<p>On Jul 9, 2011 9:46 AM, "Dylan Smith" <<a href="mailto:dylan.ah.smith@gmail.com">dylan.ah.smith@gmail.com</a>> wrote:<br>
><br>
> On Sat, Jul 9, 2011 at 10:16 AM, Misha Koshelev <<a href="mailto:misha680@gmail.com">misha680@gmail.com</a>> wrote:<br>
> > Does the attached patch work on MS Visual Studio 2010? This passes all tests.<br>
><br>
> Ya, it should work. I was only building test applications using the<br>
> cl from the platform SDK in wine; I haven't been using it for make<br>
> crosstest.<br>
><br>
> > Your version does not seem to convert properly to +NAN on some<br>
> > platforms (e.g., Linux) and results in some test failures.<br>
><br>
> That doesn't make sense, because I am primarily using gcc in Ubuntu.<br>
> Did you try it with the parentheses, because unary minus has higher<br>
> precedence than division:<br>
><br>
> -0.0f/zero => -nan (0xffc00000)<br>
> -(0.0f/zero) => nan (0x7fc00000)</p>
<p>Hi Dylan:</p>
<p>I am sorry I miswrote about Ubuntu. However, there were some test failures that occurred on Winetestbot when I tried to use -(0.0f/zero) for nan.</p>
<p>Thank you<br>
Misha</p>