Re: Re: BUG #12885: The result of casting a double to an integer depends on the database version

Michael Paquier <michael(dot)paquier(at)gmail(dot)com> writes:
> On Thu, Mar 26, 2015 at 9:19 AM, Andrew Gierth
> <andrew(at)tao11(dot)riddles(dot)org(dot)uk> wrote:
> "Tom" == Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:
>>> Tom> BTW, if memory serves we also have a pretty lazy rounding
>>> Tom> implementation for the numeric datatype. I wonder if now would be
>>> Tom> a good time to upgrade that to be round-to-nearest-even as well.

>> A data point: there have been occasional complaints on IRC about the
>> fact that numeric rounding is not round-to-even, but obviously it's
>> harder to tell if anyone is relying on the current behavior.

> Tom, if there is a patch showing up soon, would you integrate it? I
> imagine that it would be good to have the same behavior for a maximum
> of datatypes in 9.5 now that src/port/rint.c is more compliant.

Yes, that's what I was thinking --- if we're changing this in 9.5 for
float (on Windows and whatever other platforms use rint.c) then it would
be sensible to fix numeric at the same time. I will commit a patch if
it shows up, but do not have time to write one myself.

regards, tom lane