Re: pg_get__*_ddl consolidation - Mailing list pgsql-hackers

From Andrew Dunstan
Subject Re: pg_get__*_ddl consolidation
Date
Msg-id 786bceca-84c4-4a83-9c30-3ce49d81b274@dunslane.net
Whole thread
In response to Re: pg_get__*_ddl consolidation  (SATYANARAYANA NARLAPURAM <satyanarlapuram@gmail.com>)
List pgsql-hackers


On 2026-04-13 Mo 3:41 PM, SATYANARAYANA NARLAPURAM wrote:
Hi,

On Fri, Apr 10, 2026 at 1:03 PM Jeff Davis <pgsql@j-davis.com> wrote:
On Sun, 2026-04-05 at 11:06 -0400, Andrew Dunstan wrote:
> Pushed. I have moved the remaining get_*_ddl items to PG20-1

The line:

  role_settings = DatumGetArrayTypeP(datum);

should be DatumGetArrayTypePCopy(), because it's being pfree()d later.
The existing code will sometimes make a copy and sometimes not, e.g.:

  -- settings are contrived to make the datum inline
  CREATE USER u1;
  ALTER ROLE u1 SET search_path = 'public, pg_catalog, pg_temp';
  ALTER ROLE u1 SET work_mem='64MB';
  ALTER ROLE u1 SET statement_timeout='30s';
  ALTER ROLE u1 SET lock_timeout='10s';
  ALTER ROLE u1 SET idle_in_transaction_session_timeout = '60s';
  SELECT pg_get_role_ddl('u1');
  ERROR:  pfree called with invalid pointer 0x7986dd0c7cc8 (header
0x0000400600000000)

Yes, it appears to be a bug. Attached a patch to fix this. Tested with the 
attached patch and don't see server crashing after that.



Thanks, pushed.


cheers


andrew

--
Andrew Dunstan
EDB: https://www.enterprisedb.com

pgsql-hackers by date:

Previous
From: David Rowley
Date:
Subject: Re: [PATCH] Miscellaneous little fixes
Next
From: Peter Smith
Date:
Subject: Re: Add missing period to HINT messages