Last week in mysql-server development (35)

[henrikingo-oracle]

This weekly newsletter gives you a comprehensive overview of what is going on
in the MySQL open development project. It covers the week of August 24–30,
2026, ending Sunday.

Scope

• Repositories: mysql/mysql-community and mysql/mysql-server
• Date range: August 24 through August 30, 2026, inclusive (America/Chicago)
• Coverage: 17 pull requests, 9 Issues, and 2 Discussions retrieved
• Pull requests merged: 2 of 17; both are closed-unmerged pull requests included because their current labels exactly match Integrate case-insensitively.

Repository	Pull requests	Issues	Discussions
mysql/mysql-community	0	7	2
mysql/mysql-server	17	2	0
Total	17	9	2

Executive Summary

0. [Discussion]: Support DuckDB as a columnar storage engine being added to MySQL … 79 by @github-actions
   @p1p1bear from Alibaba posted the full design proposal for review. Some notable highlights: The integration depends on a forked version of DuckDB, whose modifications were not part of this design but will be documented separately. Apart from insert-select and alter statements from innodb to duckdb, cross engine SQL statements (such as joins) are not supported. Note that duckdb is a first class storage engine in its own right, not a cache or otherwise off-loading innodb.
1. Bug#121063 Members of one group assign different GNOs to the same transaction by @matias-sanchez opened a Group Replication correctness fix that retries a stale Synode reservation before proposal; it is open and marked for review.
2. Large-transaction binlog optimization: Improved Binlog Handling of Large Transactions by @HugoWenTD received follow-up from @SongLibing on compression and recovery details, while SongLibing’s related implementation was closed without merge.
3. Bug#112704: Fix TempTable CTE clone handler mismatch when TempTable falls back to InnoDB by @catalinbp was approved by @mvcc and moved to internal merge queue.
4. Process: New format for writing proposals and designs as a PR against mysql-server effective immediately.

Merged

• Bump openai/codex-action from 1.11 to 1.12 … 734 by @dependabot[bot] Editor's note: Perhaps more interesting than the minor codex-action upgrade, some readers may find rather enjoyable the superficially diplomatic yet unbending dialogue, a contemporary short story par excellence, between our tireless everyday heros, the always helpful and watchful dependabot, meeting the equally helpful, politely gatekeeping cla checker, our very own oca-bot. The latter being the more assertive one - it is automating the work of a lawyer, after all - the story takes a surprising turn and confused both of the quarreling bots as it turns out the project is using a rather non-standard workflow for merging approved patches. The end scene is the modern Romeo&Juliette, the blind Oidipus, story uncarnated into bot form: the oca-gatekeeper never realized that by forcefully closing the issue, they did in fact commission it to be merged...
• Bug#112704: Fix TempTable CTE clone handler mismatch when TempTable falls back to InnoDB … 672 by @catalinbp

Implementing / Review

Active implementation and review covered replication correctness, SQL behavior, InnoDB, client builds, and an experimental type-system prototype.

• Bug#121063 Members of one group assign different GNOs to the same transaction … 736 by @matias-sanchez OCA Verified, Replication, Pluggable, Review Requested.
  The patch retries a stale Synode reservation to prevent Group Replication members from proposing different GNOs for one transaction; reviewers are requested.
• Initialize binlog event scalar members … 732 by @SeojunKim-pumisj OCA Required, Build Passed, MTR Passed.
  Initializes binlog-event scalar members and adds gunit coverage for malformed-input paths; it fixes Issue 706 and remains open.
• [POC] Prototyping - MySQL User Defined Types … 707 by @marcalff OCA Verified, InnoDB, Pluggable, Build, Tests, Review Requested, Build Passed, User Defined Types.
  Author is iterating against review feedback from Codex bot.
• Allow building mysqlclient without server deps … 680 by @BlackDex labels were not returned.
  Proposes skipping server-only ncurses/readline and RPC dependency checks when building mysqlclient alone; requested review remains pending.
• Defer Memory Allocation of AHI Hash Tables Until AHI Is Enabled … 729 by @SongLibing OCA Verified, InnoDB, Tests, Review Requested, Build Passed, MTR Failed.
  Defers full adaptive-hash-index allocation until AHI is enabled to avoid wasting 1/64 of buffer pool size when AHI is disabled. Review and MTR follow-up remain pending.
• Add DELETE ... RETURNING clause support … 733 by @evgenyp-azm OCA Verified, Tests, Review Requested, Build Passed, MTR Failed.
  Adds RETURNING to single-table DELETE, including parser, executor, and MTR changes; this is the trunk-targeted version.
• Bug#99917: Limit connections on the administrative interface … 727 by @vidyadharchelluru OCA Verified, Build, Tests, Docs.
  Adds admin_max_connections and administrative-connection status counters so privileged administrative connections can be capped independently of ordinary connections. Maintainers push back saying the port is already intended as a backup for such situations, and should not itself be used by routine admin connections. MySQL Product Mgmt is presumably talking offline with the submitter (my assumption as discussion on the thread is quiet)
• Add DELETE ... RETURNING clause support … 725 by @evgenyp-azm OCA Verified, Optimizer.
  Carries the same single-table DELETE ... RETURNING capability to the 8.4 branch; it remains under review.

Planning / PoC

The planning activity centered on storage-engine extensibility, vector indexing, InnoDB concurrency, and binlog handling for large transactions.

• [Proposal]: Integrate SVS (Vamana ANN) as a Vector Search Index Option for MySQL … 116 by @klmckeig enhancement, AI & Cloud.
  Intel proposes SVS/Vamana as a graph-based vector index type behind MySQL’s evolving vector-index interface; @henrikingo-oracle handed the integration discussion to @gopshank and @mayprasa.
• [Discussion]: Support DuckDB as a columnar storage engine being added to MySQL … 79 by @github-actions
  @p1p1bear from Alibaba posted the full design proposal for review. Some notable highlights: The integration depends on a forked version of DuckDB, whose modifications were not part of this design but will be documented separately. Apart from insert-select and alter statements from innodb to duckdb, cross engine SQL statements (such as joins) are not supported. Note that duckdb is a first class storage engine in its own right, not a cache or otherwise off-loading innodb.
• Improved Binlog Handling of Large Transactions … 683 by @HugoWenTD
  @SongLibing raised a compression and recovery-format question during the period; the proposal remains open and is explicitly related to mysql-community Issue 77.

New

• [Feature]: Add optional namespaces for tables within a database … 133 by @sambenne enhancement.
  Newly opened request for an optional namespace level within a database to group tables without creating separate schemas or relying on name prefixes.

Bug Fixes

• Bug#121045 Server telemetry metrics backed by aggregated_stats_buffer always report 0 … 735 by @hunter-kang OCA Verified.
  Fixes missing writes to the sharded aggregate buffer used by 25 OpenTelemetry metrics and adds a Performance Schema regression test; review is requested.
• Bug#121124 Ignore unchanged CHECK columns for online DDL … 723 by @modasserbillah OCA Verified, Tests, Review Requested.
  Limits CHECK-column type comparison to columns actually changed by ALTER, avoiding an unnecessary table copy for an otherwise metadata-only ENUM extension.
• Compiler warnings in In constructor ‘Log_event::Log_event(...) … 706 by @VARGA-Peter bug.
  @SeojunKim-pumisj began investigation during the period; the issue now references the open initialization fix in pull request 732.
• Add missing perl(lib)/openssl Requires to mysql-community-test package … 728 by @sjmudd OCA Verified, Review Requested, Build Passed, MTR Passed.
  Adds the missing RPM dependencies on trunk so a minimal mysql-community-test installation can run its test harness.
• Fix typo in multi-statement announcement comment … 671 by @kazanzhy
  Corrects a multi-statement announcement comment; @harinvadodaria approved the change during the period.
• Add missing perl(lib)/openssl Requires to mysql-community-test package … 731 by @sjmudd OCA Verified.
  Applies the same RPM dependency fix to the 8.4 branch; the pull request remains open.
• Add missing perl(lib)/openssl Requires to mysql-community-test package … 730 by @sjmudd OCA Verified.
  Applies the same RPM dependency fix to the 9.7 branch; the pull request remains open.

Misc / GitHub-platform and process fixes

• Process: New format for writing proposals and designs as a PR against mysql-server effective immediately.

• Last week in mysql-server development (34) … 134 by @henrikingo-oracle
  A project-newsletter discussion was created and updated during the period; JFG asked a follow-up question about covering bug fixes done still by Oracle internal process and tracked purely in bugs.mysql.com. This was ruled out of scope (and is also going away in the future)