Closed mcabbott closed 4 years ago
I'm a bit torn here. This is simple enough a fix to just apply it and be happy to have the issue fixed. OTOH, I prefer to only have random tests in cases where some property should really hold for any input data, which is not the case here. Thus, I think I'd prefer to just use fixed, explicitly given values for A
, x
, and y
, I guess. Or alternative, restrict the randomly generated data to numerically benign cases, but I don't know how to achieve that.
This is why I didn't want to change change random behaviour for all the tests, but thought changing it for just these would be the quick fix.
Perhaps ideally you'd reject & replace random vectors too close to zero dot, or something? But that seems like too much work.
This is why I didn't want to change change random behaviour for all the tests, but thought changing it for just these would be the quick fix.
Yeah, that makes sense.
Closes #712, I hope.