Horton

Re: Segfault on ANALYZE in SERIALIZABLE isolation

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Sergei Kornilov <sk@zsrv.org>
Cc: Joe Wildish <joe-postgresql.org@elusive.cx>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2019-05-19T23:26:56Z
Lists: pgsql-hackers 
Hi,

On 2019-05-18 18:12:31 +0300, Sergei Kornilov wrote:
> Seems table_beginscan_analyze (src/include/access/tableam.h) should not pass second argument as NULL.

As hopefully explained downthread, and in the commit message, that's not
really the concern. We shouldn't use the snapshot in the first place.

> PS: also I noticed in src/include/utils/snapshot.h exactly same
> comment for both SNAPSHOT_SELF and SNAPSHOT_DIRTY - they have no
> difference?

That was copy & paste mistake. Fixed. Also expanded the comments a
bit. Thanks for noticing!

Greetings,

Andres Freund

Commits

  1. Fix and improve SnapshotType comments.

  2. tableam: VACUUM and ANALYZE support.