Increase test coverage

[HereAround]

• Some error messages not tested in the new fetch/read polytope methods. (This should be easy, I think.)
• Also, we never test reading from files. Somehow, this should be tested.

@ariostas Any thoughts on the latter?

[ariostas]

I'm not very familiar with how GitHub testing works. However, reading from a file vs string uses the same code, so as long as the few lines that read the contents of the file are not broken then it should be fine.

[HereAround]

There are multiple sets of tests being executed:

1. Running the code,
2. Executing code that is used for the documentation,
3. Setting up and deploying (i.e. shipping to the website) the documentation.

Here, I am mostly thinking about 1. Those tests are spelled out in the /test/runtests.jl file. (The doc tests are all examples that you add for the docstrings.) In the case at hand, I would like to check that the few lines that read from a file work. From what I read in the code and you are saying, the file should have the same content as the string that we are currently handing over to the method, correct?

If yes, then we could indeed add this as a test file and use that to add test coverage for those lines too.

Minor related point: Unless I am mistaken, as of now, the code does not test if the exists and is readable. So this might backfire badly. Could be improved as we add tests for this.