Re: Incorrect comment for memset() on pgssHashKey?

On 27/06/2023 07:32, Japin Li wrote:
> 
> Hi,
> 
> Commit 6b4d23feef introduces a toplevel field in pgssHashKey, which leads
> padding.  In pgss_store(), it comments that memset() is required when
> pgssHashKey is without padding only.
> 
> @@ -1224,9 +1227,14 @@ pgss_store(const char *query, uint64 queryId,
>       query = CleanQuerytext(query, &query_location, &query_len);
> 
>       /* Set up key for hashtable search */
> +
> +    /* memset() is required when pgssHashKey is without padding only */
> +    memset(&key, 0, sizeof(pgssHashKey));
> +
>       key.userid = GetUserId();
>       key.dbid = MyDatabaseId;
>       key.queryid = queryId;
> +    key.toplevel = (exec_nested_level == 0);
> 
>       /* Lookup the hash table entry with shared lock. */
>       LWLockAcquire(pgss->lock, LW_SHARED);
> 
> However, we need memset() only when pgssHashKey has padding, right?

Yep. I changed the comment to just "clear padding". Thanks!

-- 
Heikki Linnakangas
Neon (https://neon.tech)