mirror of
https://github.com/block/buzz.git
synced 2026-10-30 14:01:37 +00:00
42ff8cced0
The three Postgres fence tests in buzz-db asserted hidden_at equality against raw Utc::now() values, but timestamptz stores microsecond precision, so the nanosecond tail was truncated on round-trip and the equality assertions failed in CI. Truncate to microseconds up front. The e2e replay test (test_nipdv_message_replay_preserves_a_newer_rehide) posted two identical kind:41012 hides from B. A nostr event id is sha256 over second-resolution created_at, so when both hides land in the same wall-clock second they share an id and the relay dedupes the second one before re-running hide_dm -- B's re-hide silently vanished and the assertion saw an empty hidden set. Backdate the first hide to force distinct ids, matching the guard create_dm already uses. Both are test-authoring defects; the fence and resurface logic are unchanged. Co-authored-by: Mongo <9cfd347903944d5b85aa6c93d2ab67381b978a92a31914bca69998968752a1d7@buzz.block.builderlab.xyz> Signed-off-by: kenny lopez <klopez4212@gmail.com>