Drew Gallardo 1bead6960c fix: pin mock clock in eventual consistency test (#3547) ...

This PR fixes a flaky test I hit on Windows test in #3528.

Looks like `test_eventual_consistency_background_refresh` was failing
with `v_cached` expected 1, got 2. There was a pr which swapped
`tokio::time::sleep(300ms)` for `clock::advance_by(300ms)`, which is
pretty much fine but the test necer pinned the clock so the first
`get()` locks the `cached_at` on wall time. Therefore, if our CI is
taking long enough the ttl expires before the value assertion in the
test.

So now we can add a `pin()` and call it first `get()`. After that we can
advance the clock manually with no problems.

Also, it's worth noting that I tried pinning in `BackgroundCache::new()`
first. That broke another test `test_reload_resets_consistency_timer`,
which uses real `tokio::time::sleep` and needs wall clock after
`clear_mock()`. So the pin stays in this test only. And this should
unblock us.

Failing instances:
-
https://github.com/lancedb/lancedb/actions/runs/27567527236/job/81495265474?pr=3528
-
https://github.com/lancedb/lancedb/actions/runs/27560366489/job/81470414928

2026-06-17 11:56:40 -07:00

..

lancedb

fix: pin mock clock in eventual consistency test (#3547)

2026-06-17 11:56:40 -07:00

license_header.txt

ci: check license headers (#2076)

2025-01-29 08:27:07 -08:00