On 4/24/25 20:12, Michael Paquier wrote:
On Thu, Apr 24, 2025 at 03:34:29PM +, David Steele wrote:
Having said that, I think these tests are awfully expensive for a single
GUC. Unit tests would definitely be preferable but that's not an option for
GUCs, so far as I know.
On my laptop, 003
On Thu, Apr 24, 2025 at 03:34:29PM +, David Steele wrote:
> Done. This means there are no commas in the upper bound but I don't think
> it's a big deal and it more closely matches other GUC messages.
One thing that I have double-checked is if we have similar messages
for GUCs that are defined
On 2/14/25 02:42, Michael Paquier wrote:
On Fri, Jan 24, 2025 at 01:36:45PM +, David Steele wrote:
+ timeline = strtoull(*newval, &endp, 0);
+
+ if (*endp != '\0' || errno == EINVAL || errno == ERANGE)
{
GUC_check_errdetail
On Fri, Jan 24, 2025 at 01:36:45PM +, David Steele wrote:
> I attached the wrong patch. Oops!
Thanks for the new patch.
> + timeline = strtoull(*newval, &endp, 0);
> +
> + if (*endp != '\0' || errno == EINVAL || errno == ERANGE)
> {
>
On 1/24/25 01:44, Michael Paquier wrote:
On Thu, Jan 23, 2025 at 02:53:39PM +, David Steele wrote:
I discovered this while testing on Postgres versions < 12 where
The tests are probably excessive but I needed something to show that the
verification works as expected.
Even with your patch,
On Thu, Jan 23, 2025 at 02:53:39PM +, David Steele wrote:
> Currently check_recovery_target_timeline() converts any value that is not
> current, latest, or a valid integer to 0. So for example:
>
> recovery_target_timeline = 'currrent'
>
> results in the following error:
>
> FATAL: 22023: r
