Closed
Description
When using scikit-learn 0.16.1 with numpy 1.9.2 on i686 system certain doctests fail. They don't seem to fail on x86-64 systems though. I haven't tested with master.
In all cases what's missing is the dtype of the array. Here's an example:
File "/tmp/nix-build-python3.4-scikit-learn-0.16.1.drv-0/scikit-learn-0.16.1/sklearn/preprocessing/data.py", line 1034, in sklearn.preprocessing.data.OneHotEncoder
Failed example:
enc.feature_indices_
Expected:
array([0, 2, 5, 9])
Got:
array([0, 2, 5, 9], dtype=int32)
>> raise self.failureException(self.format_failure(<_io.StringIO object at 0xf0915b6c>.getvalue()))and here's the log
http://hydra.nixos.org/build/25129295/nixlog/1/raw