From: Tom Lane <tgl@sss.pgh.pa.us>
To: Mark Dilger <mark.dilger@enterprisedb.com>
cc: Thomas Munro <thomas.munro@gmail.com>,
        PostgreSQL-development <pgsql-hackers@postgresql.org>,
        Joshua Brindle <joshua.brindle@crunchydata.com>,
        Robert Haas <robertmhaas@gmail.com>,
        Andrew Dunstan <andrew@dunslane.net>, Jeff Davis <pgsql@j-davis.com>,
        Joe Conway <joe@crunchydata.com>
Subject: Re: New Object Access Type hooks
In-reply-to: <B188F52A-69B8-46E3-A32B-2FAFD9447FEF@enterprisedb.com>
References: <47F87A0E-C0E5-43A6-89F6-D403F2B45175@enterprisedb.com> <CA+hUKG+LOza8iSLiv6ymn=0R8S0x-R7KaqZmYfY+_b_cubcDOg@mail.gmail.com> <17BD82D7-49AC-40C9-8204-E7ADD30321A0@enterprisedb.com> <1918971.1649091790@sss.pgh.pa.us> <1922986.1649094077@sss.pgh.pa.us> <1923706.1649094673@sss.pgh.pa.us> <AACAF464-092B-4497-ADA2-AE1B122FAC6A@enterprisedb.com> <1925660.1649095626@sss.pgh.pa.us> <2BD8BDA8-A963-4D64-8B69-8F7EDF9AEA2D@enterprisedb.com> <1974785.1649098875@sss.pgh.pa.us> <1975199.1649099110@sss.pgh.pa.us> <B188F52A-69B8-46E3-A32B-2FAFD9447FEF@enterprisedb.com>
Comments: In-reply-to Mark Dilger <mark.dilger@enterprisedb.com>
	message dated "Mon, 04 Apr 2022 13:28:17 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2163176.1649105224.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Mon, 04 Apr 2022 16:47:04 -0400
Message-ID: <2163177.1649105224@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/2163177.1649105224%40sss.pgh.pa.us>
Precedence: bulk

Mark Dilger <mark.dilger@enterprisedb.com> writes:
>> On Apr 4, 2022, at 12:05 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> So scratch that.  Maybe we'd better add "could not send data to server"
>> to the regex?

> If it fails in pqsecure_raw_write(), you get either "server closed the c=
onnection unexpectedly" or "could not send data to server".  Do we need to=
 support pgtls_write() or pg_GSS_write(), which have different error messa=
ges?

Don't see why, since this test sets up a new cluster in which neither
is enabled.

> Is it possible that pgFlush will call pqSendSome which calls pqReadData =
before trying to write anything, and get back a "could not receive data fr=
om server" from pqsecure_raw_read()?

Yeah, it's plausible to get a failure on either the write or read side
depending on timing.

Perhaps libpq should be trying harder to make those cases look alike, but
this test is about server behavior not libpq behavior, so I'm inclined
to just make it lax.

			regards, tom lane