Re: Removing support for FE/BE protocol version 2

From: Tatsuo Ishii <ishii(at)postgresql(dot)org>
To: nadav(at)tailorbrands(dot)com
Cc: pgpool-hackers(at)lists(dot)postgresql(dot)org
Subject: Re: Removing support for FE/BE protocol version 2
Date: 2026-10-13 03:53:58
Message-ID: 20261013.125358.1171170128787374410.ishii@postgresql.org
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgpool-hackers

Hi Nadav,

> Hi,
>
> Thanks for v4. I applied it on master. The build is clean (no new
> warnings, also with -Wextra and -Wunused-parameter), the full
> regression suite passes on PG16 (80/80), and 045 passes. On the wire
> the reply is 'E' + "FATAL: unsupported frontend protocol 2.0\n" +
> '\0', and the log has a single FATAL line. All the v3 points are fixed.
>
> Since we are now removing the protoMajor-specific code entirely, here
> is what is still left. None of it is a bug:
>
> - pool_hba.c (around lines 1342 and 1380): two
> "frontend->protoVersion == PROTO_MAJOR_V3" checks that are now
> always true. This file was on your original list.

Oops. I forgot it. Will fix.

> - child.c (around line 1031): "new_sp->major == PROTO_MAJOR_V3 &&
> sp->application_name".

Will fix.

> - The MAJOR() macro in pool.h has no callers now. So
> pool_get_major_version(), pool_set_major_version() (called only from
> child.c) and session_context->major are dead too, as is the
> pool_get_major_version() stub in the 010 test.

Agreed on removing these.

> - pool_get_cp() and pool_discard_cp() still take protoMajor and compare
> it with sp->major. It is harmless, but the argument could be dropped
> along with the rest.

I would like to keep the protoMajor for now.

Now clients can use protocol 3.0 or 3.2. So comparing protoMajor in
pool_get_cp() is not sufficient. It needs to compare protoMinor as
well. I would like to fix this in a separate patch since this is not
related to dropping v2 support.

> - pool.h: sbuf/sbufsz in POOL_CONNECTION are now only initialized and
> freed in pool_stream.c. Their only user was pool_read_string(), which
> is gone (the comment still says "buffer for pool_read_string").

I will remove them.

> - pgproto/read.c:411 still uses %m.

Will fix.

> - pgproto/send.c write_it(): "int errsave = errno;" needs pgindent
> (alignment and a blank line after the declaration).

Yeah. There are tons of places that need pgindent. We will run
pgindent on whole source files before GA is out.

> - pool_proto_modules.h: the header comment ends with a dangling comma
> ("pool_proto_modules.c,"), and there is a double blank line after
> log_backend_messages().

Will fix.

> Other than these, the patches look good to me. Doing the items above in
> a follow-up commit is probably fine too.

I have left the pool_get_cp() and pool_discard_cp() issue as described
above. Other than that, attached v5 should address all the problems
you pointed out.

Thank you so much for your review!
--
Tatsuo Ishii
SRA OSS K.K.
English: http://www.sraoss.co.jp/index_en/
Japanese:http://www.sraoss.co.jp

Attachment Content-Type Size
v5-0001-Reject-v2-protocol-startup-packet.patch application/octet-stream 193.0 KB
v5-0002-Modify-pgproto-as-a-test-driver.patch application/octet-stream 18.5 KB

In response to

Responses

Browse pgpool-hackers by date

  From Date Subject
Next Message Nadav Shatz 2026-10-13 11:49:43 Re: Removing support for FE/BE protocol version 2
Previous Message Nadav Shatz 2026-10-12 09:10:58 Re: Removing support for FE/BE protocol version 2