Contents
PostgreSQL 19 is expected to be released on November 19, 2026, following the removal of several planned features late in the beta cycle and an extended review period. Two PostgreSQL committers say LLM-assisted bug reports have increased the work required before release. However, some of the withdrawn features also had architectural problems.
Key Points#
- As of November 16, PostgreSQL 19 RC2 is available, with the final release expected on November 19.
- Beta 4 withdrew features including SQL/PGQ, online checksum changes, updates and deletes targeting time ranges, and partition merging and splitting.
- According to committers, LLM-assisted reviews are finding bugs faster, increasing the number of issues requiring investigation and fixes.
- There is no guarantee when withdrawn features will return, so teams need to check their upgrade plans against the final release notes.
Features Removed in Beta 4#
Beta 4, released in late September, removed several features that had passed earlier testing. For teams evaluating PostgreSQL 19, these withdrawals matter more than the change in release date. Applications designed around beta features may need alternative implementations.
The removed features are as follows.
| Feature | Impact on Upgrade Plans | |---|---| | SQL/PGQ support | Applications cannot be built around the planned property graph query syntax. | | Online checksum changes | Plans to enable or disable data checksums while the database remains running need to be reconsidered. | | Updates and deletes using FOR PORTION OF | Designs relying on the planned syntax for changes targeting time ranges need an alternative approach. | | MERGE PARTITIONS and SPLIT PARTITIONS | Partition maintenance procedures cannot assume these commands will be available. | | DDL helper functions | Scripts using pg_get_role_ddl(), pg_get_tablespace_ddl(), or pg_get_database_ddl() need changes. |
Beta 4 also reverted a change that forced the postmaster process's LC_COLLATE to C. GROUP BY ALL had already been withdrawn in Beta 3.
Why the Release Was Delayed#
The community's statement on Beta 4 acknowledged the value of a predictable release schedule while making reliability the top priority. Removing unfinished features from PostgreSQL 19 gave developers time to address design and review issues.
This distinction matters. Some issues could be addressed with individual fixes, while others required substantial changes beyond what could safely be completed during the beta period. Removing features reduces release risk, but the remaining review and stabilization work still takes time.
The release notes leave open the possibility that withdrawn features will appear in future major versions. They do not, however, promise inclusion in a specific release such as PostgreSQL 20.
How AI Review Played a Role#
In his October 8 post, “The Reverts Will Continue Until Morale Improves,” PostgreSQL core team member and committer Peter Eisentraut divided the issues found into two categories: serious architectural flaws and a long list of relatively minor edge cases.
His argument was that LLM-assisted review brings these smaller bugs to light much faster. Issues that might previously have emerged gradually over five years can arrive all at once as roughly 40 bug reports. That figure illustrates his explanation; it is not a published total for PostgreSQL 19.
Committer Robert Haas made a similar point in his October 30 post, “AI Is Why Postgres 19 Is Late.” He attributed the delay to many people reporting bugs found with LLMs, leaving committers busy investigating and fixing them.
What the Evidence Supports#
These accounts support the conclusion that faster bug discovery increased the workload just before release. They do not show that AI created the original bugs or that every withdrawn feature was removed because of AI-generated reports.
The material presented also contains no measured acceptance rate for LLM-generated reports, comparison with traditional reviews, or false-positive count. Claims that AI review is always more accurate go beyond the evidence.
The practical lesson concerns review capacity. Finding a potential bug is only the starting point: maintainers must reproduce it, assess its impact, and review a fix. Faster reporting can make software quality easier to assess while increasing short-term pressure on release schedules.
How to Adjust Upgrade Plans#
Treat November 19 as a target date until it is confirmed by the official release announcement. RC2, published on November 12, remains a release candidate as of November 16.
If upgrade preparations are already underway, check the following.
1. Check required features against the final PostgreSQL 19 release notes. 2. Search application SQL and maintenance scripts for dependencies on withdrawn features. 3. Test the final release with representative workloads before deploying to production. 4. Remove assumed delivery dates for withdrawn features from internal plans until the project confirms their timing.
The upgrade may still be worthwhile without these additions. The decision should be based on the features actually included and the results of your own compatibility testing.
Next Steps#
Before setting a production deployment date, review your PostgreSQL 19 upgrade checklist and explicitly identify every dependency on withdrawn beta features.
Source: https://www.publickey1.jp/blog/26/postgresql_194ai.html
Contents
