opengl and borland C++ cause floating point exceptions
Ryan C. Stallings
winehq.com at beergeek.net
Mon May 20 17:07:04 CDT 2002
No I don't have 'proof'. It was just a guess (and probably a bad one). Maybe
Microsoft changed their opengl.dll to not produce the exceptions, or maybe in
the DLL they know where the exceptions will be so they wrap only those
specific portions in the mask/unmask pair. It doesn't look like the viewer
specifically disables the floating point exception, so it must no longer be
necessary to do so with Borland and OpenGL. I was just trying to guess why,
so wine could emulate what windows does.
On Monday 20 May 2002 03:33 pm, Lionel Ulmer wrote:
> On Mon, May 20, 2002 at 01:26:15PM -0400, Ryan C. Stallings wrote:
> > I am trying to get the neverwinter nights model viewer running under
> > wine. When viewing some models I start to get floating point errors. I
> > did a quick search on google and came up with the following post.
> Yeah, I have the same floating point errors here.
> > It looks like Microsoft
> > fixed this incompatibility by wrapping a call to turn off floating point
> > exceptions and then restore the previous mask around each opengl call.
> > Do I understand the problem correctly? Does this sound like a resonable
> > fix for wine too?
> You mean than in Windows, at each OpenGL call, the FP exceptions are
> masked and re-enabled at the end of the function ? That would seem very
> unoptimal. Do you have 'proof' of this :-) ?
More information about the wine-devel