Re: Pluggable cumulative statistics

Hi,

On Mon, Jul 08, 2024 at 07:22:32AM +0000, Bertrand Drouvot wrote:
> Except the above (which is just a Nit), 0001 LGTM.
>

Looking at 0002:

It looks pretty straightforward, just one comment:

+ ptr = ((char *) ctl) + kind_info->shared_ctl_off;
+ kind_info->init_shmem_cb((void *) ptr);

I don't think we need to cast ptr to void when calling init_shmem_cb(). Looking
at some examples in the code, it does not look like we cast the argument to void
when a function has (void *) as parameter (also there is examples in 0003 where
it's not done, see next comments for 0003).

So I think removing the cast here would be more consistent.

Looking at 0003:

It looks pretty straightforward. Also for example, here:

+ fputc(PGSTAT_FILE_ENTRY_FIXED, fpout);
+ write_chunk_s(fpout, &kind);
write_chunk(fpout, ptr, info->shared_data_len);

ptr is not casted to void when calling write_chunk() while its second parameter
is a "void *".

+ ptr = ((char *) shmem) + info->shared_ctl_off +
+ info->shared_data_off;
+
+ if (!read_chunk(fpin, ptr,

Same here for read_chunk().

I think that's perfectly fine and that we should do the same in 0002 when
calling init_shmem_cb() for consistency.

Regards,

--
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com