fix(messaging): resolve getToken() when observed service worker becomes redundant

[chan-mai]

Discussion

Fixes #10197.

getToken() intermittently fails with:

Messaging: We are unable to register the default service worker. Service worker not registered after 10000 ms (messaging/failed-service-worker-registration).

even though the service worker registers and activates successfully.

Root cause

waitForRegistrationActive (added in #8661) captures a single worker
(registration.installing || registration.waiting) at call time and resolves
only when that specific worker fires statechange → 'activated'.

registerDefaultSw calls registration.update() immediately before awaiting
waitForRegistrationActive. When that update installs a replacement worker —
common right after a redeploy, when the SW script or one of its
importScripts dependencies changed — the originally observed worker
transitions to 'redundant' and never fires 'activated'. The promise then
hangs until the 10s timeout, even though the replacement worker activates
successfully in the meantime. The 'redundant' transition is never handled,
so the call can neither recover nor fail fast.

Fix

waitForRegistrationActive now:

1. resolves as soon as registration.active is set, regardless of which
   specific worker reached the 'activated' state;
2. follows replacement workers via the registration's updatefound event; and
3. re-attaches to the next incoming worker when the one it was watching
   becomes 'redundant'.

All listeners and the timeout are cleaned up once the promise settles.

This also removes the previous immediate 'No incoming service worker found.'
rejection: a worker can still arrive via updatefound after register()
resolves, so the wait now relies on the existing timeout for the genuine
"nothing ever activates" case.

No public API changes.

Testing

• All existing @firebase/messaging tests pass.
• Added packages/messaging/src/helpers/registerDefaultSw.test.ts covering:
  • resolves when the registration already has an active worker;
  • resolves when a replacement worker activates after the originally observed
    worker becomes redundant (the regression, which previously timed out);
  • resolves for a worker that only appears via updatefound;
  • rejects when no worker becomes active before the timeout.

API Changes

None

[changeset-bot]

🦋 Changeset detected

Latest commit: c9d81c2

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 3 packages

Name	Type
@firebase/messaging	Patch
firebase	Patch
@firebase/messaging-compat	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[gemini-code-assist]

Code Review

This pull request fixes an issue where getToken() intermittently fails with a service worker registration timeout. The waitForRegistrationActive function was refactored to resolve as soon as any active worker is present, follow replacement workers via the updatefound event, and transition to watching a new worker if the currently watched worker becomes redundant. Comprehensive unit tests were also added to verify these scenarios. I have no feedback to provide as there are no review comments.