ica.get_sources seems to be broken

here is the script to reproduce:

import mne
from mne.preprocessing import ICA
from mne.datasets import sample

data_path = sample.data_path()
raw_fname = data_path + '/MEG/sample/sample_audvis_filt-0-40_raw.fif'

raw = mne.io.read_raw_fif(raw_fname)
raw.pick_types(meg=False, eeg=True, exclude='bads', stim=True).load_data()

ica = ICA(n_components=0.95, method='fastica').fit(raw)

ica_sources = ica.get_sources(raw, start=20, stop=40)
data, times = ica_sources[:]

ica_sources.plot()
assert data.shape == times.shape
assert abs(times[0] - 20) < 1

I get:

ValueError: shape mismatch: objects cannot be broadcast to a single shape

[larsoner]

This appears to be fixed by #8801

[cbrnr]

Good suggestion! tmin and tmax should always refer to seconds, and start/stop should always refer to samples.