Queue-flake anchor: src/data-nested-object-door.test.ts #27037

Activity

1. github-actions

   added

   finding

   queue-flake-anchor

   on Oct 28, 2026
2. os-zhuang

   added a commit that references this issue on Oct 28, 2026

   test(rest): give the nested-relation cap case a 30 s budget (ported f…

   ...

   6674ba8
3. objectstack-fleet commented on Oct 28, 2026

   

   objectstack-fleetBot

   on Oct 28, 2026

   Contributor

   More actions

   #27020 carries a budget for this case as a ported fix. Read 2026-10-28T02:55Z.

   Commit 6674ba8e09 gives "the cap" case in packages/rest/src/data-nested-object-door.test.ts an explicit 30_000 ms timeout. The case seeds 1,000 rows and deletes 1,001: 2.3 s alone, past the default 5 s on a loaded full-suite queue shard. A one-line comment in the test cites this issue and #27004. It is budget only: the slowdown's cause (#27004, a 1,001-row predicate delete about 18x slower since #26930) is not addressed. PRs queued behind #27020 inherit the budget once it lands.

   ---

   Generated by Claude Code
4. objectstack-fleet commented on Oct 28, 2026

   

   objectstack-fleetBot

   on Oct 28, 2026

   Contributor

   More actions

   Triage: p1 · domain:server · pm:blocked on #27004 · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-28T02:57Z

   ⛔ Not a claim.

   • Read fresh. A generated queue-flake anchor for packages/rest/src/data-nested-object-door.test.ts: three PRs ejected, two independent (refactor(driver-sql): move the declared column DEFAULT emission out of sql-driver.ts into sql-driver-schema-sync.ts (#25521 cut 19) #26960, feat(spec): PLATFORM_TOOLS_BY_PACKAGE['service-ai'] lists read_package_doc and search_package_docs (#27002) #27011).
   • Cause known: objectql: since #26930 a predicate delete walks every matched row one by one; 1,001 rows cost ~1.7 s (18×), and rest's nested-object-door cap test now nears its 5 s budget in the merge queue #27004 (p1, domain:server). Since fix(objectql)!: a predicate delete honours deleteBehavior for every matched row: restrict refuses, master_detail cascades, set_null clears (#26871) #26930, a predicate delete walks every matched row one by one, so this test's 1,001-row cap case went from about 0.1 s to about 1.7 s and crosses the 5 s budget under queue load. It is not a flake; it is a measured cost regression.
   • Blocked-by: objectql: since #26930 a predicate delete walks every matched row one by one; 1,001 rows cost ~1.7 s (18×), and rest's nested-object-door cap test now nears its 5 s budget in the merge queue #27004. This anchor closes when objectql: since #26930 a predicate delete walks every matched row one by one; 1,001 rows cost ~1.7 s (18×), and rest's nested-object-door cap test now nears its 5 s budget in the merge queue #27004's fix lands and the queue stops ejecting on this file.
   • ⛔ The fix never raises the timeout, skips the case or shrinks its rows.
   • finding removed.

   ---

   Generated by Claude Code
5. objectstack-fleet

   added

   area:recordsBusiness objects, records, the views that show data, usable forms, searchBusiness objects, records, the views that show data, usable forms, search

   domain:serverDomain lane — seat card indexed by label:pm:seatDomain lane — seat card indexed by label:pm:seat

   pm:blocked

   priority:p1High: required for production / M2High: required for production / M2

   priority:p0Critical: blocker, must ship before MVPCritical: blocker, must ship before MVP

   and removed

   finding

   priority:p1High: required for production / M2High: required for production / M2

   on Oct 28, 2026
6. objectstack-fleet commented on Oct 28, 2026

   

   objectstack-fleetBot

   on Oct 28, 2026

   Contributor

   More actions

   Triage: p1 → p0 · merge-queue red anchor · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-28T03:05Z

   ⛔ Not a claim.

   • Correction. Triage graded this p1 in R720. Under main-red convention ①, a full red in the merge queue is a p0 anchor, the same as main. It has ejected three PRs (refactor(driver-sql): move the declared column DEFAULT emission out of sql-driver.ts into sql-driver-schema-sync.ts (#25521 cut 19) #26960, feat(spec): PLATFORM_TOOLS_BY_PACKAGE['service-ai'] lists read_package_doc and search_package_docs (#27002) #27011, fix(spec): re-anchor the authorable-surface anchor so an offline, tagless spec build passes (#27015) #27020) on one check, and the [maintainer] refactor(driver-sql): split src/sql-driver.ts (22,096 lines, +2,533 in 30 days) along its SqlDriver seams, one move-only slice per PR #25521 split series is stalled. The maintainer's question that caught it: 「main 红为什么不是 p0?」.
   • This card is the anchor for that check (convention ②). The fix is root-cause card objectql: since #26930 a predicate delete walks every matched row one by one; 1,001 rows cost ~1.7 s (18×), and rest's nested-object-door cap test now nears its 5 s budget in the merge queue #27004, which is also raised to p0.
   • Stop-bleed choice: reverting the introducing commit would undo fix(objectql)!: a predicate delete honours deleteBehavior for every matched row: restrict refuses, master_detail cascades, set_null clears (#26871) #26930 (objectql: a predicate delete (multi: true with a filter) skips the referential walk, so a flow's bulk delete_record bypasses restrict, master_detail cascade and set_null #26871's referential walk, a p1 correctness fix for restrict / cascade / set_null). That trades a timeout for a data-integrity hole, so revert is ⛔ not the stop-bleed here. Fix forward in objectql: since #26930 a predicate delete walks every matched row one by one; 1,001 rows cost ~1.7 s (18×), and rest's nested-object-door cap test now nears its 5 s budget in the merge queue #27004 (walk the matched ids as sets).
   • ⛔ Never raise the 5 s budget, skip the case or shrink its 1,001 rows.

   ---

   Generated by Claude Code
7. os-warren

   added a commit that references this issue on Oct 28, 2026

   test(rest): give the nested-relation cap case a 30 s budget (ported f…

   ...

   ef55ac5
8. objectstack-fleet commented on Oct 28, 2026

   

   objectstack-fleetBot

   on Oct 28, 2026

   Contributor

   More actions

   Triage: closing as completed: the queue has recovered · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-28T05:27Z

   ⛔ Not a claim.

   • Root-cause card objectql: since #26930 a predicate delete walks every matched row one by one; 1,001 rows cost ~1.7 s (18×), and rest's nested-object-door cap test now nears its 5 s budget in the merge queue #27004 is closed. Its fix, PR fix(objectql): a predicate delete walks the rows it matches as one set, one probe per relation per 500 rows (#27004) #27073 (a predicate delete walks its matched rows as one set), merged at 05:03Z.
   • Since then, six merge-group builds have completed success (05:01–05:03Z: fix(app-shell): the permission matrix's object rows show exactly what the runtime enforces for every grant shape #27051, fix(app-shell): a screen-flow result screen draws Done, and closing it completes the run #27046, docs(qa): wave-2 checklist-accuracy close-out, slice B — six areas #27009, docs(qa): wave-2 checklist-accuracy close-out, slice A — seven areas and three RUNNER facts #26969, fix(runtime): an action or job body refused for a missing capability answers the same coded 403 a hook body does, not a bare 500 #27045, feat(security)!: resolve current_user.queues kernel-side and reserve the queues membership key #27044), with five more in progress and no ejection. Main's push run at fbc16563 (05:03Z) is green.
   • That is this anchor's close condition. If the merge-queue-triage workflow sees another ejection on data-nested-object-door.test.ts, it reopens or refreshes the anchor.
   • 4-hourly full run: red on main (CI) #27076 (main's 04:04Z full-run red, the same test) stays open until the 08:05Z 4-hourly run reads the fix. A green run there closes it.

   ---

   Generated by Claude Code
9. objectstack-fleet

   closed this as completedon Oct 28, 2026