You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Nov 21, 2026. It is now read-only.
Repository navigation
This repository was archived by the owner on Nov 21, 2026. It is now read-only.
/next: mark the tpch10 top_parts refusal as the 26.10.1 per-query heap budget defect, fixed in 26.11.1 #272
On the 26.10.1 measurement, ArcadeDB's TPC-H SF10 top_parts query (about 2,000,000 GROUP BY groups) is refused at the engine's default settings in both ArcadeDB arms of the documents analytics table: 26.10.1's per-query heap budget scales with the core count and refuses it on 12 cores (upstream ArcadeData#9402). Upstream fixed it in ArcadeData#9415 for 26.11.1, and we verified the fix at the merge commit on Temurin 25 and Temurin 21 (the query passes at 1, 2, 4, and 8 cores).
The paper is expected to pin 26.11.1. On /next, show those cells as refused at default settings, with one plain sentence naming the cause and saying that 26.11.1 fixes it and the next measurement re-measures it, the way the ArangoDB 10M dense cell is declared (an export_web declared outcome with its sentence), rather than hiding or omitting the cell. The raised-budget override from #260 is not used for the headline.
Activity