From: Jeremy Kerr <jk@ozlabs.org>
To: pgsql-hackers@postgresql.org
Subject: Re: [RFC,PATCH] SIGPIPE masking in local socket connections
Date: Tue, 2 Jun 2009 23:52:01 +1000
User-Agent: KMail/1.11.2 (Linux/2.6.28-11-generic; KDE/4.2.2; i686; ; )
Cc: Tom Lane <tgl@sss.pgh.pa.us>
References: <1243914753.517210.131510259224.0.gpush@pingu>
	<10549.1243949503@sss.pgh.pa.us>
In-Reply-To: <10549.1243949503@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Type: Text/Plain;
  charset="iso-8859-15"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200906022352.01814.jk@ozlabs.org>

Tom,

> The consideration is that the application fails completely on server
> disconnect (because it gets SIGPIPE'd).  This was long ago deemed
> unacceptable, and we aren't likely to change our opinion on that.

OK, understood. I'm guessing MSG_NOSIGNAL on the send() isn't portable 
enough here?

> What disturbs me about your report is the suggestion that there are
> paths through that code that fail to protect against SIGPIPE.  If so,
> we need to fix that.

I just missed the comment that pqsecure_read may end up writing to the 
socket in the SSL case, so looks like all is fine here. We shouldn't see 
a SIGPIPE from the recv() alone.

Cheers,


Jeremy