Horton

Re: First draft of PG 19 release notes

John Naylor <johncnaylorls@gmail.com>

View thread
From: John Naylor <johncnaylorls@gmail.com>
To: Bruce Momjian <bruce@momjian.us>
Cc: Richard Guo <guofenglinux@gmail.com>, PostgreSQL-development <pgsql-hackers@lists.postgresql.org>
Date: 2026-05-11T05:40:38Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. doc PG 19 relnotes: more fixes

  2. doc PG 19 relnotes: various corrections

  3. doc PG 19 relnotes: adjust item to mention pg_replication_slots

  4. doc PG 19 relnotes: remove "Add fake LSN support to hash index"

  5. doc PG 19 relnotes: add two optimizer hooks

  6. doc PG 19 relnotes: remove "Optionally" for CPU optimizations

  7. doc PG 19 relnotes: adjustments/removal of items

  8. Use fake LSNs to improve nbtree dropPin behavior.

  9. Clean up ICU includes.

On Wed, Apr 15, 2026 at 8:19 AM Bruce Momjian <bruce@momjian.us> wrote:

Hi Bruce, thanks for you work on this. Here are some comments:

> Allow C++ compiler mode to be used with ICU (John Naylor) §

That was just for our headerscheck and I'd be surprised if it mattered
for extension authors. Also, a later commit af2d4ca191 got rid of
including ICU headers in our local headers entirely, so that
supercedes the above anyway.

> Optionally use AVX2 CPU instructions for calculating page checksums (Matthew Sterrett, Andrew Kim) §

> Optionally use ARM Crypto Extension to Compute CRC32C (John Naylor) §

There's no option for these, the extensions are used where available.

> Improve sort performance using radix sorts (John Naylor) §

Minor quibble: I think it's more normal to refer to "sort" as a
non-count noun, when talking about a technique.

> Improve performance of numeric operations on platforms without 128-bit integer support (Dean Rasheed) §

The intent was a more of a simplification/refactoring. Server
platforms without 128-bit integer support are few and far between
(e.g. S390X on some broken versions of clang, which I tested out of
curiosity).

> Change index access method handlers to use a static IndexAmRoutines structure, rather than dynamically allocated ones (Matthias van de Meent) §
>
> This is a backwardly incompatible.

I understand this whole section to be about incompatibilities -- if
there was a follow-up mention here, it seems to have been lost along
the way.

--
John Naylor
Amazon Web Services

-- 
John Naylor
Amazon Web Services