fix integer datetime division rounding error
The attached patch seems to fix the rounding error that is causing
regression failures on machines with integer datetimes. (Source of error
discovered by Andrew@Supernews).ISTM this code needs to be given some
careful analysis - I know it makes my head spin reading it.
cheers
andrew
Attachments:
ts.patchtext/x-patch; name=ts.patchDownload+2-2
Andrew Dunstan wrote:
The attached patch seems to fix the rounding error that is causing
regression failures on machines with integer datetimes. (Source of error
discovered by Andrew@Supernews).ISTM this code needs to be given some
careful analysis - I know it makes my head spin reading it.
Ah, brilliant! I knew I was missing something fundamental, and the use
of rint() was it. Strangely enough, the 8.0 code uses rint() in that
function, but for floating point intervals, and the code was buggy,
generating negative time values for division.
Patch attached and applied. I also improved the interval multiplication
code.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073Attachments:
/bjm/difftext/plainDownload+48-53
This fixes the problem for me.
Thanks,
-rocco
Show quoted text
-----Original Message-----
From: Bruce Momjian [mailto:pgman@candle.pha.pa.us]
Sent: Sunday, July 24, 2005 12:37 AM
To: Andrew Dunstan
Cc: Patches (PostgreSQL); Rocco Altier
Subject: Re: [PATCHES] fix integer datetime division rounding errorAndrew Dunstan wrote:
The attached patch seems to fix the rounding error that is causing
regression failures on machines with integer datetimes.(Source of error
discovered by Andrew@Supernews).ISTM this code needs to be
given some
careful analysis - I know it makes my head spin reading it.
Ah, brilliant! I knew I was missing something fundamental,
and the use
of rint() was it. Strangely enough, the 8.0 code uses rint() in that
function, but for floating point intervals, and the code was buggy,
generating negative time values for division.Patch attached and applied. I also improved the interval
multiplication
code.-- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
Import Notes
Resolved by subject fallback