Horton

Re: Assertion failure in HEAD and 13 after calling COMMIT in a stored proc

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

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Michael Paquier <michael@paquier.xyz>
Cc: Jim Nasby <nasbyj@amazon.com>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>
Date: 2021-06-22T17:13:08Z
Lists: pgsql-hackers

Attachments

I wrote:
> The attached seems to be enough to resolve Jim's example.  I'd like
> to invent a test case that involves a detoast of the simple
> expression's result, too, to show that transiently pushing a
> snapshot for the duration of the expression is not the right fix.

Here we go.  This test case gives "cannot fetch toast data without an
active snapshot" in v11 and v12 branch tips.  Since those branches lack
the 73b06cf89 optimization, they push a snapshot while calling the
SQL-language function, thus it doesn't complain.  But what comes back
is toasted, and then we fail trying to detoast it.

			regards, tom lane

Commits

  1. Restore the portal-level snapshot for simple expressions, too.

  2. Centralize the logic for protective copying of utility statements.

  3. Restore the portal-level snapshot after procedure COMMIT/ROLLBACK.

  4. Avoid taking a new snapshot for an immutable simple expression in plpgsql.