mirror of
https://github.com/block/buzz.git
synced 2026-11-08 02:11:36 +00:00
6cc43c9fd4
Follow-up to 885804282 from its review. That commit added a class-level
`BuzzTestHostSettleBarrier` that measured a user-initiated-queue-then-main
hop in `setUp` and let the class start once three consecutive hops were
fast. Re-reading both 2026-11-01 CI logs showed the barrier would have
opened immediately in both failing runs: in job 124905175903 the one-hop
test `testAllowedNotificationRestoreNeedsNoAppGroupStore` passed in
0.070 s at the start of the class, and in job 124910378184 the gate would
have opened in the 1-2 s lull after its 1.287 s hop. Both stalls began
mid-class and recurred mid-test, which a one-instant gate cannot cover,
and a sustained-hop version would be a CI-only wait. The barrier, its
doc comment with the measurement figures, and its "settled/gave up" print
are removed. The per-element QoS assertion from 885804282 and everything
from f0bf2c0c8 stay.
In its place, a `waitForCallback` helper in `BuzzPushSnapshotEnrichmentTests`
wraps the three callback waits in the two affected tests. It waits with
`XCTWaiter` for the unchanged 10 s `callbackTimeout` and returns at once on
success, costing nothing on a passing run. Only when the wait does not
complete does it time one fresh user-initiated-queue-then-main-queue hop,
the path every bridge reply follows, bounded at 2 s, and fail with the
expectation's description plus the measured hop ("a fresh queue->main hop
took 705.3 ms", "took 0.4 ms", or "did not finish within 2 s"). That lets a
starved host be told apart from a callback that never fired the next time
this appears in CI. The original assertions still run after the helper, so
the nil-versus-reason evidence is unchanged.
If this recurs, the remaining levers are the test host (RunnerTests boot
the full Flutter app as TEST_HOST, which is what saturated the 3-vCPU
runner) or the 10 s `callbackTimeout`.
Verification (Xcode 27.2, iOS 26.2 simulator runtime, 16-core host):
- Forced timeout, with temporary variants not committed: making the
bridge's retry deletion never call back produced "'retry failure stays
allowed' not fulfilled within 10 s; a fresh queue->main hop took 0.4 ms"
followed by the existing nil-versus-reason assertion failure; adding a
0.7 s sleep inside the hop produced "took 705.3 ms".
- Class in isolation with the final code: 10/10 green, 0.1-0.5 s each.
- Full RunnerTests bundle with CI's `-parallel-testing-enabled NO`: 2/2
green (101 tests, 24-29 s).
- The commit touches only this test file; the pinned Dart formatter was
first on PATH so the pre-commit hook reformatted nothing else. No
swiftformat or swiftlint configuration in the repo; `git diff --check`
is clean.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Signed-off-by: Matt Toohey <contact@matttoohey.com>