Horton

Re: search_plan_tree(): handling of non-leaf CustomScanState nodes causes segfault

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Zhihong Yu <zyu@yugabyte.com>
Cc: David Geier <david@swarm64.com>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2021-01-18T21:13:28Z
Lists: pgsql-hackers 
Zhihong Yu <zyu@yugabyte.com> writes:
> I was thinking that, if sstate->ss_currentRelation is null for the other
> cases, that would be a bug.
> An assertion can be added for the cases ending with T_TidScanState.

Maybe, but there are surely a lot of other places that would crash
in such a case --- places far more often traversed than search_plan_tree.
I do not see any value in complicating search_plan_tree for that.

			regards, tom lane

Commits

  1. Remove faulty support for MergeAppend plan with WHERE CURRENT OF.

  2. Avoid crash with WHERE CURRENT OF and a custom scan plan.