Comment 26 for bug 1406458

>> but is it better than getting the 0, that should be in the gradient, past the division error?

um, we should not be thinking of this as a division error. The expression 1/0 can be correctly called a division error. The expression 0/0 is more accurately called indeterminate, meaning we don't know what it is, meaning different renderers will treat it differently.
The whole purpose of this code is to produce a unit vector of length 1. You cannot do that if both x and y are 0.