Follow-up from #2034 review: cover configured busy timeout in SpeakerDatabaseSnapshotTests #2145

Activity

1. r3dbars

   added

   waiting-on-humanNeeds a human (Justin) to look before anything else happensNeeds a human (Justin) to look before anything else happens

   cleanupOne-time repo cleanupOne-time repo cleanup

   on Oct 8, 2026
2. r3dbars commented on Oct 17, 2026

   

   r3dbars

   on Oct 17, 2026

   OwnerAuthor

   More actions

   Already fixed on main by 78d7a18 ("Prove snapshot busy timeout with an injected clock", landed via #2173).

   SpeakerDatabaseSnapshot.create now takes injected now/wait closures, and testExclusiveLockFailsWithBusyAndPreservesSource passes busyTimeout: 0.05 with a fake monotonic clock. It asserts the simulated time ends at 0.05 (±1e-6), exactly 5 waits each in (0, 0.01], a .busy error, unchanged source bytes, and no destination file. No wall-clock timing.

   Checked on current origin/main:

   • swift test --filter SpeakerDatabaseSnapshotTests in Tools/TranscriptedCLI: 7 tests, 0 failures.
   • Mutation check: forcing the deadline to the fixed 5s cap (ignoring busyTimeout) makes the test fail (simulated time 5.0 vs 0.05, 501 waits vs 5). Reverted after.

   So the promise from the #2034 review thread is covered. Closing.
3. r3dbars

   closed this as completedon Oct 17, 2026