RE: [xml] NAN problems on Win32

["Vakoc, Mark" <Mark_Vakoc jdedwards com>]

> > few reports of crashes of my program from various people.  Only the
latest
> > version of my program has the crash handler that generates that log with
the
> > stack trace, so I cannot be sure that they had the exact same problem.
If I

> The crash handler may very well be the problem.

The crash handler (installed on Win32 using SetUnhandledExceptionFilter) was
created in a recent release to generate logs when a crash occurs.  Prior to
that (when no exception filters were installed) the program would just exit.

> trio_nan() uses infinity/infinity as its fallback solution to generate
NaN.
> You have installed a trap handler, and you are receiving an invalid
operation
> trap.

No.  If that were the case the problem would occur in xmlXPathInit() where
xmlXPathNAN is assigned.  Also, the compiled version uses the USE_IEEE_754
method (which assigns a hardcoded value).  

I'm pretty sure NAN has nothing to do with this, since it occurs in one
instance on the conversion from int to double.  Every instance that i know
the offending machine has been a Dell GX150.  I have some preliminary
numbers and the program runs fine on over 1100 unique machines, with 8
people experiencing problems, and 3 of them (the only hardware descriptions
I have) being that same Dell GX150.  I'm sure that there are more crashes
that I am not aware of however.

The only advise I have seen related to this has been a sketchy usenet post
advising against doing conversions directly in function calls, i.e. instead
of
xmlXPathNewFloat((double) nodeset->nodeNr)

to do a 
double d = (double) nodeset->nodeNr;
xmlXPathNewFloat(d)

Seems stupid to me, but I will give it a try Monday when I get access to
their machine to see if this makes a difference.  Will let you know what I
find.