On Wednesday, April 18, 2012 11:18:44 Ali Çehreli wrote: > On 04/18/2012 10:13 AM, Jonathan M Davis wrote: > > On Wednesday, April 18, 2012 19:04:12 SomeDude wrote: > >> On Wednesday, 18 April 2012 at 16:36:39 UTC, bearophile wrote: > >>> Ali: > >>>> That a thousandth time I have made that mistake and still have > >>>> not learned. :( Yes, .nan may not be compared with any other > >>>> value, including .nan. > >>> > >>> Today I'll present an enhancement request to remove this > >>> problem from D. > >>> > >>> Hugs, > >>> bearophile > >> > >> I don't see how this could be enhanced. > > > > It's by design. An enhancement request is a waste of time. > > Comparisons with > > > NaN _always_ return false regardless of what they're compared against > > - even > > > NaN. It's not going to change. > > > > - Jonathan M Davis > > It shouldn't be a problem to detect comparisons against literal .nan > values. The compiler can warn with "comparison is always false".
Yes, that would make sense. Heck, I'd be tempted to argue that using == with floating point values in general should be a warning, since that's pretty much never what you actually want, but that's probably not going to fly. But the behavior itself isn't going to change. - Jonathan M Davis