Hi,
On Thu, 21 Nov 2024 at 09:32, Bertrand Drouvot
<bertranddrouvot.pg@gmail.com> wrote:
> That was in fact the main reason why I added this test. But well, just
> adding the "write_to_file" in the injection test is enough to "show" that this
> member does exist. So I'm fine with v3.
I think that the changes below (write_to_file checks) should come
after the assertion checks. Otherwise, they could mask some problems.
=== 1
- if (!info || !info->fixed_amount)
+ /* skip if not fixed or this kind does not want to write to the file */
+ if (!info || !info->fixed_amount || !info->write_to_file)
continue;
if (pgstat_is_kind_builtin(kind))
Assert(info->snapshot_ctl_off != 0);
=== 2
kind_info = pgstat_get_kind_info(ps->key.kind);
+ /* skip if this kind does not want to write to the file */
+ if (!kind_info->write_to_file)
+ continue;
+
/* if not dropped the valid-entry refcount should exist */
Assert(pg_atomic_read_u32(&ps->refcount) > 0);
===
--
Regards,
Nazir Bilal Yavuz
Microsoft
