Fixes for Octave

[gllmflndn]

These changes were necessary for compatibility with GNU Octave.

Bug reports have been submitted for the following:
fopen(-1)
and
strmatch('a',[])
which should return without error but code is cleaner anyway by not relying on these corner cases.

[agramfort]

can someone test that it still works with matlab?

ideally it would be nice to have travis running octave tests if their are users of this code.

[gllmflndn]

It would indeed be great to have Travis testing with Octave 4.2. I could only run two of the five tests because of the requirement to have access to the dataset from MNE_SAMPLE_DATASET_PATH.

[gllmflndn]

Also, @agramfort and @robertoostenveld, how is this repository synchronised with https://github.com/fieldtrip/fieldtrip/tree/master/external/mne ? Could FT use git submodules?

[robertoostenveld]

I only vaguely know git submodules and never used it. Right now I (or someone else) copies the mne-matlab code into the FT repo and commits.

[agramfort]

yes I do the sync manually you could use the mne-testing dataset used for testing mne-python on travis. PR more than welcome let me know if you need help

[larsoner]

ideally it would be nice to have travis running octave tests if their are users of this code.

We now have some basic unit tests, so I've fixed the merge conflict. I'll merge once CIs come back green, thanks in advance (and apologies for dropping the ball on this for years)!

[larsoner]

... actually it looks like this does break a test!