From: Tom Lane <tgl@sss.pgh.pa.us>
To: "'alvherre@alvh.no-ip.org'" <alvherre@alvh.no-ip.org>
cc: "iwata.aya@fujitsu.com" <iwata.aya@fujitsu.com>,
        "tsunakawa.takay@fujitsu.com" <tsunakawa.takay@fujitsu.com>,
        "'Kyotaro Horiguchi'" <horikyota.ntt@gmail.com>,
        "k.jamison@fujitsu.com" <k.jamison@fujitsu.com>,
        "pgsql-hackers@lists.postgresql.org"
 <pgsql-hackers@lists.postgresql.org>,
        Andrew Dunstan <andrew@dunslane.net>
Subject: Re: libpq debug log
In-reply-to: <3336855.1617229186@sss.pgh.pa.us>
References: <20210331220920.GA23882@alvherre.pgsql>
 <3336855.1617229186@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Wed, 31 Mar 2021 18:19:46 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <3337421.1617229905.1@sss.pgh.pa.us>
Date: Wed, 31 Mar 2021 18:31:45 -0400
Message-ID: <3337422.1617229905@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/3337422.1617229905%40sss.pgh.pa.us>
Precedence: bulk

I wrote:
> What I suspect is some Windows dependency in the way that
> 001_libpq_pipeline.pl is setting up the trace output files.

While this may have little to do with drongo's issue,
I'm going to take exception to this bit that I see that
the patch added to PQtrace():

	/* Make the trace stream line-buffered */
	setvbuf(debug_port, NULL, _IOLBF, 0);

I do not like that on two grounds:

1. The trace file handle belongs to the calling application,
not to libpq.  I do not think libpq has any business editorializing
on the file's settings.

2. POSIX says that setvbuf() must be done before any other
operation on the file.  Since we have no way to know whether
any output has already been written to the trace file, the
above is a spec violation.


... and as for an actual explanation for drongo's issue,
I'm sort of wondering why libpq_pipeline.c is opening the
trace file in "w+" mode rather than vanilla "w" mode.
That seems unnecessary, and maybe it doesn't work so well
on Windows.

			regards, tom lane