries with unified export-type and fix ref integrity
Consolidates the auto-link libraries feature into a single coherent implementation:
- Unify boolean flags (embed-assets, include-libraries, link-later)
into single ::bfc/export-type parameter
- Fix critical reference-integrity bug: pre-resolution no longer remaps
:component-file refs when no link is created (multi-match / no-permission)
- compute-link-decisions as single source of truth for auto-link logic
- 80+ backend tests covering round-trip, cross-team, permissions,
edge cases, and reference integrity
AI-assisted-by: longcat-2.0
---------
Signed-off-by: Andrey Antukh <niwi@niwi.nz>
Co-authored-by: Eva Marco <evamarcod@gmail.com>
* 🐛 Fix svg-raw child ids stored as strings in existing files
Before schema:svg-raw-attrs declared :shapes as a vector of uuid, the
JSON decoder had no type information for those child ids and left them
as plain strings on any round trip, so they got persisted as strings.
Once the schema was tightened those files fail schema validation on
load.
Add migration 0026-fix-svg-raw-shapes-uuids that coerces the :shapes
entries of svg-raw shapes back to uuid instances, in both pages and
components.
AI-assisted-by: deepseek-v4-flash
* ♻️ Drop redundant vector check in svg-raw shapes migration
:shapes is always a vector when present, so the guard is unnecessary.
AI-assisted-by: deepseek-v4-flash
Editor/WASM round-trips can truncate line-height strings
(e.g. 1.3333333333333333 → 1.33333). Exact string compare
treated that as a style change and detached typography tokens.
Extract the backend Testing section from backend/core into a dedicated
backend/testing memory, following the pattern of common, frontend, and
exporter. Update the testing skill and root testing memory to point at
the new location, and add exporter/testing to the skill's required
reading list.
AI-assisted-by: deepseek-v4-flash
The backend format check was using 'check-fmt' instead of 'check-fmt:clj',
causing CI to always fail on the fmt step.
Closes#11358
AI-assisted-by: longcat-2.0
The handler previously allowed any team member (including viewers) to
generate invitation tokens. Now requires at least edition-level
permissions (can-edit, admin, or owner).
Closes#11358
AI-assisted-by: longcat-2.0
* 🐛 Enforce ownership check on tempfile bucket access
The upload-tempfile RPC stores profile-id with tempfile objects, but
objects-handler never verified the requester was the owner. Any
authenticated user who knew the UUID could access the tempfile.
Add ownership check: tempfile bucket now requires the request's
profile-id to match the stored profile-id. Returns 404 on mismatch
(not 403) to avoid leaking object existence.
Legacy tempfiles without stored profile-id remain accessible to any
authenticated user for backward compatibility.
Closes#11269
AI-assisted-by: qwen3.7-plus
* ♻️ Extract tempfile-bucket constant and fix docstring indentation
Extract the 'tempfile' bucket string literal into a named constant
(sto/tempfile-bucket) to prevent typos and make future bucket renames
trivial. Updated 9 occurrences across 7 files.
Also fixed minor docstring indentation inconsistency in authenticated?
function.
AI-assisted-by: qwen3.7-plus
* ♻️ Refactor process-bucket! and authenticated? helpers
Replace case with cond in process-bucket! to properly resolve
sto/tempfile-bucket var from another namespace (case does not
evaluate qualified vars at compile time).
Redefine authenticated? in terms of request-profile-id to remove
duplicated lookup logic.
Closes#11269
AI-assisted-by: mimo-v2.5-pro
* 🐛 Add accumulated storage byte quota for media uploads
Add media-storage-bytes-per-team quote to prevent persistent DoS via
repeated uploads. The quota sums storage_object sizes from both
file_media_object (media + thumbnails) and team_font_variant
(otf/ttf/woff1/woff2). Default limit is 20 GiB per team, configurable
via PENPOT_QUOTES_MEDIA_STORAGE_BYTES_PER_TEAM.
The check is invoked in upload-file-media-object before processing,
looking up the team-id via file -> project -> team_id join.
AI-assisted-by: mimo-v2.5-pro
* 🐛 Fix deduplicated storage overcounting in media-storage-bytes-per-team quote
The SQL query sql:get-media-storage-bytes-per-team used UNION ALL
across six SELECT branches that each produce a so_id reference.
When deduplication causes multiple file_media_object or
team_font_variant rows to point at the same storage_object, UNION ALL
counts that objects size once per reference — inflating "used bytes"
and causing false :max-quote-reached rejections.
Change all five UNION ALL to UNION so that duplicate so_id values are
collapsed before the JOIN storage_object / SUM(so.size).
Add a test (media-storage-bytes-quote-deduped) that creates one
storage_object referenced by two file_media_object rows and asserts
the computed usage reflects the deduplicated physical size, not 2x.
AI-assisted-by: mimo-v2.5-pro
those textures across zoom for progressive previews, and invalidate
by old∪new document coverage so rotate/move edits do not leave
stale fragments on zoom-out.
* 🐛 Fix font selector dropdown takes noticeably long to open when changing font
* ♻️ Refactor detach-preview-sprite! to use atomic swap
Use idiomatic atomic swap! update instead of non-atomic
read-then-write pattern. The new implementation computes
the decremented refs inside swap! and only removes the
node when the result reaches zero.
AI-assisted-by: mimo-v2.5-pro
---------
Co-authored-by: Andrey Antukh <niwi@niwi.nz>
Ensure toast components always receive a boolean `is-html` prop so nil or
truthy notification values do not violate the Rumext schema.
AI-assisted-by: gpt-5.6-luna
The viewer's login-register modal always showed the Register link
regardless of the disable-registration server flag. The main login
page already gates this link on (contains? cf/flags :registration);
apply the same guard to the viewer modal to make the flag consistent
across all entry points.
PR #11036 added a per-request permission check to the file-media asset
endpoints (/assets/by-file-media-id/:id and the /thumbnail variant) using
bfc/get-file-permissions. Anonymous share-link viewers were then rejected
because they have neither a session nor an access token, the asset URL
carries no share context, and the 2-arg get-file-permissions short-circuits
to nil when profile-id is nil.
Make the asset endpoints share-link aware, mirroring how get-view-only-bundle
already authorizes the same scenario: read the share-id from the query
string, validate it as a UUID, and call the 3-arg perms/get-file-read-permissions
(which chains the existing 2-arg bfc lookup, the bfc share-link fallback,
and the Nitrate org-owner fallback). On the frontend, extend
cf/resolve-file-media with an optional share-id arg and pass it from the
WASM viewer render path using the share-id already present in
[:viewer-local :share-id]. Non-viewer call sites (workspace, clipboard,
code-gen) keep the original URL shape because the new arg defaults to nil.
Closes#11338
AI-assisted-by: minimax-m3
Present visible tiles via ViewportReady so zoom settle turns sharp
without waiting on the interest ring, and paint at atlas slot size so
DPR 2 does not rasterize 1024 only to downscale into 512 slots.
The get-page RPC command did not validate that the requested page-id was within the share-link's authorized :pages set, allowing share-link holders to read out-of-scope pages. The get-file-fragment command had the same issue.
This fix adds page scope validation to get-page, rejecting requests for pages not in the share-link's :pages set with a :not-found error. For get-file-fragment, share-link access is denied entirely as fragments lack direct page-id mapping.
The fix aligns these commands with the existing correct behavior in get-view-only-bundle, which already filters pages based on share-link scope.
Closes#11281
AI-assisted-by: qwen3.7-plus
Persist binfile manifest metadata in file_data on import so file
statistics are available at open-workspace time. Emit a new
open-workspace-file audit event enriched with file statistics:
page count, shape count, component count, linked libraries,
design tokens, and whether the file is a shared library.
Closes#11106
AI-assisted-by: mimo-v2.5-pro
Pan/zoom via render_from_cache while the tile atlas is still empty
left a blank workspace under the page-transition blur. Ignore
set-view-box / view-interaction-start until tiles-complete, block
pointer events on the viewport SVG, and flush any deferred local
viewport sync when the overlay ends.
Rename skill directory and update all references to follow the
same naming pattern as plan-review.
Simplify review.md command from 138 to 25 lines — remove redundant
content that duplicated what the skills already define. The command
now acts as a thin router; the skills own the methodology.
AI-assisted-by: mimo-v2.5-pro
Logout only cleared the auth-token cookie but never deleted the
server-side row because delete-fn read ::id which wrap-authz no longer
sets since 363b4e3778. Make delete-fn delete via ::session/:id attached
by wrap-authz so replayed tokens are rejected (CWE-613, GHSA-mj9f-5cwq-7p3q).
Add regression tests covering invalidation, idempotency and isolation
of other sessions. Fix verified with Red→Green TDD and full backend
suite (677 tests).
Closes#11316
AI-assisted-by: muse-spark-1.2-contributor
* 🚑 Remove internal error details from HTTP error responses
PostgreSQL exceptions, I/O exceptions, and unhandled errors were
leaking raw database messages (table names, constraint names,
SQLSTATE codes), filesystem paths, and internal exception details
to API clients via :hint, :state, and :path response fields.
Remove these fields from server-error responses while keeping
full error context in server-side logs for operators.
Closes#11287
AI-assisted-by: mimo-v2.5-pro
* 🚑 Strip internal fields and map PG errors to safe messages
Complete the security fix for GHSA-r8wx-23q6-w3gf by addressing
the incomplete redaction found in code review.
Add strip-internal-fields helper to dissoc :hint, :state, :path,
and :context from error response data in three handlers that
previously passed raw ex-data through to clients:
- handle-error :internal
- handle-exception :default (else branch)
- handle-error :assertion (else branch)
Add pgsql-state->message to map PostgreSQL SQLSTATE codes to safe,
client-facing messages (e.g. 23505 → "A conflicting entry already
exists") instead of returning raw PG error text. Include :message
in all PSQLException response branches.
Add regression tests asserting :hint, :state, :path, :context are
absent from responses for :internal and unhandled ex-info errors.
Closes#11287
AI-assisted-by: mimo-v2.5-pro
* 🚑 Keep :hint in error protocol, fix unsafe sources
Refine the security fix based on code review feedback.
Keep :hint as part of the error protocol — it is essential for
controlled error communication. Remove it from strip-internal-fields
(which now only strips :state, :path, :context).
Fix the actual sources of unsafe :hint values:
- http/middleware.clj: replace (ex-message cause) with safe static
strings for IllegalArgumentException, RequestTooBigException, and
EOFException. These :validation errors return ex-data verbatim
to clients, so raw exception messages were leaking internals.
- PSQLException handler: use :hint instead of :message for the
SQLSTATE-mapped messages, staying consistent with the error
protocol.
Update tests to assert :hint is present (with safe static values)
in :internal and unhandled ex-info responses, and absent only from
bare RuntimeException and IOException responses.
Closes#11287
AI-assisted-by: mimo-v2.5-pro
