PG-2424 PostgreSQL 19 support by artemgavrilov · Pull Request #806 · percona/pg_stat_monitor

artemgavrilov · 2026-07-30T16:33:40Z

PG-2424

Description

https://www.postgresql.org/docs/19/release-19.html#RELEASE-19-PGSTATSTATEMENTS

postgres/postgres@REL_18_4...REL_19_BETA2

Links

Fixes #773

codecov · 2026-07-31T14:12:49Z

Codecov Report

❌ Patch coverage is 86.84211% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 88.08%. Comparing base (8bd27c7) to head (96a3336).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/pg_stat_monitor.c	86.84%	0 Missing and 5 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #806      +/-   ##
==========================================
- Coverage   88.57%   88.08%   -0.49%     
==========================================
  Files           3        3              
  Lines        1304     1318      +14     
  Branches      180      183       +3     
==========================================
+ Hits         1155     1161       +6     
- Misses         75       80       +5     
- Partials       74       77       +3

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

jeltz

I have not reviewed what actually changed in PGSS.

jeltz · 2026-08-03T05:39:06Z

    name: Build
    runs-on: ${{ inputs.os }}
-    timeout-minutes: 10
+    timeout-minutes: 15


Why was this change necessary?

Also is this really the right commit?

Tests run for PG19 takes slightly more time. It was added to this commit to make it atomic, so CI will pass together with PG19 related changes.

Ah, so then it is just in the wrong commit.

So you want this to be separate commit anyway?

jeltz · 2026-08-04T10:21:57Z

Do we need to support RELKIND_PROPGRAPH in pgsm_ExecutorCheckPerms() like we support views?

artemgavrilov · 2026-08-04T17:11:12Z

Do we need to support RELKIND_PROPGRAPH in pgsm_ExecutorCheckPerms() like we support views?

Added

adeshthack · 2026-08-04T21:13:24Z

what are the concerns regarding changes in PGSS that are being referenced here?

jeltz · 2026-08-05T10:20:41Z

@adeshthack The main thing was ComputeConstantLengths() and the addition of counts of generic and custom plans.

jeltz · 2026-08-05T10:27:56Z


 - Our name and version are now in `pg_get_loaded_modules()` for PostgreSQL 18+
 - Support for `EXEC_BACKEND` builds ([PG-2547](https://perconadev.atlassian.net/browse/PG-2547))
+- Add PostgreSQL 19 support ([PG-2424](https://perconadev.atlassian.net/browse/PG-2424))


Maybe mention the two new fields?

Property graph support should maybe also be mentioned.

jeltz · 2026-08-05T10:36:19Z

-		if (rte->relkind == RELKIND_VIEW)
+		if (rte->relkind == RELKIND_VIEW
+#if PG_VERSION_NUM >= 190000
+			|| rte->relkind == RELKIND_PROPGRAPH


Us even getting here show that the concidional a bit up has a typo in it. We should fix that conditional.

if (rte->rtekind != RTE_RELATION #if PG_VERSION_NUM >= 160000 && rte->rtekind != RTE_SUBQUERY && rte->relkind != RELKIND_VIEW #endif )

Has a typo in it. It should have been something like && !(rte->rtekind == RTE_SUBQUERY && rte->relkind == RELKIND_VIEW).

Actually maybe it is even unnecessary since it cannot be true given the assertion in the PostgreSQL code, right? Maybe we should fcheck for perminfoindex or similar? But either way we should clean up the above.

/* * Only relation RTEs and subquery RTEs that were once relation * RTEs (views, property graphs) have their perminfoindex set. */ Assert(rte->rtekind == RTE_RELATION || (rte->rtekind == RTE_SUBQUERY && (rte->relkind == RELKIND_VIEW || rte->relkind == RELKIND_PROPGRAPH)));

perminfoindex fits perfectly here 👍

Make pg_stat_monitor compilable against PostgreSQL 19 PostgreSQL 19 optimized SELECT ... INTO statements execution(ce8d5fe), so some of them are no longer tracked by pg_stat_statements/pg_stat_monitor. So we update level tracking test to address this change. Build and test CI workflow required timeout increase as PG 19 requires slightly more time.

There are no PostgreSQL 19 packages yet, so add this version only to source based workflows.

pg_stat_statements introduces two new coutners in PostgreSQL 19: - generic_plan_calls - custom_plan_calls These counters track how many times a prepared statement was executed using a generic or custom plan. Backport them to pg_stat_monitor.

PostgreSQL 19 exposes ComputeConstantLengths function, so there is no more reasons to compute constant lenght with our own implemenation.

In PostgreSQL 19 some pg_stat_statements tests got updates. Backport them to pg_stat_monitor.

Backport the upstream pg_stat_statements fix (commit b1635c16669) from PostgreSQL 18. Mark same-location duplicates with length = -1 before the squashed check.

Mark property graphs with * the same way as we do for views.

Condition had and issue where it was keeping all subqueries, but it should keep only subqueries that are views or property graphs. Rewrite it with better condition: perminfoindex is non-zero for all types of RTEs that we need.

adeshthack · 2026-08-05T20:48:01Z

@adeshthack The main thing was ComputeConstantLengths() and the addition of counts of generic and custom plans.

yea sounds good

artemgavrilov force-pushed the PG-2424-pg19 branch 3 times, most recently from 19a2402 to fed60d2 Compare July 31, 2026 14:07

artemgavrilov force-pushed the PG-2424-pg19 branch from fed60d2 to 481ccb5 Compare July 31, 2026 15:35

artemgavrilov marked this pull request as ready for review July 31, 2026 16:17

artemgavrilov force-pushed the PG-2424-pg19 branch from 0c199bc to 2a116fe Compare July 31, 2026 16:25

jeltz requested changes Aug 3, 2026

View reviewed changes

This was referenced Aug 3, 2026

Add PostgreSQL 19 compatibility #654

Closed

PostgreSQL 19 support #773

Open

artemgavrilov force-pushed the PG-2424-pg19 branch 4 times, most recently from 958ddb5 to fed3a0b Compare August 4, 2026 17:04

artemgavrilov requested a review from jeltz August 4, 2026 17:17

jeltz requested changes Aug 5, 2026

View reviewed changes

artemgavrilov added 9 commits August 5, 2026 15:52

PG-2424 Add PostgreSQL 19 to ci (source only)

14aed53

There are no PostgreSQL 19 packages yet, so add this version only to source based workflows.

PG-2424 Add generic and custom plan counts

2d74158

pg_stat_statements introduces two new coutners in PostgreSQL 19: - generic_plan_calls - custom_plan_calls These counters track how many times a prepared statement was executed using a generic or custom plan. Backport them to pg_stat_monitor.

PG-2424 Use ComputeConstantLengths function for PostgreSQL 19

364f3e9

PostgreSQL 19 exposes ComputeConstantLengths function, so there is no more reasons to compute constant lenght with our own implemenation.

PG-2424 Update CHANGELOG.md

1d5bf64

PG-2424 Backport test updates from PG19

0b5c757

In PostgreSQL 19 some pg_stat_statements tests got updates. Backport them to pg_stat_monitor.

PG-2424 Fix squashing constant duplicates

9a075ec

Backport the upstream pg_stat_statements fix (commit b1635c16669) from PostgreSQL 18. Mark same-location duplicates with length = -1 before the squashed check.

PG-2424 Mark property graphs with *

696e3c4

Mark property graphs with * the same way as we do for views.

PG-2424 Fix RTE skip condition

96a3336

Condition had and issue where it was keeping all subqueries, but it should keep only subqueries that are views or property graphs. Rewrite it with better condition: perminfoindex is non-zero for all types of RTEs that we need.

artemgavrilov force-pushed the PG-2424-pg19 branch from fed3a0b to 96a3336 Compare August 5, 2026 13:52

artemgavrilov requested a review from jeltz August 5, 2026 14:00

Uh oh!

Conversation

artemgavrilov commented Jul 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Links

Uh oh!

codecov Bot commented Jul 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

jeltz left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jeltz commented Aug 4, 2026

Uh oh!

artemgavrilov commented Aug 4, 2026

Uh oh!

adeshthack commented Aug 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jeltz commented Aug 5, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

adeshthack commented Aug 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

artemgavrilov commented Jul 30, 2026 •

edited

Loading

codecov Bot commented Jul 31, 2026 •

edited

Loading

adeshthack commented Aug 4, 2026 •

edited

Loading