Re: [PATCH] Don't call ereport(ERROR) from recovery target GUC assign hooks

On Sun, Jul 12, 2026 at 7:45 PM JoongHyuk Shin <sjh910805(at)gmail(dot)com> wrote:
>
> On Wed, Jul 8, 2026 at 10:39 AM Fujii Masao <masao(dot)fujii(at)gmail(dot)com> wrote:
> > We can remove assign_recovery_target_name()?
>
> Yes, done in v9. recovery_target_name now stores directly into
> recoveryTargetName, so recovery_target_name_string and the assign hook
> are both gone.

Thanks for updating the patch!

I've updated it further. Could you review this version?

+ my $res = run_log(
+ [
+ 'pg_ctl',
+ '--pgdata' => $node_standby->data_dir,
+ '--log' => $node_standby->logfile,
+ '--options' => $options,
+ 'start',
+ ]);
+ ok($res, "server starts for $test_name");

Starting a test node with pg_ctl directly bypasses Cluster.pm's PID tracking.
As a result, cleanup via stop(), teardown_node(), etc. may not work correctly,
and the postmaster can be left running if the test fails or exits unexpectedly.
To avoid that, I added a preliminary patch that allows
PostgreSQL::Test::Cluster::start() to pass extra postmaster options through
pg_ctl --options, and then updated the test to use that start().

+sub test_recovery_standby_with_options
+{

I merged test_recovery_standby_with_options() into the existing
test_recovery_standby() by adding an optional options parameter. This removes
duplicated standby setup, catch-up waiting, row-count checks, and teardown code.

Regards,

--
Fujii Masao