Problem: When a meeting was flagged external, the row was tinted red
(pink bg + red number box — desired) but the title cell text and the
attendee list text were also forced to red (#b91c1c), which the user
did not want.
Change (artifacts/tx-os/src/pages/executive-meetings.tsx):
- Meeting title cell: always use navy text (#0B1E3F) instead of switching
to red on meeting.isExternal.
- Attendee list <ul>: always use navy text (#0B1E3F) instead of switching
to red on externalText.
- Removed the now-dead `externalText` prop end-to-end (call-site value,
type declaration, and destructured default) to keep typecheck clean.
Unchanged: the red row tint, red number box, and all external-meeting
behavior (e.g. auto-postpone rules). Internal/virtual meetings unaffected.
Verified: tx-os typecheck passes; HMR applied with no console errors.
- Attendees column no longer centres its content. AttendeeFlow's <ul>
switched from justify-center to justify-start; AttendeeGroup header
and subheading rows (both committed and pending) switched from
text-center to text-start. Restores cell-start alignment (right in
AR, left in EN) the user expected on fresh inline-add and on
existing rows.
- External meetings now render the meeting title cell text and the
attendee names in red (#b91c1c), matching the row tint family, in
addition to the existing red background. Implemented by:
* Conditional text-[#b91c1c] / text-[#0B1E3F] on the meeting <td>
based on meeting.isExternal.
* New optional `externalText` prop on AttendeeFlowSharedProps,
threaded from AttendeesCell (derived from meeting.isExternal) and
applied as a colour class on the AttendeeFlow root <ul> so all
attendee names inherit red. Group headers and subheadings keep
their neutral navy palette per scope.
- Non-external rows and the cancelled/highlighted white-on-red # cell
are untouched. Print/PDF path reuses the same renderer so the red
carries through automatically.
- No schema/API/migration changes. TS clean.
Files: artifacts/tx-os/src/pages/executive-meetings.tsx
- Attendees column no longer centres its content. AttendeeFlow's <ul>
switched from justify-center to justify-start; AttendeeGroup header
and subheading rows (both committed and pending) switched from
text-center to text-start. Restores cell-start alignment (right in
AR, left in EN) the user expected on fresh inline-add and on
existing rows.
- External meetings now render the meeting title cell text and the
attendee names in red (#b91c1c), matching the row tint family, in
addition to the existing red background. Implemented by:
* Conditional text-[#b91c1c] / text-[#0B1E3F] on the meeting <td>
based on meeting.isExternal.
* New optional `externalText` prop on AttendeeFlowSharedProps,
threaded from AttendeesCell (derived from meeting.isExternal) and
applied as a colour class on the AttendeeFlow root <ul> so all
attendee names inherit red. Group headers and subheadings keep
their neutral navy palette per scope.
- Non-external rows and the cancelled/highlighted white-on-red # cell
are untouched. Print/PDF path reuses the same renderer so the red
carries through automatically.
- No schema/API/migration changes. TS clean.
Files: artifacts/tx-os/src/pages/executive-meetings.tsx
Two issues reported on the new "اجتماع خارجي" toggle from #635:
1. "Doesn't save." Investigation: the server schema, PATCH/POST handlers,
GET endpoint (full-row select), and DB column are all correct, and
`docker/migrate.sh` runs `pnpm --filter @workspace/db run push-force`
on every `docker compose up`, so the production schema does receive
the `is_external` column. The likeliest real-world cause is a stale
API image (column existed but the post-review API code that persists
the flag wasn't deployed yet), or a transient stale-state render
where `state.isExternal` was undefined and silently dropped from the
PATCH body. Hardened the client so neither can happen again:
- Both save sites (Manage dialog `save()` ~6700, inline Schedule
edit save body ~3421) now send `Boolean(state.isExternal)` instead
of the raw value, so an undefined state coerces to `false` rather
than dropping the field.
2. "Overlaps other fields." The toggle row was sharing a row with the
date input in the 2-col grid and used a fixed `h-9` wrapper that
sat awkwardly next to the taller time pickers on narrow screens.
- Made the FormRow `full` so it spans both columns on its own row,
above the time pickers.
- Dropped the fixed `h-9` and switched to `py-1` + `gap-3` + `text-sm`
so it matches the form's vertical rhythm on iPhone, iPad portrait,
iPad landscape, and desktop.
Out of scope / not changed
- No server changes (the server already accepts boolean | number and
coerces to boolean; `Boolean(data.isExternal ?? false)` on POST and
`if (data.isExternal !== undefined)` on PATCH are unchanged).
- No new badge / copy — that lives under separate proposed work.
- Postpone / cascade rules unchanged.
Two issues reported on the new "اجتماع خارجي" toggle from #635:
1. "Doesn't save." Investigation: the server schema, PATCH/POST handlers,
GET endpoint (full-row select), and DB column are all correct, and
`docker/migrate.sh` runs `pnpm --filter @workspace/db run push-force`
on every `docker compose up`, so the production schema does receive
the `is_external` column. The likeliest real-world cause is a stale
API image (column existed but the post-review API code that persists
the flag wasn't deployed yet), or a transient stale-state render
where `state.isExternal` was undefined and silently dropped from the
PATCH body. Hardened the client so neither can happen again:
- Both save sites (Manage dialog `save()` ~6700, inline Schedule
edit save body ~3421) now send `Boolean(state.isExternal)` instead
of the raw value, so an undefined state coerces to `false` rather
than dropping the field.
2. "Overlaps other fields." The toggle row was sharing a row with the
date input in the 2-col grid and used a fixed `h-9` wrapper that
sat awkwardly next to the taller time pickers on narrow screens.
- Made the FormRow `full` so it spans both columns on its own row,
above the time pickers.
- Dropped the fixed `h-9` and switched to `py-1` + `gap-3` + `text-sm`
so it matches the form's vertical rhythm on iPhone, iPad portrait,
iPad landscape, and desktop.
Out of scope / not changed
- No server changes (the server already accepts boolean | number and
coerces to boolean; `Boolean(data.isExternal ?? false)` on POST and
`if (data.isExternal !== undefined)` on PATCH are unchanged).
- No new badge / copy — that lives under separate proposed work.
- Postpone / cascade rules unchanged.
User asked for three things in the executive-meetings module:
1. Remove the "daily number" input from the add/edit dialog — numbering
is auto-assigned by the server already, so the field was just noise.
2. Add an "اجتماع خارجي" (external meeting) toggle. When ON, the row
is force-tinted red in the schedule grid.
3. External meetings must be excluded from the auto-postpone flow:
the alert "Postpone" button is disabled, and the cascade shift
skips them as followers.
Changes
- lib/db/src/schema/executive-meetings.ts
+ `is_external boolean not null default false` column. Pushed via
drizzle-kit (no destructive migration; defaults backfill).
- artifacts/api-server/src/routes/executive-meetings.ts
+ `isExternal` added to base zod fields, POST insert, PATCH update.
+ POST/PATCH force `rowColor='red'` when `isExternal=true`.
+ POST /:id/postpone-minutes rejects external meetings with
HTTP 409 + code `external_meeting_no_auto_postpone`.
+ `computeCascadeShift` filters out external followers (preview +
writer stay in lockstep, as the comment promises).
- artifacts/tx-os/src/pages/executive-meetings.tsx
+ Meeting type + MeetingFormState gain `isExternal`; dropped the
`dailyNumber` field from the form state entirely.
+ `emptyMeetingForm` / `openEdit` updated; ManageSection.save() and
the inline ScheduleSection save body now send `isExternal` and
never send `dailyNumber`.
+ MeetingFormDialog: removed the daily-number FormRow and added a
Switch-based external-meeting FormRow with localized hint.
+ `rowColors` memo coerces to "red" whenever `m.isExternal`.
+ Audit-diff "interesting" list now includes `isExternal`.
- artifacts/tx-os/src/components/executive-meetings/upcoming-meeting-alert.tsx
+ Meeting type gains optional `isExternal`.
+ "Postpone" button disabled + tooltip when meeting is external.
- locales: added `field.isExternal`, `field.isExternalHint`,
`field.isExternalOn/Off`, and `alert.externalCannotPostpone` in
both ar.json and en.json. Existing dailyNumber keys kept (still
referenced by the manage-table column header / cell display).
Post-review hardening (from architect pass)
- PATCH now coerces rowColor → 'red' for any PATCH that touches a row
that is (or becomes) external. Prevents PDF/export paths that read
raw rowColor from showing a non-red tint for externals.
- Schedule row quick-actions popover: the "Postpone" button is now
disabled with a localized tooltip when the meeting is external,
matching the upcoming-meeting alert behavior.
- PostponeDialog.handleErr maps the 409 code
`external_meeting_no_auto_postpone` to the localized
`alert.externalCannotPostpone` toast so users see a real reason
instead of the raw English error string.
Notes
- The `dailyNumber` column stays in DB (heavily used for slot
persistence and the unique index); only the form input was removed.
Server's `nextDailyNumber()` already auto-assigns when the field is
absent on POST, and PATCH leaves it untouched when absent.
- Two pre-existing TS errors remain in api-server (`isHighlighted`
type-inference quirk on `z.union(...).transform(...)` and a
font_settings comparison) — not in scope for this task.
- Tests not added; verification by typecheck on changed surfaces and
workflow-restart smoke. No e2e harness was wired.
User asked for three things in the executive-meetings module:
1. Remove the "daily number" input from the add/edit dialog — numbering
is auto-assigned by the server already, so the field was just noise.
2. Add an "اجتماع خارجي" (external meeting) toggle. When ON, the row
is force-tinted red in the schedule grid.
3. External meetings must be excluded from the auto-postpone flow:
the alert "Postpone" button is disabled, and the cascade shift
skips them as followers.
Changes
- lib/db/src/schema/executive-meetings.ts
+ `is_external boolean not null default false` column. Pushed via
drizzle-kit (no destructive migration; defaults backfill).
- artifacts/api-server/src/routes/executive-meetings.ts
+ `isExternal` added to base zod fields, POST insert, PATCH update.
+ POST/PATCH force `rowColor='red'` when `isExternal=true`.
+ POST /:id/postpone-minutes rejects external meetings with
HTTP 409 + code `external_meeting_no_auto_postpone`.
+ `computeCascadeShift` filters out external followers (preview +
writer stay in lockstep, as the comment promises).
- artifacts/tx-os/src/pages/executive-meetings.tsx
+ Meeting type + MeetingFormState gain `isExternal`; dropped the
`dailyNumber` field from the form state entirely.
+ `emptyMeetingForm` / `openEdit` updated; ManageSection.save() and
the inline ScheduleSection save body now send `isExternal` and
never send `dailyNumber`.
+ MeetingFormDialog: removed the daily-number FormRow and added a
Switch-based external-meeting FormRow with localized hint.
+ `rowColors` memo coerces to "red" whenever `m.isExternal`.
+ Audit-diff "interesting" list now includes `isExternal`.
- artifacts/tx-os/src/components/executive-meetings/upcoming-meeting-alert.tsx
+ Meeting type gains optional `isExternal`.
+ "Postpone" button disabled + tooltip when meeting is external.
- locales: added `field.isExternal`, `field.isExternalHint`,
`field.isExternalOn/Off`, and `alert.externalCannotPostpone` in
both ar.json and en.json. Existing dailyNumber keys kept (still
referenced by the manage-table column header / cell display).
Post-review hardening (from architect pass)
- PATCH now coerces rowColor → 'red' for any PATCH that touches a row
that is (or becomes) external. Prevents PDF/export paths that read
raw rowColor from showing a non-red tint for externals.
- Schedule row quick-actions popover: the "Postpone" button is now
disabled with a localized tooltip when the meeting is external,
matching the upcoming-meeting alert behavior.
- PostponeDialog.handleErr maps the 409 code
`external_meeting_no_auto_postpone` to the localized
`alert.externalCannotPostpone` toast so users see a real reason
instead of the raw English error string.
Notes
- The `dailyNumber` column stays in DB (heavily used for slot
persistence and the unique index); only the form input was removed.
Server's `nextDailyNumber()` already auto-assigns when the field is
absent on POST, and PATCH leaves it untouched when absent.
- Two pre-existing TS errors remain in api-server (`isHighlighted`
type-inference quirk on `z.union(...).transform(...)` and a
font_settings comparison) — not in scope for this task.
- Tests not added; verification by typecheck on changed surfaces and
workflow-restart smoke. No e2e harness was wired.
User-reported regression from #630: opening "Send note" on iPad PWA
showed the recipient list pushed to the left side of the screen,
narrower than designed, with the dialog title + search input clipped
off the top. Dismissing the keyboard restored normal centered layout.
Root cause: the `keyboardFallback` branch (iPad PWA, where
visualViewport doesn't update) set `top: "6vh"` in *layout viewport*
coordinates. On iPad PWA Safari, opening the keyboard scrolls the
page down behind the keyboard, so layout-viewport `6vh` ends up
above the visible visual viewport — clipping the top half of the
dialog. Separately, the inline `transform` replaced Tailwind's
translate-x/y on Radix's content, and without an explicit `left:50%`
the animation/RTL combination shifted the box left.
Fix (artifacts/tx-os/src/components/notes/send-note-dialog.tsx):
- `keyboardFallback`: `top: calc(${vv.offsetTop || 0}px + 6vh)` so the
anchor follows the visible visual viewport when Safari scrolls the
page behind the keyboard. Add explicit `left: 50%`. Add
`overflowY: auto` so if dialog content exceeds 50dvh, the dialog
scrolls internally instead of clipping the header.
- `keyboardActive` branch: also add explicit `left: 50%` defensively
for the same animation/RTL reason (no behavior change on devices
where it was already working, but eliminates the asymmetry).
No changes to the shared `dialog.tsx`, composer, or executive-meetings
dialog — scope strictly limited to the regression site.
TypeScript: clean. Pre-existing errors in api-server/src/routes/push.ts
unrelated to this diff.
User-reported regression from #630: opening "Send note" on iPad PWA
showed the recipient list pushed to the left side of the screen,
narrower than designed, with the dialog title + search input clipped
off the top. Dismissing the keyboard restored normal centered layout.
Root cause: the `keyboardFallback` branch (iPad PWA, where
visualViewport doesn't update) set `top: "6vh"` in *layout viewport*
coordinates. On iPad PWA Safari, opening the keyboard scrolls the
page down behind the keyboard, so layout-viewport `6vh` ends up
above the visible visual viewport — clipping the top half of the
dialog. Separately, the inline `transform` replaced Tailwind's
translate-x/y on Radix's content, and without an explicit `left:50%`
the animation/RTL combination shifted the box left.
Fix (artifacts/tx-os/src/components/notes/send-note-dialog.tsx):
- `keyboardFallback`: `top: calc(${vv.offsetTop || 0}px + 6vh)` so the
anchor follows the visible visual viewport when Safari scrolls the
page behind the keyboard. Add explicit `left: 50%`. Add
`overflowY: auto` so if dialog content exceeds 50dvh, the dialog
scrolls internally instead of clipping the header.
- `keyboardActive` branch: also add explicit `left: 50%` defensively
for the same animation/RTL reason (no behavior change on devices
where it was already working, but eliminates the asymmetry).
No changes to the shared `dialog.tsx`, composer, or executive-meetings
dialog — scope strictly limited to the regression site.
TypeScript: clean. Pre-existing errors in api-server/src/routes/push.ts
unrelated to this diff.
Bug: order creator (services manager) saw new orders instantly with
sound, but other users holding orders.receive — even with the app
foregrounded on the Orders screen — got nothing until they force-quit
and reopened the PWA, at which point the orders appeared. Root cause:
Socket.IO silently drops behind Tailscale/NAT on iPad PWA and the
client neither detected it nor refetched missed state on reconnect.
The creator was unaffected because their POST mutation updates their
queries locally without depending on the realtime channel.
Client (artifacts/tx-os/src/hooks/use-notifications-socket.ts):
- Tighten io() options: reconnection {Delay 500, DelayMax 3000,
Attempts Infinity, timeout 10s}.
- Track `wasDisconnected` flag; on `disconnect` flip it and log the
reason (skipping the clean "io client disconnect" path).
- On `connect`, if previously disconnected, invalidate every realtime-
driven query key: notifications, home stats, my/incoming orders,
notes, note-folders, exec meetings (list/alert-state/notifications),
apps, /me, roles, permissions. Warmup window is reset BEFORE the
refetch so any flushed notification_created events post-reconnect
don't chime.
- Add a `visibilitychange` listener: on foreground return, if the
socket is disconnected, force `socket.connect()`; if it's "connected"
but possibly half-open (silent drop), round-trip a 3s-timeout
`client_health_probe` ack — on timeout, `disconnect().connect()`.
- `connect_error` logger for field debugging.
Server (artifacts/api-server/src/index.ts):
- Tighten Socket.IO pingInterval=10s, pingTimeout=5s (was default
25s/20s) so dead-connection detection cycle drops from ~45s to ~15s.
- Add `client_health_probe` handler that acks immediately — pairs
with the client-side half-open probe.
Deviations from plan: skipped the optional UI connection indicator
(point 5) — not necessary to fix the reported bug; can ship later if
users still feel uncertain about connection state.
Architect approved with one minor caveat: severe (>3s) transient
latency on foreground could trigger a one-off socket cycle. Acceptable
tradeoff and explicitly documented in the comment.
Pre-existing TS errors in api-server/src/routes/push.ts are unrelated
to this task and not touched by this diff.
Bug: order creator (services manager) saw new orders instantly with
sound, but other users holding orders.receive — even with the app
foregrounded on the Orders screen — got nothing until they force-quit
and reopened the PWA, at which point the orders appeared. Root cause:
Socket.IO silently drops behind Tailscale/NAT on iPad PWA and the
client neither detected it nor refetched missed state on reconnect.
The creator was unaffected because their POST mutation updates their
queries locally without depending on the realtime channel.
Client (artifacts/tx-os/src/hooks/use-notifications-socket.ts):
- Tighten io() options: reconnection {Delay 500, DelayMax 3000,
Attempts Infinity, timeout 10s}.
- Track `wasDisconnected` flag; on `disconnect` flip it and log the
reason (skipping the clean "io client disconnect" path).
- On `connect`, if previously disconnected, invalidate every realtime-
driven query key: notifications, home stats, my/incoming orders,
notes, note-folders, exec meetings (list/alert-state/notifications),
apps, /me, roles, permissions. Warmup window is reset BEFORE the
refetch so any flushed notification_created events post-reconnect
don't chime.
- Add a `visibilitychange` listener: on foreground return, if the
socket is disconnected, force `socket.connect()`; if it's "connected"
but possibly half-open (silent drop), round-trip a 3s-timeout
`client_health_probe` ack — on timeout, `disconnect().connect()`.
- `connect_error` logger for field debugging.
Server (artifacts/api-server/src/index.ts):
- Tighten Socket.IO pingInterval=10s, pingTimeout=5s (was default
25s/20s) so dead-connection detection cycle drops from ~45s to ~15s.
- Add `client_health_probe` handler that acks immediately — pairs
with the client-side half-open probe.
Deviations from plan: skipped the optional UI connection indicator
(point 5) — not necessary to fix the reported bug; can ship later if
users still feel uncertain about connection state.
Architect approved with one minor caveat: severe (>3s) transient
latency on foreground could trigger a one-off socket cycle. Acceptable
tradeoff and explicitly documented in the comment.
Pre-existing TS errors in api-server/src/routes/push.ts are unrelated
to this task and not touched by this diff.
Problem: opening Send Note on iPad, tapping the search field opened the
soft keyboard which covered the recipients list and the Send button —
the dialog stayed centered on the full layout viewport.
Fix (artifacts/tx-os/src/components/notes/send-note-dialog.tsx):
Two-tier keyboard detection:
1. Primary — `vv.keyboardInset > 0` from useVisualViewport. Works on
desktop touch laptops, iPad Safari (regular tab), Android Chrome.
When true, override DialogContent style with
top = vv.offsetTop + vv.height/2 − 48 (QuickType inset)
transform: translate(-50%, -50%)
so the dialog recenters around the visible viewport region.
2. Fallback — `searchFocused && matchMedia('(pointer: coarse)')`.
Only used when visualViewport does NOT report an inset, which is
the iPad PWA (Add to Home Screen) case explicitly called out in
components/ui/dialog.tsx comment. Top-anchors at 6vh with
maxHeight 50dvh so the dialog never crosses the screen midpoint.
3. Neither — pass NO style at all so Radix's default centered layout
is fully restored (no leftover transition or top).
Touch-only fallback gating (matchMedia pointer:coarse) prevents the
earlier desktop-regression flagged by code review — focusing the
search on a desktop never repositions the dialog.
Other changes:
- onFocus/onBlur on the search input feed the fallback signal.
- Blur is debounced 150ms so tapping a recipient row mid-blur doesn't
recenter the dialog out from under the finger.
- useEffect cleanup clears pending blur timeout on unmount.
Verification: tsc clean. Architect approved the touch-gated fallback;
follow-up review reconciled the diff with the spec's primary
visualViewport-driven path.
Problem: opening Send Note on iPad, tapping the search field opened the
soft keyboard which covered the recipients list and the Send button —
the dialog stayed centered on the full layout viewport.
Fix (artifacts/tx-os/src/components/notes/send-note-dialog.tsx):
Two-tier keyboard detection:
1. Primary — `vv.keyboardInset > 0` from useVisualViewport. Works on
desktop touch laptops, iPad Safari (regular tab), Android Chrome.
When true, override DialogContent style with
top = vv.offsetTop + vv.height/2 − 48 (QuickType inset)
transform: translate(-50%, -50%)
so the dialog recenters around the visible viewport region.
2. Fallback — `searchFocused && matchMedia('(pointer: coarse)')`.
Only used when visualViewport does NOT report an inset, which is
the iPad PWA (Add to Home Screen) case explicitly called out in
components/ui/dialog.tsx comment. Top-anchors at 6vh with
maxHeight 50dvh so the dialog never crosses the screen midpoint.
3. Neither — pass NO style at all so Radix's default centered layout
is fully restored (no leftover transition or top).
Touch-only fallback gating (matchMedia pointer:coarse) prevents the
earlier desktop-regression flagged by code review — focusing the
search on a desktop never repositions the dialog.
Other changes:
- onFocus/onBlur on the search input feed the fallback signal.
- Blur is debounced 150ms so tapping a recipient row mid-blur doesn't
recenter the dialog out from under the finger.
- useEffect cleanup clears pending blur timeout on unmount.
Verification: tsc clean. Architect approved the touch-gated fallback;
follow-up review reconciled the diff with the spec's primary
visualViewport-driven path.
Problem: opening Send Note on iPad PWA, tapping the search field popped
up the soft keyboard which covered the entire people list and the Send
button. The previous fix relied on visualViewport to shrink the list,
but iPad PWAs don't reliably resize visualViewport when the keyboard
opens (see the note in components/ui/dialog.tsx referencing the
reverted #622–#624 global attempts).
Fix (artifacts/tx-os/src/components/notes/send-note-dialog.tsx):
- Track search-input focus directly — the most reliable "keyboard is
about to open" signal on touch devices, independent of visualViewport.
- Gate the override behind `matchMedia('(pointer: coarse)')` so it
only fires on touch-primary devices (iPad/iPhone/Android) and never
on desktop or trackpad laptops — addressing the architect's
cross-platform regression flag.
- When the gate is active, override DialogContent's Radix center
anchor (top:50%; translate(-50%,-50%)) with `top:6vh;
translate(-50%,0); maxHeight:50dvh`. The dialog grows downward from
a fixed top edge so it always fits in the upper half of the screen,
leaving the lower half free for the keyboard.
- Cap `listMaxHeight` to ~50% of innerHeight in the same gated
condition, as a backstop where visualViewport stays full-height.
- Debounce blur by 150ms so a tap on a recipient row doesn't lose
the row mid-tap to a center-jump animation; clear the timeout on
unmount.
- Added 120ms ease transition on top/transform so the layout swap
doesn't snap.
Deviations: original plan used vv.offsetTop+vv.height/2 to reposition,
but that depends on the same unreliable visualViewport. Switched to
a focus + coarse-pointer driven approach which works on iPad PWA.
Verification: tsc clean. Architect re-review pending after the touch
gate was added (first review flagged the missing platform guard,
which this revision addresses directly).
Problem: opening Send Note on iPad PWA, tapping the search field popped
up the soft keyboard which covered the entire people list and the Send
button. The previous fix relied on visualViewport to shrink the list,
but iPad PWAs don't reliably resize visualViewport when the keyboard
opens (see the note in components/ui/dialog.tsx referencing the
reverted #622–#624 global attempts).
Fix (artifacts/tx-os/src/components/notes/send-note-dialog.tsx):
- Track search-input focus directly — the most reliable "keyboard is
about to open" signal on touch devices, independent of visualViewport.
- Gate the override behind `matchMedia('(pointer: coarse)')` so it
only fires on touch-primary devices (iPad/iPhone/Android) and never
on desktop or trackpad laptops — addressing the architect's
cross-platform regression flag.
- When the gate is active, override DialogContent's Radix center
anchor (top:50%; translate(-50%,-50%)) with `top:6vh;
translate(-50%,0); maxHeight:50dvh`. The dialog grows downward from
a fixed top edge so it always fits in the upper half of the screen,
leaving the lower half free for the keyboard.
- Cap `listMaxHeight` to ~50% of innerHeight in the same gated
condition, as a backstop where visualViewport stays full-height.
- Debounce blur by 150ms so a tap on a recipient row doesn't lose
the row mid-tap to a center-jump animation; clear the timeout on
unmount.
- Added 120ms ease transition on top/transform so the layout swap
doesn't snap.
Deviations: original plan used vv.offsetTop+vv.height/2 to reposition,
but that depends on the same unreliable visualViewport. Switched to
a focus + coarse-pointer driven approach which works on iPad PWA.
Verification: tsc clean. Architect re-review pending after the touch
gate was added (first review flagged the missing platform guard,
which this revision addresses directly).
Refactor notification sound playback to use per-bucket throttling, adjust socket warmup, and ensure push notifications are sent even if the client is considered connected.
Refactor notification sound playback to use per-bucket throttling, adjust socket warmup, and ensure push notifications are sent even if the client is considered connected.
Replit-Commit-Author: Agent
Replit-Commit-Session-Id: 77cfe984-2c65-4152-bb7a-0df28274fe66
Replit-Commit-Checkpoint-Type: full_checkpoint
Replit-Commit-Event-Id: 3dcee04b-6717-4c1f-a172-b1f9c2febbe0
Replit-Commit-Screenshot-Url: https://storage.googleapis.com/screenshot-production-us-central1/c3c252e4-c83d-40ca-9fff-99a3ea60701e/77cfe984-2c65-4152-bb7a-0df28274fe66/qrzw3bH
Replit-Helium-Checkpoint-Created: true
- /auth/logout: delete all push_subscriptions rows for the user before
destroying the session. Best-effort with logger.warn on failure so an
operator can correlate any leaked-ring report to a real DB error.
- home.tsx handleLogout: call pushSub.disable() before the logout
mutation, raced against a 1.5s timeout so a stalled service worker
cannot block sign-out.
- executive-meeting-scheduler: switch the eligibility filter from
denylist (ne cancelled + ne completed) to whitelist
(eq status='scheduled'). Postponed / rescheduled / future statuses
can no longer trigger false 5-minute reminders.
- docker-compose.yml: pin TZ=Asia/Riyadh on postgres, api, and web
services so the scheduler's naive date/time math matches the
operator's wall clock instead of UTC.
Gitea push failed with TLS error during this session — code committed
locally, needs manual `./scripts/publish-to-gitea.sh --push` retry
when the desktop-11cj93j tunnel recovers.
- /auth/logout: delete all push_subscriptions rows for the user before
destroying the session. Best-effort with logger.warn on failure so an
operator can correlate any leaked-ring report to a real DB error.
- home.tsx handleLogout: call pushSub.disable() before the logout
mutation, raced against a 1.5s timeout so a stalled service worker
cannot block sign-out.
- executive-meeting-scheduler: switch the eligibility filter from
denylist (ne cancelled + ne completed) to whitelist
(eq status='scheduled'). Postponed / rescheduled / future statuses
can no longer trigger false 5-minute reminders.
- docker-compose.yml: pin TZ=Asia/Riyadh on postgres, api, and web
services so the scheduler's naive date/time math matches the
operator's wall clock instead of UTC.
Gitea push failed with TLS error during this session — code committed
locally, needs manual `./scripts/publish-to-gitea.sh --push` retry
when the desktop-11cj93j tunnel recovers.
Root causes the prior attempts missed:
1) Safari iOS/iPadOS ignores `interactive-widget=resizes-content`,
so `100dvh` never shrinks when the soft keyboard opens. Every
previous override that used `dvh` to size the dialog was sized
to the full screen and ran behind the keyboard.
2) The CSS rule from #622 (`[role="dialog"]:has(:focus)`) was
unconditional. On desktop, focusing any input inside a Radix
DialogContent instantly stretched it to width:100vw + top:0,
which collided with Radix's open-state slide-animation
transforms and rendered the dialog as a broken narrow strip
pinned in the top-left corner — visible in Tahani's laptop
screenshot when opening "Add Meeting".
3) `AdminFormDialog` puts `role="dialog"` on a NON-fixed inner
card, so the rule's `top/left/right` had no effect there
anyway; the card stayed centred in a layout-viewport-sized
wrapper which iOS does not shrink.
Fix:
- New `useVisualViewportVars` hook mounted in
NotificationsSocketBridge. Writes `--vv-height` and
`--vv-offset-top` on `<html>` from `window.visualViewport`,
updated on resize / scroll / orientationchange. Falls back
to `innerHeight`/0 if visualViewport is absent.
- Rewrote the index.css dialog rule:
* Gated inside `@media (any-pointer: coarse)` so desktop is
completely unaffected.
* Uses `var(--vv-height, 100vh)` and `var(--vv-offset-top, 0px)`
instead of `100dvh` / `0`, so dialogs actually track the
visible band on iOS.
- New `.dialog-vv-wrapper` helper class that applies the same
vv-sized top/height to bespoke full-screen overlay wrappers.
- Applied `.dialog-vv-wrapper` to AdminFormDialog and the admin
ConfirmDialog wrappers; switched their `fixed inset-0` to
`fixed inset-x-0 top-0 h-screen` and their inner cards from
`max-h-[90vh]`/`max-h-[85vh]` to `max-h-full` so the card
shrinks with the wrapper.
Kept `useScrollFocusedDialogField` unchanged — works correctly
once the dialog itself is sized to the visible band.
`tsc --noEmit` clean.
Replit Agent