logicalWidth/viewportHeight can go non-positive for a narrow container
on a heavily zoomed-out board, where outerCardInset (which grows as
canvasScale shrinks) can exceed the container's own un-inset size.
That non-positive scale then feeds into a division (Infinity/NaN into
--board-canvas-ui-scale) and is used as zoom directly (invalid or
collapsed content) - flooring it well above zero avoids both.
Removing it in favor of viewportHeight for the vertical fit ratio left
the declaration itself dead code, which the CI lint step flags as an
error rather than a warning.
- Guard canvasScale before dividing by it in outerCardInset, matching
calculateBoardUiScale's own fallback.
- Fix containerContentScale's vertical fit ratio for scrollable
containers: it was derived from the full (all-rows) content height,
which drifts toward 1 as more rows are added regardless of the fixed
inset, under-scaling relative to what the visible viewport actually
needs and clipping the bottom of the visible rows. Use
viewportHeight/viewportRowCount instead, which describe the real
visible card size in both the scrollable and non-scrollable case.
A container's own visible card is inset from its allocated board cell
by the board's standard per-item gap, but items inside it are
positioned in a coordinate space sized to the un-inset column/row
count - so without correcting for that, items (not just the grid's
own box) spilled past the container's right/bottom edge by that
inset amount, worse the busier/more zoomed-out the board is.
Scales the container's inner grid content down to fit the actual card
size, using the same zoom mechanism the board's own top-level canvas
uses for its whole-board zoom, just one level deeper. Corrects
--board-canvas-ui-scale to compensate so icon/text/custom-CSS sizing
inside the container isn't affected - computed as a plain JS number
rather than a CSS calc() referencing the property's own inherited
value, since custom properties don't have a sequential/temporal order
within the same element and doing so is a circular reference that
silently invalidates the whole property for every descendant.
Also fixes the collapsible container's label bar poking above the
card's rounded top border via a negative offset - it now sits flush
inside it.
Review follow-ups on the session query cache:
- Cache the per-payload size verdict and reuse a single TextEncoder. The
persist subscription re-dehydrates the whole cache on every cache event,
so every widget payload was being superjson-serialized several times per
refresh cycle just to measure it.
- Skip the size measurement on the restore path. It runs before any query is
allowed to fetch and the payload was already budgeted when it was written.
- Reject stored queries whose dataUpdatedAt is not a positive finite number;
`undefined <= 0` is false, so those were hydrating with a NaN timestamp.
- Add `stop()` and tear the persister down before `queryClient.clear()`, so
the clear cannot overwrite the saved snapshot with an empty cache.
- Remove every other scope's cache when a persister is created: signing out
navigates away before the scope guard can react, which left the previous
account's dashboard data in the tab.
- Remove the entry instead of storing an empty cache.
Refresh indicator:
- Make it decorative instead of an aria-live region; every widget refreshes
on a timer, so it made assistive technology repeat "loading" forever.
Also broadens the e2e hydration-error matcher to production React's minified
error codes, hoists the single WidgetTimeProvider in ClientBoard, memoizes the
resolved local time zone, and drops the exports this branch made dead.
Rebuild the integration marquee on Mantine's Marquee, SimpleGrid and
ThemeIcon instead of the hand-rolled grid, track and keyframes, and
reduce the stylesheet to the mask and reduced-motion rule.
Drop the useMounted guards from the login form and let the buttons rely
on the pending state alone, and use Group for the restore flow's file
summary.
Replace the stacked install modals (browse -> details -> import review)
with page-based flows so the primary action is always reachable:
- /manage/custom-widgets/workshop browses submissions as a card grid
with large previews, sitting in a tab next to the installed widgets
- /manage/custom-widgets/workshop/[id] merges the listing, security
review and source setup onto one page with a sticky install footer
- /manage/custom-widgets/publish/[id] replaces the publish modal
- /manage/workshop redirects; its navbar entry is gone
- the Custom CSS picker collapses from three modals to one
Extract useCustomWidgetImport and ImportReviewContent so the Workshop
page and the file/clipboard dialog share one import pipeline.
Votes now show the viewer's own choice filled Reddit-style, backed by a
new user-votes query, and screenshots get a zoomable lightbox.
Rework the installed list onto ManageCollectionItem: show the widget
description, replace the cryptic "JSX - 2" badge with plain request and
origin facts, and swap the toggle and bare pencil for a status badge,
a labelled Edit button and a menu entry for enable/disable.
Fixes found while reviewing:
- importing could run twice and duplicate a widget; guard on pending and
succeeded, and return to the installed list instead of the heavy edit
form
- a background refetch of the submission (window focus, or the
invalidation a vote triggers) reset the source setup and silently wiped
part-typed credentials; key the reset on the sources, not the object
- breadcrumbs raised MISSING_MESSAGE for unmapped dynamic segments; fall
back to the raw segment via t.has
- publishing had no in-flight guard and could post duplicate listings
- tooltips on disabled vote and report controls never fired, making the
sign-in hint unreachable
Thomas
Bergasha
ajnart