Date: Wed, 28 Oct 2020 16:46:54 -0400
From: Bruce Momjian <bruce@momjian.us>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: kensuke.okamura@kantei.co.jp, pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #16688: psql removes only LF without CR at end of backquotes
 on windows.
Message-ID: <20201028204654.GD3239@momjian.us>
References: <16688-c649c7b69cd7e6f8@postgresql.org>
 <1224124.1603855625@sss.pgh.pa.us>
 <20201028032942.GA21830@momjian.us>
 <1441558.1603910379@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1441558.1603910379@sss.pgh.pa.us>
User-Agent: Mutt/1.10.1 (2018-07-13)
Precedence: bulk

On Wed, Oct 28, 2020 at 02:39:39PM -0400, Tom Lane wrote:
> Bruce Momjian <bruce@momjian.us> writes:
> > On Tue, Oct 27, 2020 at 11:27:05PM -0400, Tom Lane wrote:
> >> But rather than mess with that newline-chomping code, I'm
> >> inclined to wonder why we're using PG_BINARY_R for input
> >> that we clearly expect to be textual.  Most of our other
> >> popen's do not do that.
> >> 
> >> Bruce, this seems to date to 98e9775a3 ... don't suppose
> >> you remember that?  I see the point about control-Z in text
> >> files, but I wonder how plausible that is for popen cases.
> 
> > It seems safe for popen to use TEXT mode, and it might help with
> > encoding conversion.  I don't think I made a popen distinction when I
> > was working in this area.
> 
> I double-checked, and verified that our only other popen() calls
> that use binary mode are dealing with COPY data.  It seems appropriate
> to do so in that case, partly because the data might indeed be binary
> and partly because we have adequate newline recognition logic in copy.c.
> But it does seem best to uniformly use plain "r" or "w" for other
> popen's.  So I've fixed this that way.

Thanks.

-- 
  Bruce Momjian  <bruce@momjian.us>        https://momjian.us
  EnterpriseDB                             https://enterprisedb.com

  The usefulness of a cup is in its emptiness, Bruce Lee