Fix race condition in ginHeapTupleFastInsert (#2412)

Browse Source

Because the metadata was not locked, it could be updated concurrently
such that we wouldn't actually have the tail block.

The current ordering works better, as we still only start XLogBeginInsert()
once we have all potentially interesting buffers loaded in memory, but
still have correct lock lifetimes.

See also: access/transam/README section Write-Ahead Log Coding

...

This commit is contained in:

MMeent

2022-09-08 12:57:30 +02:00

committed by GitHub

parent 9e3136ea37

commit 1351beae19

2 changed files with 2 additions and 2 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download Patch File Download Diff File Expand all files Collapse all files

2

vendor/postgres-v14 vendored

Submodule vendor/postgres-v14 updated: a4963aa6df...e8518d3fc8

2

vendor/postgres-v15 vendored

Submodule vendor/postgres-v15 updated: 26c6466873...313769bb62