No node outside a root's structural subtree may intrude into that bloodline's
[minX, maxX] horizontal span — the contiguity guarantee that fixes the smeared
bloodline symptom.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
O(n^2) sweep over canonical + synthetic: any two nodes sharing a y are at least
NODE_W + COL_GAP apart.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fixture-wide loop over the canonical forest and a synthetic tree: each unit's
run centre is within [min, max] of its child-unit centres — the ancestor
centring invariant, asserted on real data.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A 5-generation single bloodline fanning out wide at the bottom: the apex
great-great-grandparent (and every ancestor in the chain) sits at the centre of
the descendant span, the exact symptom the old per-generation packer produced
in reverse (apex pinned to the left edge).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
StammbaumConnectors takes the layout's crossLinks and draws those parent->child
connectors with a 2 6 dash at reduced opacity — deliberately distinct from the
ended-marriage spouse dash (4 4) and from a solid parent drop. Geometry still
lands on the child top, so the meaning is carried redundantly (WCAG 1.4.1).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When the two spouses' parents sit at different structural levels, the
structural owner keeps its hierarchy edge and the other parent->spouse edge is
recorded in layout.crossLinks (rendered with a distinct dash). The couple still
sits exactly adjacent in the owner's run and B keeps a real position.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Extends the existing adjacency contract: the couple is exactly adjacent in the
run AND, because both parents are roots (same structural level), the displaced
parent edge stays solid — layout.crossLinks is empty for this case.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
buildLayout now builds the family forest, packs it bottom-up via tidyTree, and
maps each unit's run x back to per-person positions (x from structure, y from
rank). assignRanks, the generations map, and computeViewBox are reused
unchanged. The unknown-id guard now covers PARENT_OF as well as SPOUSE_OF, and
displaced cross-level edges are exposed as crossLinks for distinct rendering.
The ~210-line block packer (and its block/merge helpers) is gone.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Net-new ordering coverage: roots and every unit's children sort by birthYear
ASC (undated last), then displayName, then stable id — so horizontal x never
depends on Map iteration order.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Assigns every person to one unit: a primary, or a spouse absorbed into the
primary's run (marriage-year order, #361 preserved). Wires the parent/child
hierarchy from each primary's structural-owner parent and records displaced
parent edges as cross-links (classified same-level vs cross-level for later
distinct rendering). Unknown-id guard covers PARENT_OF and SPOUSE_OF.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Structural-owner rule for couples: earlier birth year wins, missing year sorts
last, ties break on stable id. The single definition reused by the cross-link,
cycle and intra-family paths.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
New domain-agnostic bottom-up tidy-tree module (Reingold-Tilford contour pack)
operating on abstract { id, width, children } nodes — zero generated-API
imports. First rung of the TDD ladder: a single leaf lays out at x=0. The full
contour/centring machinery is in place; subsequent commits add tests that
exercise it.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The existing node() factory never sets birthYear, but the new sibling/branch
comparator (birthYear ASC NULLS LAST) needs it. Add makeNode(id, name,
{birthYear, generation}) alongside it; unblocks every ordering test.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The component-test browser env (src/test-setup.ts) loads no Tailwind
stylesheet, so the footer buttons' min-h/min-w-[44px] classes have no
layout effect there and the elements collapse to their 16px icon —
making the getBoundingClientRect size assertions fail in CI.
Assert the sizing utility classes instead; they are the exact mechanism
that produces the WCAG 2.2 §2.5.8 target size in the real app. The
compiled pixel size remains covered by the full-app e2e.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Fix a stale test title that still claimed a delete button is visible.
- Strengthen the two "never renders a delete button" contract tests
(AnnotationShape + AnnotationLayer specs) to assert the annotation
element has zero descendant <button> elements, not just the absence of
the removed testid (a near-tautology now that the testid is gone).
- Harden the e2e delete test: guard countBefore > 0 so a missing seed
fails clearly instead of asserting toHaveCount(-1), and capture the
deleted annotation's testid to assert that specific element is gone
(identity check) alongside the count drop.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The panel footer's delete and review-toggle controls were icon-only ~16px
hit areas. After #722 removed the on-canvas delete button, the panel delete
button became the only touch-reachable delete path, so it must meet the WCAG
2.2 §2.5.8 minimum target size (44×44px). Give both icon-only footer actions
a >=44px inline-flex hit area with negative margins so the row layout and the
visible icon size are unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The per-annotation delete button (a 44px circular control pinned to the
box's top-right) overlapped the box below and obscured the underlying
document text. It was redundant: every user-drawn annotation has a
transcription block, and the right-hand panel already offers a
non-overlapping delete per block that cascades to the annotation.
Remove the visible button and its `deleteVisible` derived. Keep the
keyboard Delete shortcut (and its `showDelete`/`onDeleteRequest`/
`deleteAnnotation` wiring) — it obscures nothing and remains a
power-user path and the only cleanup route for orphan annotations.
Tests: replace the button-render/click specs with contract tests
asserting no delete button ever renders; repoint the e2e delete flow
to the keyboard shortcut + confirm dialog.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
CI showed the single/many a11y tests failing with count 0: init()'s async
fetchUnreadCount resolved to {count:0} AFTER setNotifications() ran,
clobbering the seeded count (the flake Sara predicted in review). Stub
fetch to never settle so the announced count is driven solely by
setNotifications — deterministic, no race. Also rewrites the 'error' test
to seed a count then fail the load and assert the count SURVIVES, so it is
a meaningful state distinct from 'empty' (was byte-identical, flagged by
Felix/Sara/Leonie). Part of #560.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
CI proved cross-file sharing of a virtual-module mock body cannot work in
@vitest/browser-playwright 4.1.6: the static-import spread fails the hoist
("no top level variables"), and the await-vi.hoisted-import form fails to
parse ("Unexpected identifier 'vi'"). vi.hoisted has the same hoist
constraint as vi.mock, so there is no way to thread an external module's
body into the factory here.
Reverts Phase 1: restores the 4 $app/forms/$app/navigation specs to their
inline factories, inlines NotificationBell.spec's forms stub, deletes the
src/__mocks__/$app/* modules and the $mocks alias (vite, vitest-coverage,
kit). The no-factory-ban meta-test stays (no-factory vi.mock is still
banned). ADR-012 amended to record the infeasibility. Everything else
($app/state migration, confirm context-inject, notification refactor, the
pin, the meta-test) is unaffected. Part of #560.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
CI caught that vi.mock('$app/forms', () => ({ ...formsMock })) with a
static `import * as formsMock` fails: vitest hoists vi.mock above the
import, so the factory references an uninitialised binding
("no top level variables inside"). Load the shared mock module via
`const formsMock = await vi.hoisted(() => import('$mocks/...'))` instead —
the factory may reference a vi.hoisted binding, and the dynamic import runs
at collection time (not in the lazily-invoked factory), so it stays clear
of ADR-012's birpc race and the no-async-mock-factories guard. Applies to
all 5 shared-mock consumers ($app/forms x4, $app/navigation x1). Part of #560.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Converts the module-singleton notificationStore into a context-provided
store so its specs can drive it without mocking the module. notifications.svelte
now exports createNotificationStore() (the former singleton body), plus
provideNotificationStore()/getNotificationStore()/NOTIFICATION_KEY mirroring
the confirm service. Root +layout provides it; NotificationBell and the
Chronik page read it via getNotificationStore().
Tests:
- notifications.svelte.spec drives a fresh createNotificationStore() per test
(replacing __resetForTest/__setNavigateForTest with setNavigate()).
- notification.test-fixture.svelte wraps the bell, provides the store, and
exposes setNotifications(items) via onReady (option b).
- NotificationBell.svelte.spec asserts the announced unread count across the
empty / single / many / error a11y states (AC#5), stubbing EventSource+fetch.
- aktivitaeten page spec injects a real store via render context.
Per the recorded Phase-2b decision (full context refactor). Part of #560.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replaces the vi.mock('$lib/shared/services/confirm.svelte') stub with a
real createConfirmService() provided through render's context map, mirroring
the existing admin/tags/[id]/page.svelte.spec.ts pattern. The generic
confirm.test-fixture.svelte renders only ConfirmDialog and cannot wrap an
arbitrary page; none of these specs trigger confirm(), so the children's
getConfirmService() simply reads the provided context instead of a module
mock. No vi.mock of confirm.svelte remains in these 5 specs. Part of #560.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replaces the local beforeNavigate-capture plumbing and simulateNavigate
helper with the shared $mocks/$app/navigation module via a sync factory.
The per-test reset now comes from the shared module's embedded beforeEach.
Part of #560.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Completes Phase 1a after the load-bearing ChronikFuerDichBox spec proved
the pattern. ChronikFuerDichBox.test and NotificationDropdown.test (rich
result-firing interceptors) keep their submit-fired assertions
(optimisticMarkRead/MarkAllRead) and use formsMock.setFormResult for the
failure branch. NotificationBell.spec used the simpler intercept-only
factory and renders no form of its own, so it adopts the shared superset
purely as a render-time stub. Part of #560.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Load-bearing first migration (ADR-012): this is the hardest case — its
enhance submit callback actually fires and reads the form result. Replaces
the duplicated 23-line interceptor factory with vi.mock('$app/forms',
() => ({ ...formsMock })) via $mocks, and the per-test mockFormResult
mutation with formsMock.setFormResult({ type: 'failure' }). The reset now
comes from the shared module's embedded beforeEach. The existing
optimisticMarkRead/optimisticMarkAllRead-on-submit assertions remain as the
positive proof the callback fired. Part of #560.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Drop the /api/documents/conversation path and its getConversation
operation from the generated client to match the removed backend
endpoint.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
With the Briefwechsel view gone, lib/conversation/ held a single shared
component whose only consumer is lib/document/ThumbnailRow. Move it (and
its spec) into lib/document/, update the import, delete the now-empty
lib/conversation/ folder, and fix the stale frontend/CLAUDE.md lib map.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Person nodes rendered in `nodes` array order (backend/DB row order), so
Tab focus hopped between nodes unrelated to their on-screen position,
failing WCAG 2.4.3 Focus Order (Level A).
Render the node loop in reading order instead: sort by layout y (top
generation first) then x (left-to-right within a row), via a
`nodesInReadingOrder` derived. Nodes without a layout position sort last
(mirroring the `{#if pos}` guard); node.id is the final tie-break for a
total, deterministic comparator. Shift+Tab and reload-stability fall out
for free (reversed render order; x/y independent of backend order).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Addresses the clean-agent review of PR #717:
- C1: the hidden pencil was opacity-0 only, which still hit-tests; its 44px box
overhangs adjacent text, so a click in the gap between two mentions could land
on the invisible button and spuriously open the dropdown (AC-8 hole). Add
pointer-events-none while hidden, re-enabled with the opacity reveal on
hover/focus.
- C2/N1: editor.setEditable() emits "update", not a ProseMirror transaction, so
the NodeView's 'transaction' listener missed a mid-session disable flip (stale
aria-disabled/tabindex; the comment was wrong). Listen on 'update' instead —
which also skips selection-only changes, so it fires far less often.
- N2: track the node across update() so the pencil opens with the live
displayName (hardening; relink only swaps personId today).
Tests: structural guard that the hidden pencil is pointer-events-none + reveals,
and a mid-session disable-flip test (fixture gains an onReady setDisabled hook).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Passes editingDisplayName into MentionDropdown; the persistent aria-live region
announces person_mention_editing_announce({displayName}) on re-edit open and
falls back to the prompt/empty/count copy once the user edits or results arrive.
Routed through the SAME sr-only region as the result count — no second live
region (avoids the double-announce bug Leonie S-2 fixed). Fresh-@ passes an
empty editingDisplayName, so its announcements are unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- AC-7: disabled editor → pencil is disabled + aria-disabled + tabindex -1, and
neither keyboard nor pointer activation mounts a dropdown (WCAG 2.1.1, not just
pointer-events-none).
- AC-8: plain text shows no pencil/dropdown; two adjacent mentions each keep one
pencil with no spurious gap pencil and no auto-open; a doc-start mention still
renders its pencil.
- Security: an oversized stored displayName clips the search query to 100 chars
while the preserved node text stays full-length; re-link sources personId
solely from the picked Person (p-anna), never the reflected/clipped text.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Locks in the single-owner controller guarantees: pencil→pencil, fresh-@→pencil
and pencil→fresh-@ all leave exactly one dropdown open; the request-token bump
on open discards a superseded open's in-flight fetch (open A → open B → A
resolves, deterministic, no sleeps). Plus a #380 AC-1 regression guard that the
fresh-@ path still inserts the typed text as displayName after the controller
refactor.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds a visible × dismiss control to MentionDropdown (shared by the fresh-@ and
re-edit paths) and, for the re-edit path which has no Tiptap suggestion plugin
to forward keys, focuses the search input on open and handles its own keyboard:
Escape dismisses (AC-4), Arrow/Enter reuse the exported selection logic so the
dropdown is navigable on its own (AC-9 parity with the fresh-@ dropdown).
Both close paths (Escape + ×) leave the mention node attrs + text byte-identical
(AC-4) — close() never touches the document. Controller wires ondismiss=close
(+refocus editor) and focusOnMount only for the re-edit open.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Hosts each mention as a Tiptap NodeView (mentionNodeView.ts) that renders the
@displayName token (textContent — never innerHTML) plus a contenteditable=false
pencil button in a fixed-width slot, revealed on whole-token hover and keyboard
focus (instant opacity swap, no reflow). Activating the pencil (click or Enter/
Space) opens the single mention dropdown via the controller, anchored at the
token and pre-filled with the stored displayName.
commitRelink swaps ONLY personId in place via setNodeMarkup, sourcing the id
solely from the selected Person — the stored displayName is preserved by
construction (AC-3), even after the search input is edited (AC-5, the #380 AC-1
invariant). renderHTML/renderText stay for serialization + clipboard.
AC-1/AC-2/AC-3/AC-5 + serializer round-trip covered by browser tests.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Pulls mountedDropdown / requestId / debouncedSearch / dropdownState ownership
out of Tiptap's suggestion.render() closure into one createMentionController().
render() becomes a thin adapter: onStart→open, onUpdate→update, onExit→close.
This is the single-owner structure #628 needs for the AC-6 single-dropdown
invariant — the upcoming pencil re-edit affordance opens via the same
controller.open() instead of racing the suggestion plugin over module state.
open() now also bumps the request token so an open-A→open-B sequence discards
A's in-flight fetch (preserved increment-on-open semantics). No behaviour
change for the fresh-@ path — existing browser suite is the regression guard.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Address the remaining UI/UX polish: add a warning-triangle icon so the
failure is signalled by shape, not colour alone (WCAG 1.4.1); give the
recovery download link a full 44px tap/focus target (inline-flex
min-h-[44px]); and soften the message copy in de/en/es.
Addresses re-review: Leonie (colour-only, undersized link, copy warmth).
Refs #708
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The error block was a colour-only, visually-small dead end. Add
role="alert" so screen readers announce the failure, bump the message to
text-base and the recovery download link to text-sm with a py-2 tap
target — the only escape hatch, sized for the archive's older readers.
Addresses re-review: Leonie (a11y of the error state).
Refs #708
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The "render failure" test rejected getDocument().promise — the load
path, not the render path — and only asserted a template constant. Now
the fake loads the document successfully and rejects the page render
(the actual #708 wasm-decode failure class), plus a negative companion
asserting the message is absent on a successful render. Also reset
renderTask to null on the render-error path.
Addresses re-review: Felix, Sara (mislabeled test / asserted a constant).
Refs #708
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The render path was localized but loadDocument still stored the raw
pdf.js message (and an untranslated English fallback), contradicting the
"never leak raw error text" principle. Both load and render failures now
set the localized doc_render_failed message.
Addresses re-review: Felix, Nora (raw error leak on the load path).
Refs #708
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The in-browser pixel-render fixture test was green locally but flaky in
CI: the real pdf.js worker could not fetch /pdfjs-wasm/ in the CI
Chromium container, so the CCITT canvas stayed blank (0 sampled pixels)
and failed the suite — green locally, red in CI, root cause not locally
reproducible. A flaky gate is worse than none.
This bug is a build/serve parity failure, so guard it deterministically
where it actually breaks: a postbuild assertion that jbig2.wasm and
openjpeg.wasm shipped into build/client/pdfjs-wasm/ (non-empty). It runs
after `npm run build` — including the Docker build stage — and fails the
build loudly if a future pdfjs bump makes the static-copy glob match
nothing. Combined with the getDocument(wasmUrl) unit guard and the
negative-path render test, the regression is covered without CI flake.
Addresses re-review: Tobias (no automated parity check), Sara (pixel
test not pinned). Render-decode correctness verified manually via
`node build` serving /pdfjs-wasm/jbig2.wasm as application/wasm.
Refs #708
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Render committed synthetic fixtures through PdfViewer with the REAL
pdf.js loader and assert the canvas is non-blank (sampled dark-pixel
count). The CCITT (G4 fax) fixture exercises the shared jbig2.wasm
decode path — the same module pdf.js uses for JBIG2 — so it transitively
covers the JBIG2 acceptance criterion (the archive sample found zero
true JBIG2 docs and jbig2enc is unavailable to synthesize one). The
JPEG/DCTDecode fixture guards against regressing the natively-decoded
path. Verified the CCITT case goes red when wasmUrl is removed.
Fixtures are hermetic, committed assets (~2-5 KB each), generated with
ImageMagick — never fetched from staging at test time. CI browser mode.
Refs #708
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The error-state download link opened with target="_blank" but no rel,
exposing the opener to reverse tabnavbabbing. Add rel="noopener
noreferrer". Same-origin so low severity, but a one-token fix in a file
this issue already touches.
Refs #708
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The error state showed a hardcoded German string ("Fehler beim Laden
der PDF" / "Direkt öffnen") to all users regardless of locale. Use the
localized doc_render_failed and doc_download_link messages so the
recovery path (message + working download link) is honest in de/en/es.
Refs #708
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
renderCurrentPage swallowed every render rejection with a bare return,
so a decode failure left a blank white viewer with no feedback. Now a
non-cancellation rejection sets a localized doc_render_failed message,
which routes into the existing error UI (message + download link).
Cancellation (page-nav / zoom) still returns silently — no error.
Refs #708
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
getDocument was called with a bare src string, so pdf.js 5.x had no
`wasmUrl` and could not initialise the JBIG2/CCITTFax wasm decoder —
CCITT (G4 fax) scans painted a blank canvas. Pass
{ url, wasmUrl: '/pdfjs-wasm/' }; the directory URL (trailing slash
required) is the single source of truth next to the worker config.
Refs #708
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
With the visible "Originaltext" line gone from every view, the
date_original_label message has no remaining references — remove it from
de/en/es. Also drop the now-inaccurate comments in documentDate.ts that
described the raw cell as "preserved separately as the visible secondary
line"; the raw cell now only feeds the SEASON word and is never shown.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
