Test Code coverage
It seems that Fleur has a number of bugs, that are not detected by our tests, which makes it a frustrating experience for users. To combat this I believe we should improve our tests. I have a few ideas, but I would also like to hear what you think:
Lots of more tests. If this is to slow, we could split this into a core set (run on every commit) and an extended set (run over night)
Using coverage tools such as gcov https://en.wikipedia.org/wiki/Gcov to determine and improve our test coverage. We probably test certain sections of our code 10 times, while we neglect others completely.
Unit tests. I think certain parts of the code can be tested independently of each other (e.g. all the different eigenvalue calculations). This one is more tricky, because Fortran doesn't have well maintained tools here (that I know of).
Before I dive into a fun weekend of figuring out gcov, I wanted to hear if this is something that is needed & wanted.