Hi, Peter On Tue, 22 Jul 2025 at 17:46, Peter Smith <smithpb2...@gmail.com> wrote: > Hi. > > Here are the latest v14 patches. > > Changes include: > > PATCH 0002. > - Fixes the enable_seqscan PANIC bug reported by Japin [1] > - Adds a new regression test case for this > > ====== > [1] > https://www.postgresql.org/message-id/ME0P300MB04457E24CA8965F008FB2CDBB648A%40ME0P300MB0445.AUSP300.PROD.OUTLOOK.COM >
After some investigation, I found that the VACUUM assertion failures were caused by an uninitialized HeapTupleFreeze structure. PFA. -- Regards, Japin Li
>From 452996a7c9b2ae08df464007ecbb81a728638bbb Mon Sep 17 00:00:00 2001 From: Japin Li <japi...@hotmail.com> Date: Wed, 23 Jul 2025 11:48:58 +0800 Subject: [PATCH] Initialize HeapTupleFreeze structure prior to use --- contrib/vci/expected/bugs.out | 8 ++++++++ contrib/vci/sql/bugs.sql | 10 ++++++++++ contrib/vci/storage/vci_ros_command.c | 2 +- 3 files changed, 19 insertions(+), 1 deletion(-) diff --git a/contrib/vci/expected/bugs.out b/contrib/vci/expected/bugs.out index 2a77ab55214..231fad69f28 100644 --- a/contrib/vci/expected/bugs.out +++ b/contrib/vci/expected/bugs.out @@ -10,3 +10,11 @@ SELECT * FROM t1 WHERE id = 100; 100 | f899139df5e1059396431415e770c6dd (1 row) +-- Bug reported by Japin Li that resulted TRAP +-- See https://www.postgresql.org/message-id/SY8P300MB0442BEC3F5CF432F0121ACC4B642A%40SY8P300MB0442.AUSP300.PROD.OUTLOOK.COM +CREATE TABLE t2 (id int, info text) WITH (autovacuum_enabled = off); +CREATE INDEX t2_id_idx ON t2 USING vci (id); +INSERT INTO t2 SELECT id, 'test' || id FROM generate_series(1, 1000) id; +DELETE FROM t2 WHERE id % 10 = 0; +VACUUM t2; +DROP TABLE t2; diff --git a/contrib/vci/sql/bugs.sql b/contrib/vci/sql/bugs.sql index 3db6e6d37e0..cd45756ad00 100644 --- a/contrib/vci/sql/bugs.sql +++ b/contrib/vci/sql/bugs.sql @@ -6,3 +6,13 @@ CREATE INDEX t1_id_idx ON t1 USING vci (id); INSERT INTO t1 SELECT id, md5(id::text) FROM generate_series(1, 1000) id; SET enable_seqscan TO off; SELECT * FROM t1 WHERE id = 100; + +-- Bug reported by Japin Li that resulted TRAP +-- See https://www.postgresql.org/message-id/SY8P300MB0442BEC3F5CF432F0121ACC4B642A%40SY8P300MB0442.AUSP300.PROD.OUTLOOK.COM + +CREATE TABLE t2 (id int, info text) WITH (autovacuum_enabled = off); +CREATE INDEX t2_id_idx ON t2 USING vci (id); +INSERT INTO t2 SELECT id, 'test' || id FROM generate_series(1, 1000) id; +DELETE FROM t2 WHERE id % 10 = 0; +VACUUM t2; +DROP TABLE t2; diff --git a/contrib/vci/storage/vci_ros_command.c b/contrib/vci/storage/vci_ros_command.c index b3a802a77a0..e56a5f803c4 100644 --- a/contrib/vci/storage/vci_ros_command.c +++ b/contrib/vci/storage/vci_ros_command.c @@ -3324,7 +3324,7 @@ freezeWos(vci_RosCommandContext *comContext, vci_MainRelVar wosType, Snapshot sn BlockNumber nblocks, blkno; - frozen = palloc(sizeof(HeapTupleFreeze) * MaxHeapTuplesPerPage); + frozen = palloc0(sizeof(HeapTupleFreeze) * MaxHeapTuplesPerPage); oid = vci_GetMainRelVar(&comContext->info, wosType, 0); -- 2.43.0
