On Sat, Sep 19, 2026 at 4:22 AM Haibo Yan <[email protected]> wrote: > I found what looks like a regression from e28cf96e935 > ("Fix stale copies of PHVs in subqueries"). In particular, I think there is > still a way for an upper-level PHV containing an unprocessed SubLink to get > into a subquery after preprocess_subquery_phvs() has already run.
This is the same bug reported by Fujii-san earlier, which is fixed in my v1-0001-Fix-preprocessing-of-PHV-copies-pushed-down-into-.patch. - Richard
