Re: [xml] The NaN dilemma

[Daniel Veillard <veillard redhat com>]

  thanks for the details !

On Tue, Jul 31, 2001 at 06:14:11PM +0000, Bjorn Reese wrote:
>  2. We can use the C99 nan() and isnan() functions. This of course
>     assumes their availability. However, we still have to ensure that
>     we do not perform any arithmetic operations on NaN elsewhere in
>     the library.

  I would shoot for this one, if they are not available AND the 
Invalid Operation flag have been changed ... then there are troubles.
Basically we control only a part of the framework, so my suggestion
is to try to cascade all possibilies in order to minimize the probablility
of getting to the problem. 

  1/ nan() and isnan()
  if not
  2/ if PCcee or being able to detect the hardware use the predefined
     value
  else
  3/ use the existing code (i.e. run the divide).

If people are really interested in being sure that the port to a given
platform is rock solid, they just have to provide detection rules and
the predefined value for phase 2.

It's gonna be an ugly piece of code, but heck when the hardware or
underlying OS is broken, then it's better to cope with it at upper layers.
Adding your explanation in the comment before that section will help
whoever have to maintain this in 10 years.

 Daniel

-- 
Daniel Veillard      | Red Hat Network http://redhat.com/products/network/
veillard redhat com  | libxml Gnome XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/