Repository navigation
Releases
v13.18
In short
Admin Panel / Problems says what its numbers count - "New
problems: 1" and "Checks with findings: 8 / 31", in every language - and a
report row's address, location and user always belong to the same occurrence.
Thanks to above GitHub users for their contributions.
v13.17
In short
The Admin Panel answers one question per tab: Settings gains
Appearance, Features, Feature profiles, Security, Activities and
notifications, and Delete, Problems keeps only what is going wrong, and every
old
address redirects. The docs now have a page for every menu entry, and cards of
an archived swimlane no longer appear in the first swimlane.
Thanks to above GitHub users for their contributions.
v13.16
In short
ListLockBulkBleed is fixed, and a list that refuses a role
now also refuses that role's bulk writes, dependency lines, rule actions and
changes to the list or swimlane itself. Administrators can give chosen users,
teams and organizations a smaller WeKan with opt-in feature profiles. A
server no longer runs out of memory in the hourly History integrity audit,
scheduled rules can create cards again, and Card Excel export and
clearing a card's text work again.
Security
Bulk REST card writes and server-side card writes ask the list's role exception. Thanks to immanuwell and xet7.
GHSA-83w9-74mx-7v83,
CWE-863. A board admin can refuse a role write access in one list (#4527).
DELETE /api/boards/:boardId/cards/bulk and
POST /api/boards/:boardId/cards/labels take their cards from the body, so no
URL names a list, and asked only for board write access; the methods
archiveSelectedCards, copyCard and createLinkedCard write from server
code, past the Cards allow/deny rules. A normal member the list refused could
permanently delete its cards in one request, relabel and archive them, and copy
or link new cards into it.
The bulk routes now read every card and ask each list before the first write:
any refusal is a 403 and nothing changes, and each removal and relabel is keyed
on the list that was checked. The refusal shows in Admin Panel / Problems as
ListLockBulkBleed (authz.list-locked-bulk, medium, never disables). The
methods answer not-authorized and are not recorded, because Copy Card and
Multi-Selection reach them in ordinary use. The same shape was fixed in twelve
more paths: converting a card to a board, adding a subtask,
createCardWithDueDate, the iCalendar import, sending and receiving board
email, creating a version-control issue, moving board objects, History and
position-history undo and redo, renaming an attachment, the scoped import and a
card's rule button. Dependency lines keep their own board-wide rule.
tests/listLockBulkBleed.test.cjs runs the gate on the reporter's board and
pins every path; tests/playwright/specs/list-lock-bulk.e2e.js replays the
proof of concept.
Thanks to above GitHub users for their contributions.
v13.15
In short
RaceBleed is fixed: deleting a card while somebody moves it
to another board no longer destroys its checklists, comments and attachments
while the card itself survives there, and a member deleting a card that was
just moved keeps their account. REST handlers now answer 404, not 500, for
a swimlane, user or board that is not there.
Security
A card's children are destroyed only for a card whose own removal happened. Thanks to xingmuse and xet7.
GHSA-3vjg-3p4q-8fwh,
CWE-367. The REST card delete looked the card up on the board the caller was
authorised for, destroyed its checklists, checklist items, comments,
attachments and subcards by card id, and removed the card itself last, keyed
on that board and list. A move to another board landing in between left the
card alive on a board the deleter cannot read, stripped of every child, with
both requests answering 200. Every hooked removal had the same shape:
collection-hooks runs Cards.before.remove, where the cascade was, and then
removes by the caller's original selector - a swimlane, a delete batch, a
parent - which a moved card no longer matches. The custom-field link clean-up
did the same to the cards linked to it.
The card document is now the claim (models/lib/claimedCardRemoval.js): the
single and bulk REST deletes remove the card first and destroy its children
only when that removal matched, and the hooked removals cascade from
Cards.after.remove, only for a card that is really gone. A list deletion
removes a card only while it is still in that list, and the REST board move
that loses to a delete is a 404 instead of a 500.
A delete that lost the race, or arrived after the card was moved away from
the named board, is refused and shows in Admin Panel / Problems as RaceBleed
(authz.card-delete-race, medium). It never disables the account: a stale
view is not an attack, and PurgeBleed's critical record used to disable a
member who deleted a card somebody had just moved.
tests/claimedCardRemoval.test.cjs replays the reported race and scans
server/ for any cascade before or without the card's removal;
tests/playwright/specs/race-bleed.e2e.js races twelve moves against
deletes and checks that a surviving card always keeps its checklist.
Thanks to above GitHub users for their contributions.
v13.14
In short
a checklist folded on a minicard stays folded, and every
profile preference survives a reconnect. The card shortcuts open only their
popup, one shared list header row really hides the other swimlanes'
headers, and the editors no longer show "comment-placeholder" after the
i18next 27 update. Deleting a card keeps its attachments recoverable.
Thanks to above GitHub users for their contributions and translators for their translations.
v13.13
In short
Copy Checklist and the other destination pickers list the
cards of a large board again: a board that loads its cards lazily used to leave
the card dropdown empty. Dependabot's updates, among them i18next 27 and
the mongodb driver 7.8.0, are kept.
Thanks to above GitHub users for their contributions.
v13.12
In short
WeKan is updated to Meteor 3.6.1, whose DDP server refuses
malformed messages without stalling the connection. One shared list header
row now shows the same lists in every swimlane, scrolling together, and
Shift+M and Shift+C open Move Card and Copy Card. Those popups
remember the swimlane and list, not only the board. The embedded-engine
benchmark takes OpenTelemetry 1.47.0 for a Dependabot alert. A REST
API write that changes nothing answers 404, and a card deletion naming the
wrong list no longer strips the card's checklists (ListLockBleed).
Security
A card deletion naming another list no longer strips the card's checklists and comments. Thanks to xet7.
A follow-up to ListLockBleed,
GHSA-4734-8vrg-8j3w,
found while checking the REST deletions. DELETE /api/boards/:boardId/lists/:listId/cards/:cardId found the card by board
alone, removed its checklists, checklist items, comments and attachments, and
only then removed the card by board AND list. With the wrong list id the card
stayed, emptied, and the reply was 200. Only the list in the URL was asked
whether the caller may write, so a member a list refuses could strip the cards
of that list by naming one they may write. The card now has to be in the named
list, its own list is asked as well, and the refusal shows in Admin Panel /
Problems as authz.list-locked. tests/restWriteRoutes404.test.cjs runs the
handler on both cases and requires the list check before anything is removed,
and tests/playwright/specs/rest-write-routes-404.e2e.js checks that the
checklist survives against a running server.
Thanks to above GitHub users for their contributions.
v13.11
In short
The file-integrity scan stops reporting attachments and
avatars that WeKan deleted itself as missing, which had buried real
StorageBleed findings under thousands of false ones. The recurring-checklist
reset no longer scans every checklist on each tick. In the REST API, a
refused custom-field update changes nothing, ownership transfer takes only the
boards the user administers, and deleting a missing integration answers 404.
Thanks to above GitHub users for their contributions.
v13.10
In short
The release's Snap Store credential check bounds each
snapcraft install attempt, so a Snap Store that hangs no longer stops a release.
Thanks to above GitHub users for their contributions.
v13.09
In short
The release's Snap Store credential check retries installing
snapcraft, so one Snap Store timeout no longer stops a release.
Thanks to above GitHub users for their contributions.