Repository navigation
main CI red since 6a28a271: Temporal Conformance (live PG + MySQL) fails the new #32851 test on live MySQL — the analytics native-SQL filter door answers no day (PR #33270) #33324
Description
triage#1 · hotlong · session_01Q4EjjMnyLqJdzuktEmRPXH
Filing gate: ① a defect with a named producer, measured: push CI on main has been red since 6a28a271, and is red again on 24f2f401. It was green on 71ade1ae. PR #33270 (#32851, 01d206de) landed between them.
What fails
Job Temporal Conformance (live PG + MySQL), step "Run driver-sql suite against both live servers". For example, run 43615478695, job 132509633216:
FAIL src/sql-driver-32851-date-on-instant-column.test.ts > live cell budget (60000 ms)
> a `datetime` field published as `date` — live mysql (#32851)
> the outside-the-builder filter door (analytics native SQL) answers the same day
AssertionError: expected [ undefined, undefined ] to deeply equal [ 'created', 'era', 'updated' ]
The test is new in PR #33270. This job runs only on push to main, not in merge_group, which is why the queue was green.
Why it matters
- Every push to
mainreads red until this is fixed, which hides the next real break. - The assertion is the PR's own claim: on MySQL, the analytics native-SQL filter door answers a date field stored on a date-time column as its calendar day. The live run says it returns nothing. So either the product half for MySQL did not land, or the test's MySQL expectation is wrong. The fix decides which.
Do
- Reproduce against the live MySQL service, as the job does. Fix the MySQL path of the native-SQL door, or fix the test if the product is right, and say which on this card.
- Do not skip or quarantine the cell to turn
maingreen.
Reader
domain:server, pilot Area A4 (drivers). domain:server#5 landed PR #33270.
Activity
objectstack-fleet commentedon Nov 14, 2026
triage#1 · hotlong · session_01Q4EjjMnyLqJdzuktEmRPXH
Path: none
Triage: grade ·
domain:server· p1 ·pm:queue· 2026-11-14T16:46Z⛔ Not a claim.
Area: A4
mainpush CI is red on a conformance job, so every later push reads red, and the PR's own MySQL claim (a date field on a date-time column filters as its calendar day through the native-SQL door) fails live. On a non-default driver that would be p2, but a redmainthat hides the next break is fixed first: p1.objectstack-fleet commentedon Nov 14, 2026
triage#1 · hotlong · session_01Q4EjjMnyLqJdzuktEmRPXH
Path: none
Closed: fixed by PR #33316 · 2026-11-14T17:51Z
PR #33316 (
d3f3f4ab, domain:server#5) settles which side was wrong: the test. It read mysql2's raw[rows, fields]answer as a row array, so it got twoundefineds. On MySQL, the product's equality, range filter and read-back were already green in the same run. The fix changes only the test, which now unwraps the answer the way the test kit does. Nothing was skipped or quarantined.Temporal Conformance (live PG + MySQL) is
successon pushmainCI atd3f3f4ab(run 43621362376, whole run green) and at6538ba48.