Thomas Kluyver

Possibly. Or maybe it didn't have matplotlib installed in that env - the `--matplotlib=inline` option makes it try to load matplotlib on start.

I've opened #20 to avoid relying on matplotlib

Some extra packages are needed for testing, and some of the tests are not meant to be run normally (we're testing a testing tool, so naturally it has some sample...

I think that fix sounds reasonable if you want to send a PR. We don't have any good option to compare PDFs, unless nbdime can do that now (@vidartf?).

Is the cell that it crashes on doing anything specific? Some operations can be much slower on Travis than on a normal system, though a few seconds to 10 minutes...

It might be dependent on the kernel reporting when the output from a cell is finished, but the IPython kernel should do that correctly. I don't know why that code...

IIRC, differences in images are ignored unless you're using the nbdime integration, because when you're creating a plot or something, even running the exact same code often doesn't produce a...

I'd be OK with doing that based on cell metadata. Using comments for this feels like a bit too much magic parsing.

At present, no. nbval works only on the outputs that the kernel sends to the notebook, not the Python objects inside the kernel. #26 is about giving it a way...

I'm not so sure on the parallel stuff, but the marker comments for nbval can be anywhere in the cell. You can also use cell tags instead of comments: https://nbviewer.jupyter.org/github/computationalmodelling/nbval/blob/0.9.1/docs/source/index.ipynb#Using-tags-instead-of-comments