array[::-1] size reported as 0

[nbecker]

I have a function which accepts nd::Array<complex,1,0> const&. I get the size from
boost::size()

I just noticed that when I pass it a python array
x[::-1]

then boost::size(x) -> 0!

Is this expected? I find it hard to believe that I've been using ndarray for years and never saw this, since I have lots of code that works this way (accepts nd::Array<T,1,0> and uses boost::size)

[TallJimbo]

I don't think there are any tests for negative strides in ndarray, so I wouldn't be surprised if there is code somewhere that assumes that strides are positive.