Message-ID: <e6a27751-8c91-4bff-ba07-2e97de99e34f@pgbackrest.org>
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: Improve checks for GUC recovery_target_xid
To: Michael Paquier <michael@paquier.xyz>
Cc: Fujii Masao <masao.fujii@gmail.com>, =?UTF-8?Q?H=C3=BCseyin_Demir?=
 <huseyin.d3r@gmail.com>, Pg Hackers <pgsql-hackers@postgresql.org>
References: <f14463ab-990b-4ae9-a177-998d2677aae0@pgbackrest.org>
 <CAHGQGwEnakTrosMp3Y=Trya4MTC-h8Lqx3FDU1mOALRQ0rz59Q@mail.gmail.com>
 <9e8b7ee9-4a16-477a-baa5-0cdf37a04798@pgbackrest.org>
 <CAHGQGwG44vZbSoBmg076G+xkR6n=Tj2=q+fVkfP7yEsyF1daFA@mail.gmail.com>
 <c2e97e35-1227-4351-8faa-fa95208f6cf9@pgbackrest.org>
 <aakAmNjKyBWRU1vR@paquier.xyz>
Content-Language: en-US
From: David Steele <david@pgbackrest.org>
In-Reply-To: <aakAmNjKyBWRU1vR@paquier.xyz>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Date: Thu,  5 Mar 2026 04:21:16 +0000 (UTC)
Archived-At: 
 <https://www.postgresql.org/message-id/e6a27751-8c91-4bff-ba07-2e97de99e34f%40pgbackrest.org>
Precedence: bulk

On 3/5/26 11:03, Michael Paquier wrote:
> On Thu, Mar 05, 2026 at 03:40:44AM +0000, David Steele wrote:
>> I wrote the tests for recovery_target_timeline but I was not too satisfied
>> with them because starting Postgres is fairly expensive.
> 
> +# Invalid recovery_target_xid tests
> +$node_standby = PostgreSQL::Test::Cluster->new('standby_10');
> +$node_standby->init_from_backup($node_primary, 'my_backup',
> +    has_restoring => 1);
> +$node_standby->start;
> +
> +my ($result, $stdout, $stderr) = $node_primary->psql('postgres',
> +    "ALTER SYSTEM SET recovery_target_xid TO 'bogus'");
> +like(
> +    $stderr,
> +    qr/is not a valid number/,
> +    "invalid recovery_target_xid (bogus value)");
> 
> Smart move to rely on ALTER SYSTEM to check how the GUC callback is
> reacting on incorrect input values.  Why do you need to create and
> start a new standby if it is not used, then?

The prior standby is not running because of the invalid config. I 
figured it was better to start clean but when I update the 
recovery_target_timeline tests I was planning to use the same standby 
for all the new tests.

Regards,
-David