Closed rboston628 closed 5 months ago
One issue discovered in this was with the matrix inversion, which was surprisingly not caught by the tests. Incorrect choices in the indices of the re-substitution part were causing some mode calculations to fail.
The code has individually re-written rootfinding methods all throughout it. This is not a desirable situation, and the better thing would be a single library with rootfinding methods that are all called where needed.
To do this, some careful testing is necessary, to ensure none of the routines relying on these methods accidentally break.
Testing was starting in PR #13. These changes are only to models covered by tests.