* feat(focus-mode): focus screen UX overhaul
Major rework of the focus mode timer screens (#7349):
- Shared <focus-clock-face> drives Pomodoro / Flowtime / Countdown /
Break with a single visual chrome; size tokens scale fluidly via
clamp() + vmin/vh, no discrete breakpoints.
- Single source of truth: --clock-time-size and --control-offset
derive from --clock-face-size.
- Countdown: click-to-edit duration, draggable handle on the ring,
hybrid 5-min/15-min snap with 6h-per-rotation above 1h
([useFlexibleIncrement] on input-duration-slider, opt-in for other
consumers).
- Pomodoro prep inherits the click-to-type input; drag handle hidden
so dragging the ring can't silently shift the value.
- Pause keeps the selected task on screen (displayedTask falls back
to the paused task while currentTask is null).
- Notes panel acts as a modal: clock stays put, backdrop dims and
closes on outside-click.
- Session controls row below the circle (pause / complete / reset
cycles); buttons fade via shared --revealed-opacity gated on host
hover + document.hasFocus().
- Break screen mirrors focus-mode-main layout; cycle counter inside
the circle on both focus and break; back-to-planning unified.
- Flowtime settings dialog: all fields render with proper
disable/enable, stable dialog width when switching modes.
- arrow_backward -> arrow_back: fixes glitched glyph in repeat-type
context menus (boards, simple counters, take-a-break, flowtime).
* fix(focus-mode): silence naming-convention lint on formly 'props.disabled'
Formly's expressionProperties path-string keys ('props.disabled') aren't
camelCase and the rule has no requiresQuotes exemption; matches the
existing pattern in src/app/features/issue/common-issue-form-stuff.const.ts.
* test(focus-mode): mock pomodoroConfig signal on FocusModeService
The component's initialization effect now reads
focusModeService.pomodoroConfig() in Pomodoro+Preparation mode, but the
three mocked FocusModeService instances in the spec didn't provide it,
causing TypeError in 47 tests on CI (somehow not surfaced locally).
* test(focus-mode): align specs with unified back-to-planning flow
- focus-mode-break.spec: exitBreakToPlanning -> cancelFocusSession
- focus-mode-session-done.spec: drop obsolete hideFocusOverlay assertion
(cancelFocusSession now handles both clearing tracking and hiding the
overlay, matching the production component)
- focus-mode-main.spec: storeSpy gains selectSignal returning a signal,
needed by the displayedTask paused-task fallback
* fix(focus-mode): restore E2E selector hooks on refactored controls
The shared <focus-clock-face> refactor moved pause/complete buttons out
of the clock face into a new .circle-controls row but didn't carry the
class names forward; 23 E2E specs key off them. Also re-add
.task-title-placeholder on the no-task FAB so prep-state checks find it.
- .pause-resume-btn on pause/resume in focus-mode-main + focus-mode-break
- .complete-session-btn on the done_all button in focus-mode-main
- .task-title-placeholder added to .select-task-cta FAB
* fix(focus-mode): aria-label icon buttons; align break E2Es with new flow
- focus-mode-break: pause/resume/skip/reset icon buttons now carry
[attr.aria-label] in addition to matTooltip — icon ligature alone
isn't an accessible name and breaks getByRole locators.
- pomodoro-break-timing-bug-6044.spec: skipButton uses getByRole with
accessible name rather than hasText (mat-icon ligature is "skip_next",
not "skip break").
- focus-mode-break.spec: "exit break to planning and change timer mode"
and "Back to Planning should NOT auto-start next session" now match
the unified back-to-planning flow — overlay closes on click, user
re-opens focus mode to change settings or verify prep state.
* fix(focus-mode): address PR #7586 review feedback
Maintainer review (johannesjo):
- Debounce the Pomodoro work-duration write so editing it emits one
synced config op instead of one per keystroke; flush on session start
so a value typed inside the debounce window is not lost. (A1)
- Replace the local ::ng-deep restyling of the shared input-duration-slider
with opt-in [bareRing] and [hideHandle] inputs that own the chrome
overrides in the slider's own styles; the four other consumers keep the
default look. (A2)
- Add unit tests for the flexible drag math (_setValueFromRotationFlex):
the A<->B boundary anchoring at 55/60 min and both +/-180 degree wrap
branches. (A3)
- Delete the orphaned exitBreakToPlanning action, its
stopTrackingOnExitBreakToPlanning$ effect, reducer case and specs;
cancelFocusSession already unsets the current task. (B1)
- Drop the unreferenced CONTINUE_TO_NEXT_SESSION and BREAK_RELAX_MSG
i18n keys. (B2)
- Collapse the duplicated clock-size clamp() into a single
--clock-face-size-default token. (B4)
Smaller focus-screen fixes (beerkumquatpome):
- Hide the break task title when "pause tracking during breaks" is on. (C7)
- Commit and close the duration editor on Enter. (C9)
- Rename "Back to Planning" to "Exit focus session". (C11)
- Show Flowtime breaks as a neutral "Break". (C13)
Break-circle vertical alignment (C1) is only partially addressed here
(matched the top reservation); exact alignment is a follow-up.
* refactor(focus-mode): share a layout shell across timer screens and auto-start Flowtime breaks
Extract a presentational focus-mode-layout component (4-row content-projection
skeleton: [fmTop]/[fmTask]/[fmClock]/[fmBottom]) shared by the focus-session and
break screens, so both keep a stable clock baseline across the focus<->break and
prep<->in-progress transitions. focus-mode-main and focus-mode-break now consume
the shell instead of each maintaining their own absolute layout.
Replace the Flowtime "break offer" step with an auto-started break, mirroring
Pomodoro:
- Remove the BreakOffer UI state and the offerFlowtimeBreak action/reducer.
- endFlowtimeSession now dispatches completeFocusSession(isManual:false) +
startBreak (unsetting the task first when tracking-pause-on-break is on), so
the break starts automatically and the session is logged exactly once via
logFocusSession$.
Also reorder the bottom controls (Back to Planning leftmost), restore the
BACK_TO_PLANNING label to "Back to Planning", and drop the now-orphaned
FLOWTIME_BREAK_TITLE / START_BREAK i18n keys.
* test(layout): restore document.activeElement after focus-restoration specs
The LayoutService "Focus restoration" tests override document.activeElement
with Object.defineProperty, which shadows the native (inherited) getter with an
own property on document. The afterEach only removed the mock DOM node, so the
override leaked into later specs: once it ran, document.activeElement was frozen
at the mock element and subsequent .focus() calls could no longer move it.
Depending on Karma's spec order this broke the task.service focusTaskById tests
(#7120), which then saw the stale activeElement instead of the element they
focused. Delete the shadowing own property in afterEach to restore native
behavior.
Repro: ng test --include layout.service.spec.ts --include task.service.spec.ts
* refactor(focus-mode): add interactive tracking widget and polish timer-screen layout
- Replace the read-only task-tracking-info with focus-mode-task-tracking
(vertical time stack + play/pause), wired through the shared layout shell
- Center the task title and floor the task-row height so the clock baseline
stays aligned across focus <-> break
- Spacing polish: task-title-row and layout gaps to --s2,
segmented-button-group padding to 0
- Drop redundant safe-area-bottom padding on the action row (the overlay
already reserves it for the fixed shell)
- Add "Take a moment to relax" break message and a clock-digit edit affordance
* refactor(focus-mode): share timer/break layout, drop dead tracking toggle
- Extract a shared <focus-mode-layout> skeleton and <focus-mode-task-row> used by both the focus session and the break.
- Remove the in-view tracking play/pause toggle (start/stop stays on the global header button); strip focus-mode-task-tracking to read-only and drop RESUME_TRACKING.
- Pin the mode selector out of flow and center the task·clock·bottom group; equal reserved task/bottom rows keep the clock vertically centered, with the selector kept on top.
- Tighten sizing: horizontal selector segments, settings cog matched to the in-session controls, and a fluid clock clamp.
Reconstructs a single user's AppDataComplete by replaying their op log up to a chosen serverSeq and decrypting E2E-encrypted payloads — the recovery path the server's restore endpoint cannot offer for encrypted accounts (it throws EncryptedOpsNotSupportedError). Read-only on the DB; documented in docs/backup-and-recovery.md. Unverified against real encrypted data — see the script header.
- Add accounts-only pg_dump (users + passkeys) to backup script for
lightweight disaster recovery when clients still have data
- Add pipefail to backup script to catch silent dump failures
- Add test endpoint to simulate partial server revert (ops-after/:serverSeq)
- Add 6 e2e tests covering all disaster recovery scenarios:
complete data loss, partial revert, accounts-only restore (API + SQL),
full dump restore + reset account, and all-clients-lost recovery
- Add backup-and-recovery.md with setup, recovery procedures, and
decision tree
Replace separate JWT_EXPIRY_MAGIC_LINK (365d) and JWT_EXPIRY_PASSKEY (7d)
constants with a single JWT_EXPIRY (365d). The auth method only matters
during login — once a JWT is issued, it represents a verified session
regardless of how the user authenticated.
- Use JWT_EXPIRY_PASSKEY (7d) for replaceToken instead of 365d magic link
expiry — token replacement is a security action, shorter lifetime is safer
- Replace localeCompare with locale-independent comparator in vector clock
tie-breaking to ensure deterministic behavior across environments
- Fix 5 additional stale MAX=30 references in docs and tests (now 20)
- Update authentication.md to reflect dual JWT expiry tiers
- Clean up isLikelyPruningArtifact references in docs and LEGACY_MAX in tests
Both LUKS and TDE encryption don't work on OpenVZ:
- LUKS: requires dm-crypt kernel module (not available)
- TDE: technically not feasible during testing
Reverted TDE commit (1fdcc9a90) and moved all encryption scripts
to archive/encryption-attempts-openvz-incompatible/ for reference.
Decision: Operating without encryption-at-rest for now.
Future options documented in archive README:
- KVM migration (€2-5/month) enables LUKS
- Application-level encryption if staying on OpenVZ
- Managed PostgreSQL service (~€10-20/month)
Kept backup-encrypted.sh (encrypts backup files with passphrase,
separate from database encryption-at-rest).
Internal compliance documentation has been moved to a private location.
These documents contain sensitive operational procedures and security
analysis that should not be public.
Files moved:
- GDPR compliance analysis
- Incident response playbooks
- Data subject request procedures
- DPIA screening decisions
- Records of processing activities
- Infrastructure verification documents
Update incident response, data subject request, and DPIA procedures to
accurately reflect that database encryption at rest is NOT implemented
for non-E2EE users.
Changes:
- INCIDENT-RESPONSE-PLAYBOOK.md: Clarify E2EE is optional throughout,
add physical server compromise scenarios, update risk assessments to
differentiate E2EE vs non-E2EE users, document encryption gap in
prevention measures
- DATA-SUBJECT-REQUEST-PROCEDURES.md: Add encryption status disclosure
to access responses, clarify data export formats, add security notice
about unencrypted storage for non-E2EE users
- DPIA-SCREENING-DECISION.md: Document encryption gap as additional
consideration, update risk level to LOW-MEDIUM, add encryption gap
to conclusion and re-assessment triggers
All procedures now consistently acknowledge 85% compliance score and
risk variance based on E2EE usage, while maintaining that DPIA is not
required per Art. 35.
Update GDPR compliance documentation to accurately reflect that database
encryption at rest is NOT implemented for non-E2EE users. This critical
finding required:
- Update compliance score from 92% to 85% (10% deduction for encryption gap)
- Add comprehensive encryption disclosure to privacy policies (German & English)
- Document risk: unencrypted PostgreSQL data on disk
- Update GDPR analysis with compensating controls (optional E2EE)
- Revise Records of Processing Activities with encryption status
- Add context to Alfahosting verification tracker
Changes prioritize GDPR transparency by honestly documenting security
limitations rather than overstating compliance.
Explain design decisions and security trade-offs:
- Why stateless JWTs (not stored in DB)
- Why token versioning over blacklisting
- Why verification tokens are plain strings
- Why bcrypt with 12 rounds
Include security features table, token lifecycle diagram,
and configuration reference.
