From: Andrew Gierth <andrew@tao11.riddles.org.uk>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: victor@magic.io,  pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #15519: Casting float4 into int4 gets the wrong sign instead
 of "integer out of range" error
In-Reply-To: <13162.1543014227@sss.pgh.pa.us> (Tom Lane's message of "Fri, 23
	Nov 2018 18:03:47 -0500")
Message-ID: <87zhtzqs9o.fsf@news-spur.riddles.org.uk>
References: <15519-4fc785b483201ff1@postgresql.org>
	<874lc7sb1a.fsf@news-spur.riddles.org.uk>
	<13162.1543014227@sss.pgh.pa.us>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/25.1 (berkeley-unix)
Date: Fri, 23 Nov 2018 23:42:50 +0000
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

>>>>> "Tom" == Tom Lane <tgl@sss.pgh.pa.us> writes:

 Tom>     /*
 Tom>      * Range check.  We must be careful here that the boundary values are
 Tom>      * expressed exactly in the appropriate float domain; we assume float4
 Tom>      * is going to round off INT_MAX to a power of 2.  Also note assumption
 Tom>      * that rint() will pass through a NaN or Inf unchanged.
 Tom>      */
 Tom>      if (unlikely(num < (float4) INT_MIN || num >= (float4) INT_MAX || isnan(num)))
 Tom>         ereport(...);

Looking around, another approach I've seen (which I like better) is to
use

    if (num < (float4)INT_MIN || num >= -(float4)INT_MIN || ...

which doesn't rely on assumptions about how the compiler is going to
round INT_MAX. (It does rely on two's complement representation of ints,
but that assumption is known to be safe.)

-- 
Andrew (irc:RhodiumToad)