In J601 p beta there was a bug fix where an implicit
conversion of an extended precision number y would
signal error if y was not infinite but is larger in
magnitude than the largest finite 64-bit floating
point number, approximately 1e308.  In this case,

   xfmt cat2 1000
|domain error: xfmt
|   c=.1j1 1 1,(1e12    <:y)#1 1

y is cat2 1000, approximately 2.05e597, and so an
error is signalled.

xfmt needs to be updated to be one of:
   c=.1j1 1 1,(1e12<:x:^:_1 y)#1 1
   c=.1j1 1 1,((10^12x)<:y)#1 1
   c=.1j1 1 1,(12<#t)#1 1



----- Original Message -----
From: Fred Bone <[EMAIL PROTECTED]>
Date: Tuesday, August 1, 2006 2:48 am
Subject: Re: [Jbeta] J601 xfmt error on Catalan 1000

> On 31 Jul 2006 at 18:58, Ernst Goetze said:
> 
> > Sorry guys, but I was just (innocently) working my way through the
> > Catalan number Lab and got an error (probably minor) on xfmt. I'm
> > using 32-bit Darwin, bash kernel, j601binpbeta_darwin. Please e-mail
> > me if you need more info. Sincerely yours, Ernst.
> > 
> >    xfmt cat2 1000
> > |domain error: xfmt
> > |   c=.1j1 1 1,(1e12    <:y)#1 1
> 
> Confirmed (Win2KProSP4, J601p).
> 
> Didn't happen on J601o.


----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to