Remove hard-coded du tests

[jeromekelleher]

Current tests/test_core.py::test_du test is very stupid. Refactor.

[jeromekelleher]

This is more complicated that I thought, and I now realise there's a good reason I did things the way I did in the first place. As we see in #331, for some reason getting the du results on CI to agree is hard. It works fine on my machine, so I can only assume it's something to do with the file system on the CI runners. I've tried moving the files to /tmp, but that didn't work either.

I would be most grateful if someone could figure this one out!

[tomwhite]

What we really want to do is test the implementation of du. So test it on a fixed set of files, perhaps generated in the test - to avoid the problem of having to update the numbers every time we add more test data.

But Unix du is problematic in itself as a measure of file sizes since it depends on the way data is laid out by the underlying filesystem (see e.g. https://unix.stackexchange.com/a/106290), so perhaps we shouldn't worry too much about trying to get an exact bytes match - the test could have a tolerance? Or have you tried using du --apparent-size?