Horton

Re: BUG #17385: "RESET transaction_isolation" inside serializable transaction causes Assert at the transaction end

Dmitry Koval <d.koval@postgrespro.ru>

View thread View on postgr.es
From: Dmitry Koval <d.koval@postgrespro.ru>
To: Dilip Kumar <dilipbalaut@gmail.com>, Masahiko Sawada <sawada.mshk@gmail.com>
Cc: andrewbille@gmail.com, PostgreSQL mailing lists <pgsql-bugs@lists.postgresql.org>
Date: 2022-02-06T07:37:47Z
Lists: pgsql-bugs 
> In  if (source != PGC_S_DEFAULT) we are overwriting newval with
> conf->reset_val so I think we should free the newval or can we even
> avoid guc_strdup in case of (source != PGC_S_DEFAULT)?

I think we can not avoid guc_strdup in case of (source != PGC_S_DEFAULT) 
because function call_string_check_hook() contains "free(*newval);" in 
PG_CATCH-section.
Probably we should free the newval in block

+                    if (source != PGC_S_DEFAULT)
+                    {
+                        /* Release newextra as we use reset_extra */
+                        if (newextra)
+                            free(newextra);

With best regards,
Dmitry Koval.

Commits

  1. Renumber GUC flags for a bit more sanity.

  2. Introduce GUC_NO_RESET flag.