https://issues.dlang.org/show_bug.cgi?id=20451
--- Comment #2 from jacob <[email protected]> --- As for why it works on 64-bit and Linux, they don't use the FPU there. So essentially the pow() function is returning a double in a XMM register. So if you run the above code I posted with -m64, it does a check using the FPU. Which causes the assert to fail, cause of course it is now testing a double (from pow()) and a real. The whole situation is really messed up and it seems that it is flip flopping between double and real cause in new hardware nothing actually supports 80-bit floats it's dead. Really real should mean 128-bit float, it is the next logical step over 80-bit floats. --
