Skip to content

Phone app for 1.4.0: fixes, sync dating, Material restyle, Report a problem - #137

Merged
j0nsh1n merged 48 commits into
mainfrom
fix/phone-1.4.0
Nov 20, 2026
Merged

j0nsh1n merged 48 commits into
mainfrom
fix/phone-1.4.0

Conversation

@j0nsh1n

@j0nsh1n j0nsh1n commented Nov 18, 2026

Copy link
Copy Markdown
Owner

The phone app work Jonathan asked for before 1.4.0 (inspection 2026-11-17).

  • 20 phone fixes: a tick marks one session (not the whole homework), a block ending at midnight no longer crashes Today, ending focus early keeps the minutes, Running late names the homework and its new end, refusals are shown instead of doing nothing, sync messages say what is wrong, permissions are read from Android, Week refreshes after sync, Back and rotation keep state, engine calls off the main thread.
  • Sync: a phone edit is stamped when it is made, not at the next sync (an older offline edit could undo a newer computer change or bring deleted homework back); records set aside by either side are retried on every sync, and the phone says how many are waiting.
  • Restyle, layout B "Agenda list" (Jonathan's pick): the computer's Material 3 Blue scheme, real icons, Today as an agenda with an Up next card, one-page Add homework, Week with session times and other weeks, Settings in groups.
  • Report a problem on the phone and the computer (both Settings designs), opening Jonathan's form with version and device filled in; hidden until the form link is set.

Review: Opus 5.5 review 1 FIX-PASS on 8f9b19f; its fixes come as further commits before this leaves draft.
Local: phone unit tests (326), cargo test -p flexweek-ffi -p flexweek-sync, ruff/mypy and backend tests passed; the full desktop suite runs here in CI.

🤖 Generated with Claude Code

j0nsh1n and others added 30 commits November 17, 2026 16:44
RealFlexWeekCore sent a homework row's tick to set_homework_done, which finishes the homework and every session in every week. RowTicks now sends every tick to set_block_done for that row's day; the widget tick goes through the same call. Verified: RowTicksTest green; breaking the day makes it red.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
RustPhoneJson.time used LocalTime.parse, which rejects the engine's 24:00; TimeLabel read it as 12 PM. Both now treat hour 24 as midnight; Running late's clock label uses TimeLabel. Verified: TimeLabelTest and RustPhoneJsonTest green; with hour % 24 removed the Today test fails.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The phone's End session threw the time away (focus::finish never credits) while the dialog promised 'You have done N min'. focusFinish now credits FocusJson.earlyCredit, the same whole minutes the dialog shows and the desktop's end_focus_early credits. Verified: EarlyCreditTest green; with the credit disabled it fails.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
open() called focusRestore without a handler; the core refuses a timer whose homework is gone, so startup showed an error text at every launch. FocusRestore.keep drops the saved timer on that refusal. Verified: FocusRestoreTest green; catching the wrong exception type fails it.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… Start focus or Sync now cannot act

The core's refusals were dropped: Plan opened 'Placed 0 of 0' with a dead Accept, Accept and Move it failed silently, Start focus ignored ok:false, and Sync now threw the answer away. result() no longer forces ok:true over a refusal, ffi errors lose their 'details=' prefix, and each screen shows the message in its existing sheet or line. Verified: RefusedAnswersTest and PlanRefusalTest green; with each fix removed they fail.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ach move says from and to

The engine's current block was the inserted 'Running late' block when nothing was on now, and its end was the late block's end. current_late_block now picks the session on now, else the next one today, and ends at where it moved to; the phone writes each move as 'from → to'. Verified: cargo test -p flexweek-ffi running_late (9 ok), fmt, clippy; RunningLateRealCoreTest green; with the next-session fallback or the moved end removed the Rust tests fail.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Step 2 and the step crumb said FlexWeek picks the time, but Save leaves the homework in No time yet until Plan runs. Wording only (AddHomeworkWords); Save is unchanged. Verified: AddHomeworkWordsTest green; with the old sentence it fails.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The top line and the Devices row came from different places, so a phone with no linked computer showed 'Paired · not reachable now' under 'No computer linked', and any reachable computer read 'same Wi-Fi'. phoneSettings now carries paired and last_sync from the sync answer; DeviceWords words the row. Verified: DeviceWordsTest and RemindersViewModelTest green; with the old words it fails.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…other sync errors

Every sync failure read 'isn't reachable' and every pairing failure 'That code didn't match'. SyncWords maps the engine's stable error kind to words for Sync now, the status line (the last kind is kept on the phone) and the pairing screen. Verified: SyncWordsTest, PairViewModelTest, RustPhoneJsonTest green; with the pairing default back to 'didn't match' SyncWordsTest fails.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The foreground sync loop refreshed Today, Settings and the widget but not the Week tab, so computer edits stayed missing and a new Monday never showed. SyncRefresh now refreshes all four. Verified: SyncRefreshTest green; without the week call it fails.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A tapped day stayed selected for the life of the process, so Today showed yesterday the next morning, and the jump-to-today check compared weekday numbers so the same weekday of another week hid it. DaySelection drops a chosen day when the real date changes, and the answer carries selected_is_today computed from dates. Verified: DaySelectionTest, RustPhoneJsonTest, TodayViewModelTest green; with either fix removed DaySelectionTest fails.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…sion

The banner came from a saved flag, so it came back after every cold start once granted and never again after Not now; Exact reminders saved 'on' without checking canScheduleExactAlarms and nothing replanned when it was granted. SystemPermissions asks Android, the banner re-offers after 3 days, Allow opens the system switch when asking is exhausted, and ExactAlarmPermissionReceiver replans on the permission-changed broadcast. Verified: PermissionStateTest, RemindersViewModelTest, ManifestReceiversTest green; removing the re-offer or the permission check fails PermissionStateTest. Not verified on a device.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…n thread

The planner (up to 150 ms a call) and the Week read ran on the UI thread and could wait on a sync's database lock. Each ViewModel now takes a worker dispatcher (IO by default) and applies results on the main thread; a Running late answer that arrives after a newer slider step is dropped; a second Save while one runs is ignored. Verified: ui.v2 and TodayViewModel tests green with a MainThreadRule; ThreadingTest fails when the Running late call stays on the main thread.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
DueChips.forToday appended a 'Pick a day' chip that took a slot when tomorrow and Friday/Monday coincided, and it silently set the due date to Monday; the real Pick a day button is below the chips. The chip is gone. Verified: DueChipsTest red before the fix (new Thursday/Sunday test), green after.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Back left the app from anywhere, and a rotation reset the tab, closed sheets, dropped a half-typed Add and rebuilt paired from the start-up value. Added a BackHandler driven by BackSteps, saved tab/sheet/paired state, a keyed AddHomeworkViewModel with a rebound callback, and FlexWeekApp.setPaired so Unpair and Pair keep the start-up answer in step. Verified: BackStepsTest green and red when GoToToday is removed; the rotation and Unpair parts are Compose state, not reachable from a JVM test: lead to check on the device.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…mer is never credited in full

TodayViewModel read the saved timer once, so the card kept running after Pause from the notification and at zero credited the whole session. refresh() now re-reads the timer with the day, and FocusGuard refuses to finish a paused timer (both the card and the end alarm). Verified: FocusExternalChangeTest red before the refresh change, green after; FocusAdvanceTest and TodayViewModelTest green; disabling the guard fails the guard test.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…un as a foreground service

F13: a local edit's one-shot sync used REPLACE and cancelled the running sync (and set the shared cancelled error); it is KEEP now. F5: SyncWorker gave a whole sync 10 s against the desktop's 45 s; the foreground loop stays at 10 s but pauses 5/10/20/40/60 s while syncs fail (SyncBackoff). F4: SyncWorker.getForegroundInfo uses the existing sync channel, with the FOREGROUND_SERVICE permissions and the dataSync service type WorkManager needs. Verified: SyncSchedulingTest, ForegroundSync tests, whole testDebugUnitTest and assembleDebug green; with each change reverted the tests fail. Android 8-11 behaviour not run on a device.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
week_rows counted every homework session, finished ones included, so a day with 4 h done and 1 h left said heavy. The engine's occurrence 'done' flag now keeps finished sessions out of minutes and heavy (they are still named); the phone reads the engine's heavy answer instead of re-applying its own threshold. Verified: cargo test -p flexweek-ffi week_rows (6 ok), fmt, clippy; WeekRowsHeavyTest green; with either change reverted the matching test fails.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… keeps no Activity

UpdateViewModel stored lastChecked in a finally block, so one offline check blocked the daily check for 24 h; it is now stored only when the check got an answer (the old test asserted the opposite on purpose and was rewritten for B19). PackageUpdateInstaller kept the Activity it was built with inside a ViewModel that outlives it; it now keeps the application context. Verified: UpdateViewModelTest and UpdateInstallerContextTest green; the failed-check test was red before the change and the context test fails when MainActivity passes the Activity again.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The Sound row had no control and its note about alarms did not match the app; the sample said 'starts in N min / then free until 10 PM' though a real reminder is the block title with 'start · length'. The row is removed (no engine setter exists to give it a control) and the sample uses RustPhoneJson.reminderBody. Verified: ReminderSampleTest green; with the old sample text it fails.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ns and a normal bottom bar

Full Material 3 Blue light and dark scheme from m3-tokens.json, block colours school blue / homework amber / activities green (widget palette and its test moved with them), edge to edge with matching status and navigation bar icons, window themes for light and dark, Material Symbols Rounded vector icons, NavigationBar at its default height with filled icons for the selected tab, extended Add button. Verified: gradlew testDebugUnitTest assembleDebug green (313 tests).

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Week strip with a filled selected circle and a due dot that does not move the number; Up next as a primaryContainer card with a full-width Start focus and Done / Running late; Later today, Earlier (done struck through), No time yet with a tonal Plan button and the waiting homework in a tinted group; time column split into clock and AM/PM. TodayLayout decides the groups and whether the two buttons stack: at 360 dp and font scale 1.3 they stack full width so no label is cut. Verified: TodayLayoutTest (6 new; broke the width estimate on purpose, the 1.3 test went red, restored), gradlew testDebugUnitTest assembleDebug green.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Close X, Save in the top bar (on only with a title, a due day and a length), outlined What is it? field focused on open, Due chips plus Pick a day, wrapping How long chips with Other taking minutes, Sessions as a segmented button, recent titles as small chips and the Plan line. The three steps, their crumbs and Back/Next are gone from the view model. Kept: no bogus Pick a day chip, the draft survives rotation in the keyed view model, Back closes it. Tests updated because they walked the old steps: AddHomeworkViewModelTest (steps replaced by save rule, Other length, one save), RefusedAnswersTest (calls save instead of three Next), AddHomeworkWordsTest (the removed crumb). Save rule broken on purpose (length ignored), red, restored. gradlew testDebugUnitTest assembleDebug green.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…h, and arrows for other weeks

The engine's week_rows now also lists each day's sessions (start, length, title, category, done; a due line has no start). The phone shows the title with previous and next arrows, the planned and done line, each day as a heading with its total and its sessions, and free days collapsed into one free line; tapping a day opens Today on that day for the current week only, since Today shows only the current week. Fixture regenerated. Verified: cargo test -p flexweek-ffi 101 passed, fmt and clippy clean; Rust session test and Kotlin arrow test each broken on purpose and red, restored; gradlew testDebugUnitTest assembleDebug green.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…port a problem

Reminders switch follows what Android says now, Remind me before homework is a segmented 5/10/15, On-time reminders keeps the permission logic, the Your computer card shows the name, a state dot and line from the single sync answer, a tonal Sync now with its result under it and Unpair (Pair again when no computer is linked, since it is still the only way back to the pairing screen), Help opens the online form from FEEDBACK_FORM_URL with the app version and the device filled in, and Updates is compact with its existing controls. Verified: FeedbackFormTest (fill broken on purpose, 2 red, restored), PermissionStateTest additions, gradlew testDebugUnitTest assembleDebug green.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…shows each time once

Plan result, Running late and the Plan error sheet are full width on surfaceContainerLow with insets and the keyboard respected, text button then filled button on the right; the focus End question and Replace dialogs make the safe choice the filled button. The Plan result line no longer says the time twice (it printed a raw 16:00 beside Mon 4 PM; the detail is now the length). Verified: new RustPhoneJsonTest case red with the old detail, green after; gradlew testDebugUnitTest assembleDebug green.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…dd button removed; changelog line

Buttons that had a fixed 48 dp height now have a minimum, so a label that needs two lines at font scale 1.3 is not cut; the snackbar sits above the Add button instead of 76 dp from a bottom edge that is now above the bar; the unused old Add button path in Today is gone; CHANGELOG Changed line for the phone's new look. Verified: gradlew testDebugUnitTest assembleDebug green (327 tests), cargo test -p flexweek-ffi 101 passed, fmt and clippy clean.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ext sync

The phone only scanned (stamped) its edits inside a sync session that had reached the computer, so an edit
made offline at 10:00 and synced at 18:10 was stamped 18:10 and beat newer computer edits (an old due date
came back, a homework deleted on the computer came back). Every phone write now ends in
FlexWeek::local_change, which stamps the account's records with flexweek_sync::scan before telling the app.
This also makes the comment in engine/sync/src/net/mod.rs ("the phone scans after its own writes") true.

Verified: focus::tests::an_offline_phone_edit_keeps_the_time_it_was_made red before (stamp 8 h late),
green after; disabling the stamp makes it red again; cargo test -p flexweek-ffi 100 passed; fmt and clippy clean.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…c, and the phone says so

A record the receiver rejected (an older app that does not know a new field, or a sender clock more than a
day ahead) went to sync_quarantine and was never read again; the sender's cursor had moved past it, so it
was lost while the phone said "Synced just now". flexweek_sync::retry_quarantine re-applies held records at
the start of every exchange (scan_and_export, both sides): one that now passes applies, one a newer edit
replaced is dropped by the usual newest-wins rule, one still bad is held again, an unreadable row goes. The
phone's status carries held_back and its synced line says how many changes are waiting and what to do.

Verified: merge::held_back_records_apply_on_a_later_try_or_go_when_replaced (red with the retry disabled);
sync_api a_phone_pairs... now checks held_back 0 then 1 with the notice (red with the count forced to 0);
cargo test -p flexweek-sync (50 + 36 + 1) and -p flexweek-ffi (100) green; fmt and clippy clean.
Not done here: the desktop does not show its own held-back count yet.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…he version filled in

desktop/native/feedback.py holds the form's pre-filled link (made with the sample answers 1.4.0 and
"test") and swaps those answers for this app's version and this computer. The Widgets Settings gets a
"Report a problem…" button under Account, setup and updates; Material Settings > Computer gets a
Report a problem row. Both are shown only when a form link is set, so no dead control ships; the
link itself comes from Jonathan.

Verified: desktop/tests/settings/test_report_problem.py (4) green, red with the answer swap or the
hidden rule broken; test_get_developer_builds_switch.py and test_settings_page.py green; ruff and mypy
clean on the new module.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
j0nsh1n and others added 2 commits November 17, 2026 17:27
…es until it is set

The restyle left a placeholder address (FORM_ID, entry.VERSION_ID) that would open a dead page. The phone
now keeps the form's pre-filled link exactly as the form gives it (sample answers 1.4.0 and "test", the same
link as the computer's desktop/native/feedback.py) and swaps those answers for the app version and the phone;
Settings shows the Report a problem row only when a link is set.

Verified: FeedbackFormTest (whole-URL compare, no-link rule) green; testDebugUnitTest 326 green.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…he tabs

Layout B puts Add on Today and Week only (on Settings it covered Sync now), and the full-screen Add
homework form hides the bottom tabs. MainTabsTest's old "Add on every tab" expectation (from the 1.1
pick 1 A) changes to the picked layout B.

Verified: MainTabsTest red with the old rule, green with the new; testDebugUnitTest 326 green; emulator.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@coderabbitai

coderabbitai Bot commented Nov 18, 2026 •

Copy link
Copy Markdown

Important

  • Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

Run configuration

  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: defaults
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 023240eb-d19d-4573-b1aa-8bcc6da3b47a
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

@j0nsh1n j0nsh1n mentioned this pull request Nov 18, 2026
j0nsh1n and others added 14 commits November 18, 2026 01:23
…ock ends

The phone's "stays" list counted from where the moved homework ends, so a fixed
block between the late block and the moved session was left out although it did not
move. It now counts from the end of the late block.

Verified: new test a_fixed_block_between_the_late_block_and_the_moved_homework_still_stays
fails with the old end-of-moved-session rule (broken on purpose, red) and passes with the
fix; cargo test -p flexweek-ffi green. The 'Now ends' day label for a session moved to
another day is not done (see review1-fixes REPORT).

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Records set aside for "account identity differs" or a refused write came back on every
sync, so the phone said "update FlexWeek and check its date and time" for ever and the
rows were deleted and re-inserted each round. Those rows are now left alone and not
counted; write refusals are tagged "could not be saved:" so they can be told apart.

Verified: new test records_no_retry_can_fix_are_neither_counted_nor_churned went red with
the filter removed and green with it; cargo test -p flexweek-ffi -p flexweek-sync green.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…he Add draft is emptied when the form closes

Other = 20 (or 999) kept Save on and the engine then refused with its raw text under the
title. Save now needs a multiple of 15 up to a day, and the form says "Use 15-minute
steps, for example 45 or 75." The Add form now keeps one draft view model that reset()
empties on close, instead of leaving one view model per Add opened (the MainActivity side
of this is in the next commit).

Verified: otherLengthsOffTheFifteenMinuteStepsKeepSaveOff and resetEmptiesTheDraft... went
red with validLength / reset() broken on purpose, green when restored; full gradle
testDebugUnitTest green.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…undo or Week load shows the engine's message; ticks say session done; one Add draft; dismiss icon

F3: a refused early credit (its homework deleted on the computer) no longer leaves the
saved timer for ever; the timer is removed and finished, then the reason is shown.
F7: Today's tick and undo and the Week load catch the engine's refusal and show it
instead of crashing. F11: "Session marked done", and the tick says
"Mark <title> session done". F14: the snackbar dismiss uses ic_close with a Dismiss
description; MainActivity empties the single Add draft when the form closes.

Verified: FocusEndTest, RefusedTicksTest and WeekNavigationTest's refused-week test went
red with the catches / wording broken on purpose, green when restored; full gradle
testDebugUnitTest green. The dismiss icon is checked only by compiling.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
A computer-side handler error (busy database, no progress) also arrives as a protocol
error and said "needs an update". Only "version_mismatch" and the unsupported protocol
version message now do; other protocol errors get the general message.

Verified: onlyAVersionMismatchSaysToUpdate red with the old kinds restored, green with the fix.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…c one

A failed check stored no time, so every resume hit GitHub twice while offline or
rate-limited. A separate last-failed-attempt time now holds automatic checks for an hour;
the Check now button always runs.

Verified: aFailedCheckWaitsAnHourBeforeTheNextAutomaticOne red with the wait removed, green with it.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…e the app

The widget now reads the app's categoryKind and the amber constants are named HOMEWORK_*.

Verified: everyCategoryNameTheAppColoursGreenIsGreenOnTheWidgetToo red with the old
"sports"-only match, green with the fix.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Verified: noPairedComputerSaysSoOnTheRowToo red with the old text, green with the fix.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
On-time reminders, the Your computer card, the hour wait after a failed update check,
the retry on both devices and the held-back count. The Report a problem lines are left
as they are: they stay untrue until the form link (F2) is set.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…generation; apply and delete are one transaction

scan_and_export now returns the retry's ApplyReport; both session paths add it to
applied/quarantined and the generation, so the desktop names the phone when held
records land. retry_quarantine_report deletes the retried rows and applies in one
immediate transaction (delete first), so a crash cannot double the held count.
Tests: net_sync (desktop and phone retry), merge (failed delete applies nothing).
Verified red: dropping retried.applied fails both net_sync tests; apply-then-delete
order fails the merge test; restored green (sync suite 52+38+1 pass).

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The running-late answer's current gains end_day (weekday index, null when the session
stayed on today). The phone says 'Now ends at 7 PM' as before, or 'Now ends Thu 7 PM'
for another day (TimeLabel.nowEnds; the phone has one 12-hour style, no 24-hour setting).
Phone fixture rewritten (end_day: null). Tests: ffi current_late_block (moved vs same day),
TimeLabelTest, RunningLateRealCoreTest parse. Verified red: ignoring end_day in the engine
filter fails the ffi test; ignoring it in the label/parse fails both Kotlin tests; restored green.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…merging main

Merging main (where Unreleased became 1.4.0) put this branch's 24 lines inside
the released 1.4.0 section. They are moved back under Unreleased with their
Added/Changed/Fixed headings; the 1.4.0 section is identical to main's again.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… aside

Lead review of the review-1 fixes (F8, 5a38975): each exchange added the
retry's re-held records to the summary's and the answer's 'quarantined', so
a record still a day ahead counted as newly set aside on every sync. Only
records set aside by this exchange count now; the held-back notice already
counts the stored rows.

Verified: new net_sync test fails with the old counting (1 failed), passes
restored; cargo fmt --check, clippy -D warnings and the ffi + sync tests,
196 passed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@j0nsh1n
j0nsh1n marked this pull request as ready for review November 20, 2026 16:07
j0nsh1n and others added 2 commits November 20, 2026 08:10
CI's select_map check (verify-engine) failed: desktop/native/feedback.py, added for the phone's bug
form, was in no tools/fwtest/areas.toml rule. Its tests are desktop/tests/settings/test_report_problem.py,
so it joins the settings rule. The desktop jobs on that run were cancelled after it, and mutate (4)
failed installing clippy on the runner (not the code).

Verified: cargo test --test select_map in tools/fwtest, 19 passed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Conflicts:
- desktop/native/settings.py: main removed the Widgets settings page, so its
  Report a problem button goes with it; main's file is kept.
- bridge_settings.py: both import lists kept (feedback plus main's new ones).
- tools/fwtest/areas.toml: main's shorter list plus desktop/native/feedback.py.
- CHANGELOG.md: the phone entries go into main's Added/Changed/Fixed lists;
  Report a problem now says Settings > Computer only.
- test_report_problem.py: the Widgets settings page checks are removed; the
  Material checks stay.

Verified: test_report_problem.py and test_get_developer_builds_switch.py,
8 passed; ruff check on the touched files passed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@j0nsh1n
j0nsh1n merged commit f175f3c into main Nov 20, 2026
33 checks passed
j0nsh1n added a commit that referenced this pull request Nov 20, 2026
CHANGELOG conflict: both Added lists kept (the #118 entries and Report a problem).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant