Postgres 19 Slips Toward Nov. 19 After Eight to 10 Features Are Pulled
A.I. / news
Postgres 19 Slips Toward Nov. 19 After Eight to 10 Features Are Pulled
Core team member Peter Eisentraut ties part of the reverts to LLM-assisted code review, and his own post names a spring-to-summer security backlog as another cause.
The PostgreSQL project is now aiming to release version 19 around Nov. 19, after eight to 10 features were reverted from the release and pushed to PostgreSQL 20, The Register reported on Friday.
The release was originally slated for late September or early October. The project's roadmap page still says October 2026, and carries a banner announcing PostgreSQL 19 Release Candidate 1 on Oct. 15.
What Peter Eisentraut wrote on Oct. 9
The account behind the story is a blog post by Peter Eisentraut, a PostgreSQL core team member, dated Oct. 9 and titled "The reverts will continue until morale improves." He wrote that one revert per cycle is normal and that this cycle saw eight to 10 after beta began, some of them late.
His argument is that LLM-assisted code review now finds problems far faster than they used to surface. Issues that once appeared slowly, over about five years by his account, now arrive together. "With LLM-assisted code review, these kinds of things can get found much faster," he wrote, as quoted by The Register.
The arithmetic he gives is a backlog of "forty defect reports." Even at roughly three lines of fix each, he wrote, the fixed overhead of handling each one can use up the time left in beta.
Eisentraut also wrote that the problem is not "vibe coding," because the reverted code was written before that practice was available. The effective quality bar has risen since the features were committed, he argues.
The other causes in the same post
The Register's headline puts the AI reviewers first. The post lists more causes than that, and the reporting on them is thinner.
| Cause named in the post | Detail |
|---|---|
| Architectural defects | Some features had design flaws that were hard to fix during beta |
| Security workload | Senior developers spent April to August on security fixes, which he calls the "Vulnpocalypse", cutting beta review |
| Complexity | Features that touch many combinations of database objects are hard to test |
| No experimental status | The project has discussed committing features as experimental but has not settled on a method |
He wrote that two more months might have resolved the remaining issues, but the community kept the release schedule. LWN's write-up of the post, dated Oct. 13, records commenters disagreeing on how much LLM review contributed. One thought the coverage overstated it. Another pointed to LLM-driven vulnerability reports and fuzzing consuming the beta period.
Which features were pulled and which stayed
The Register names one casualty: SQL Property Graph Queries, known as SQL/PGQ, a graph-query syntax in the SQL standard since 2023. A September thread on the pgsql-hackers mailing list shows Eisentraut agreeing to revert it, and Robert Haas asking that it come out of both the 19 and master branches.
Tom Lane, a PostgreSQL contributor, warned in September that if the feature shipped in 19, "there will be post-release bug discoveries that are unfixable until v20," per The Register.
Still in the release are REPACK CONCURRENTLY, which reclaims disk space without the table lock that VACUUM FULL takes, logical replication of sequences, which was itself reverted once before, and the pg_plan_advice extension. Eisentraut wrote that he is "confident that many of the reverted features will come back before too long."
Other open-source projects face the same release arithmetic. Forgejo shipped 17.0 on Oct. 15 with a webhook change that needs manual migration, and AWS's AI-DLC workflow has 230 issues open.
What is not known about the Nov. 19 date
The Register does not name the AI tools used in the reviews, and neither does Eisentraut's post as summarised by LWN. It gives no total bug count beyond the forty-report illustration. The Terminal found no project announcement of Nov. 19 on the roadmap page, which still reads October.
PostgreSQL 20 was expected to contain only bug fixes. It will now likely carry the features pushed out of 19. Eisentraut wrote that unless LLM-assisted review keeps improving at its current rate, the situation should sort itself out, per The Register.
Sources
More in A.I.
- 01Qwen3-4B-Deslop Cuts Its Slop Score From 121 to 12.5 on a MacBook Air, and Two of 40 Answers Came Out DamagedThe Show HN fine-tune by Erik Markus Kannike uses a rank-16 LoRA and a 175-byte neuron edit; its blog gives no licence, its Hugging Face card says Apache-2.0.
- 02HybridAgent Paper Claims 10.4% Gain and 2.7x to 3.7x Lower Cost, but Its Own Text Gives 10.4% Two Different MeaningsThe Stanford, UIUC and Together AI preprint ships MIT-licensed 8B and 14B agents, and no independent run of its numbers has been reported.
- 03Gemini 4 Argon Reaches the API at $2 and $10 Per Million Tokens, With a No. 1 Ranking and No ScoreGoogle's October 13 post says Argon tops the Vals Index but publishes no figure for it, and the introductory price doubles on November 30.
- 04DeepSeek Swapped V4.1-Flash-1017 Behind Its API Name, and the New Card Lists NL2Repo at 65.4 Where the Old One Said 64.0The MIT-licensed update calls itself minor and is now what deepseek-flash serves, with no active-parameter count and no release date on the card.