fix(extension): re-arm reminder alarms after an update, and log reminder activity - #354
Merged
kYem merged 4 commits intoSep 20, 2026
Conversation
…der activity Chrome clears every chrome.alarms entry when an extension updates, and the service worker never re-armed them from storage, so each release silently killed every pending reminder until the user edited one. The worker now reconciles armed alarms against stored reminders on install/update and browser start, arming only the missing ones so a kept alarm cannot fire twice. A 20-entry reminderActivity ring buffer in chrome.storage.local records armed/fired/skipped/snoozed/done/reconciled/advanced from both realms, so a missed fire can be diagnosed after the fact. The reminder-lifecycle e2e drives one reminder through the real extension and prints that trace. Fixes ENG-118, ENG-124
…s once per start Review round 1. The activity log exists to diagnose a missed fire, yet every failure path wrote nothing to it: a notify, re-arm, snooze or button write that failed now records a 'failed' entry naming the step; the reconcile summary names the ids it could not arm and records its own failure; a read that fails no longer wipes the log, and a corrupt value is replaced. onInstalled and onStartup both fire on an update applied at launch, so the reconcile is deduped behind one in-flight promise. The page's load-time re-arm only moves the wakes it advanced, and the macOS launch re-arm warns again when one fails. The e2e specs share one harness, settle each launch's reconcile before seeding, and assert the reconcile record rather than the alarm's mere presence. Comments say what the code does, in two lines.
…g at start Review round 2. Chrome drops a one-shot alarm before dispatching it, so at start an overdue reminder's fire is exactly the alarm getAll() no longer lists; the reconcile re-armed it in the past and the user got it twice. The worker now tracks fires in flight and treats them as armed. The page's load-time advance decides which reminders moved inside the lock, not from the snapshot a pull may have overtaken, so it neither logs a false 'advanced' nor re-arms a stale date. A reconcile or advance that fails is a 'failed' entry like every other failure, and the fire path's lookup, persist and multi-failure cases are pinned.
…ncile's discriminator Review round 3. A second overdue reminder not in flight is re-armed while the firing one is left alone; the cancel-failure trace has its own test; one comment split, one docstring reworded.
kYem
deleted the
kes/eng-118-reminders-dont-fire-chromealarms-are-never-re-armed-after-an
branch
September 20, 2026 02:19
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes ENG-118, ENG-124
The bug
Chrome clears every
chrome.alarmsentry when an extension updates (documented: alarms "persist until the extension updates"), and the service worker never re-armed them from storage —ChromeScheduler.persistsAcrossRestarts = trueswitched the only re-arm loop off. Every release, and every crxjs dev reload, silently killed every pending reminder until the user happened to edit one.Reproduced against the real built extension:
e2e/reminder-alarms-survive-update.spec.tsseeds a reminder, arms it, stages a version bump, relaunches — red before, green after. (In Playwright's Chromium the alarm is gone after a plain relaunch too, with storage intact.)The fix
armMissingReminderAlarms()in the shared reminder service, run by the worker ononInstalled+onStartup, seeded fromchrome.alarms.getAll(). Ensure-exists, not recreate, so a Chrome that does keep an alarm can't fire it twice. Deliberately not at worker top level — that runs on every alarm wake and would re-create an alarm mid-fire. The macOS page-load re-arm now shares the same function.The activity log (ENG-124)
The logger is console-only and the worker's console dies with the idle worker, so a missed fire was undiagnosable.
recordReminderActivity()keeps the last 20 events inchrome.storage.local.reminderActivity—armed / cancelled / fired / skipped / toasted / done / snoozed / reconciled / advanced, tagged with the realm — under the collection lock, never throwing into the reminder path. Local-only, never synced.e2e/reminder-lifecycle.spec.tsdrives one reminder through the real UI — add → fire → snooze → staged update → done — asserts the trace in order, and prints it:Still open on ENG-118
Re-arm after a sync pull (the
reminder-store.tsknown gap), what an overdue-on-load reminder should do, and the page-toast + OS-notification double. Each is its own decision.Verification