From: Tom Lane <tgl@sss.pgh.pa.us>
To: Peter Eisentraut <peter.eisentraut@enterprisedb.com>
cc: Andres Freund <andres@anarazel.de>,
        pgsql-hackers <pgsql-hackers@postgresql.org>
Subject: Re: libpq error message refactoring
In-reply-to: <d5dddca7-9bdc-ee18-1ec9-ee21f8302d4a@enterprisedb.com>
References: <7c0232ef-7b44-68db-599d-b327d0640a77@enterprisedb.com>
 <20220922154225.q4rtgzzsq6pd4f5x@awork3.anarazel.de>
 <d5dddca7-9bdc-ee18-1ec9-ee21f8302d4a@enterprisedb.com>
Comments: In-reply-to Peter Eisentraut <peter.eisentraut@enterprisedb.com>
	message dated "Thu, 22 Sep 2022 22:00:00 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2032644.1663900649.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Thu, 22 Sep 2022 22:37:29 -0400
Message-ID: <2032645.1663900649@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/2032645.1663900649%40sss.pgh.pa.us>
Precedence: bulk

Peter Eisentraut <peter.eisentraut@enterprisedb.com> writes:
> On 22.09.22 17:42, Andres Freund wrote:
>> It's not the cause of this failure, I think, but doesn't appendPQExpBuf=
ferVA
>> need to be added to exports.txt?

> I don't want to make that function available to users of libpq, just use=
 =

> it inside libpq across .c files.  Is there no visibility level for that?=
 =


Should "just work", I should think.  I agree with Andres that that's
not the cause of the build failure.  I wonder if somehow the failing
links are picking up the wrong libpq.a.

Separately from that: is it really okay to delegate use of a va_list
argument like that?  The other call paths of
appendPQExpBufferVA[_internal] write va_start/va_end directly around it,
not somewhere else in the call chain.  I'm too tired to language-lawyer
out what happens when you do it like this, but I'm suspecting that it's
not well-defined portable behavior.

I think what you probably need to do is export appendPQExpBufferVA
as-is and require libpq_append_error to provide the error loop.

			regards, tom lane