PostgreSQL: Memory Safety and Query Optimization Fixes

Transcript

Good morning, it's June 3rd, 2026. Today's PostgreSQL activity centers on memory safety improvements and query optimizer fixes that address several edge cases affecting production stability.

The most significant pattern involves fixing memory management issues across multiple components. Commit a77bdb1 addresses several memory leaks in the xml2 extension where libxml2 objects weren't being properly released, particularly in XPath evaluations. These leaks would accumulate across SQL calls, potentially degrading performance over time. Separately, commit 273a3c7 fixes an uninitialized variable in hash record extended that could affect extension-provided hash functions, though core PostgreSQL functions remain unaffected.

Query optimization saw two important fixes for edge cases in aggregation handling. Commit ffeda04 resolves an issue where eager aggregation was incorrectly pushed down to relations on the inner side of semi-joins or anti-joins. This affected aggregates like count-star that reference no table columns, potentially causing incorrect results. Additionally, commits eb8e76e and d58ec50 improve overflow and underflow handling in regression functions regr_intercept and regr_r2, building on previous improvements to the correlation function.

Two fixes address specific operational issues. Commit d21604e resolves assertion failures in psql pipeline mode when deferred constraint violations occur during sync message processing. This was causing incorrect state tracking and potential crashes. Meanwhile, commit f2081a7 fixes a race condition in replication slot release for ephemeral slots, where dropped slot memory could be corrupted by subsequent updates.

The remaining fixes target maintenance tools. Commit 95b6ec5 ensures that vacuum database's analyze-in-stages option properly processes partitioned tables as documented, while commit 298bdd3 fixes an internal inconsistency in pg_dump's index handling.

These changes collectively improve PostgreSQL's reliability in production environments, particularly around memory management and query planning edge cases. Developers using XML processing, pipeline operations, or complex aggregation queries should see more stable behavior.

That's your PostgreSQL update for June 3rd.