Keith Whitwell <[EMAIL PROTECTED]> writes:
> [EMAIL PROTECTED] wrote:
>
> > I did apply it but it did not solve the problem of the floating exception on
> > my VMS system.
>
> Is the execption still apparent in the transform_points_3d_no_rot, or
> has it moved to other parts? Is 0x0 a valid float on VMS? I expect
> there are other arrays which would require initialization - eg the
> normal arrays.
You can duplicate the bug on a Pentium if you put
`__setfpucw(_FPU_DEFAULT & ~0x1f);' after the START_FAST_MATH in
gl_transform_vb(). I haven't tested your patch yet.
Jouk, does it work if you initialize the clipped vProj coordinates in
cliptest_points4()? Here's the patch.
*** clip_tmp.h.old Mon Mar 8 15:40:00 1999
--- clip_tmp.h Mon Mar 8 15:57:58 1999
***************
*** 43,48 ****
--- 43,52 ----
tmpAndMask &= mask;
clipMask[i] = mask;
tmpOrMask |= mask;
+ vProj[i][3] = 0.0;
+ vProj[i][0] = 0.0;
+ vProj[i][1] = 0.0;
+ vProj[i][2] = 0.0;
#if 0
} else if (cw == 0) {
/* only get here for 0,0,0,0 - not really sure what
