Re: [xslt] [vincent@vinc17.org: Bug#206549: xsltproc: Incorrect result due to computations in extended precision]

[Daniel Veillard <veillard redhat com>]

On Thu, Sep 04, 2003 at 11:08:57AM -0500, Ardo van Rangelrooij wrote:
> Hi,
> 
> Also the bug report below has been filed against the Debian package of
> libxslt.  Although the bug report is against 1.0.30, I ran the provided
> sample code through 1.0.32 and the problem still exists.  And I also used
> libxml2 2.5.10.
> 
> Is this also a bug and do I need to file it with bugzilla, or is this a
> not-yet supported feature?
[...]
>     <xsl:when test="9007199254740992 + 1.00001 = 9007199254740992">
[...]
> On my machine, this gives:
> 
> $ xsltproc test.xsl test.xml
> equal
> 
> But it seems that the XPath recommendation requires the computations
> to be performed in double precision only; in this case, "different"
> should have been output (i.e. when double rounding is avoided).

  XPath uses double ! Look at the code if you don't believe me.
If someone else gets a different answer then that means that double
computations occurs in a different way and lead to different results
hard to blame libxslt for that....

> As another user gets "different" on an x86 machine, I'm filling the
> bug now. This may be a Debian/unstable-only bug. You can have a look
> at the thread:

  No. If libxslt bugs must be discussed, the reporting rules are 
clear http://xmlsoft.org/XSLT/bugs.html ... I'm not gonna search
for a discussion where I have only a SMTP message ID to look it up !

  BTW thanks for acting as a gateway, but you should educate them
to go report upstream and discuss there once the bug has been identified,

Daniel

-- 
Daniel Veillard      | Red Hat Network https://rhn.redhat.com/
veillard@redhat.com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/