Hi Thomas,
I think we still use that code, although I wouldn't be averse to "fixing"
the problem, if someone can come up with the appropriate fix. The last
time I checked, we couldn't use the values in std::numeric_limits on
Solaris, because they didn't yield the correct results.
Thanks!
Dave
|---------+----------------------------------->
| | [EMAIL PROTECTED]|
| | oftware.com |
| | |
| | 05/27/2004 06:14 AM |
| | Please respond to |
| | xalan-c-users |
|---------+----------------------------------->
>----------------------------------------------------------------------------------------------------------------------------------------|
|
|
| To: [email protected]
|
| cc: (bcc: David N Bertoni/Cambridge/IBM)
|
| Subject: Dynamic loading of Xalan library on Solaris
|
>----------------------------------------------------------------------------------------------------------------------------------------|
I was just curious to know if anyone knows if the issue below has been
fixed in the latest XalanC version or with a specific Solaris patch (since
it is really a Solaris problem at the end of the day).
The issue is a message (square root of negative number) printed on stderr
at XalanC library initialization time.
I did a few searches in the XalanC bug database without any success.
Here is the last comment from Dave Bertoni on the subject, exchanged end of
2002:
Well, that is a problem with the Xalan library, although it's really an
underlying bug in Sun's run-time.
Xalan uses the sqrt() function to determine the value of NaN when it starts
up by calling the function with a negative number. This is supposed to
fail silently and return the correct number according to the ANSI standard.
We have a similar problem on AIX which was fixed in the 1.4 release.
Unfortunately, nobody has ever reported the problem on Solaris, so I didn't
think it was a problem.
Thanks
Thomas