| From: | jie wang <jugierwang(at)gmail(dot)com> |
|---|---|
| To: | Ning Sun <classicning(at)gmail(dot)com> |
| Cc: | pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: Add ParameterDescription message to libpq frontend long message types |
| Date: | 2026-04-02 01:13:57 |
| Message-ID: | CAJnZyeBGwBCixjdFdS4MVVYHYDsLjNuGaqFE_z88DZ615TST+A@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Ning Sun <classicning(at)gmail(dot)com> 于2026年4月2日周四 08:52写道:
> Hi,
>
> I'm maintaining a Rust library pgwire to implement postgres wire
> protocol in rust. While doing a corner case test, I noticed the
> inconsistency for ParameterDescription between backend and frontend.
>
> The backend allows up to 65535 parameters in a prepared statement. But
> when running Describe on the statement, there is a size limit of 30000
> bytes for ParameterDescription on the frontend. This means we can only
> describe statements with at most ~7500 parameters. For statements exceed
> the limit, it ends up with error about the message size.
>
> This patch simply adds ParameterDescription to VALID_LONG_MESSAGE_TYPE
> whitelist to remove the cap.
>
>
Hi,
Based on my understanding, this approach is safe for the following reasons:
1. The message format is clearly defined and has been properly handled.
2. Similar message types (such as RowDescription) are already allowed to
exceed this limit.
3. No changes have been introduced at the protocol level.
Thanks!
wang jie
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Andreas Karlsson | 2026-04-02 01:16:33 | Re: Check some unchecked fclose() results |
| Previous Message | David Rowley | 2026-04-02 01:12:55 | Re: More speedups for tuple deformation |