Host unit tests moved - Githubissues

This completes the changeover of all oops/hosts unit tests to tests/hosts. I needed it so that I could more clearly identify the code coverage of various modules once you exclude the test code itself. Summary of changes...

In oops/hosts, I removed unit tests from every module. This includes gold_master.py and standard_obs.pywherever they appeared.
Previously, tests/hosts was just a copy of oops/hosts, because that's where I had left things when I moved all the other unit tests. Now, each file just contains the unit tests. This is also where you find all files gold_master.py, standard_obs.py, and unittester.py. Conceptually, things are set up so that tests/hosts/mission/instrument/unittester.py tests the given instrument, tests/hosts/mission/unittester.py tests all the instruments on the mission, and tests/hosts/unittester.py tests all hosts.
For my own convenience, I added tests/unittester_with_hosts.py, which runs all the OOPS unit tests including hosts.
Where there were random files or unused bits of code in oops/hosts that had once been used for testing but are now unused, I moved them to tests/hosts but they are still unused.
In scripts/automated_tests/oops_main_test.sh, I changed the location for the hosts tests from oops/hosts/unittester.py to tests/hosts/unittester.py.
In tests/*.py, I made sure each setUp() function loaded the kernels that are needed for that test. The philosophy is that each test is responsible for its own kernels to be present (and must-be-excluded kernels to be absent). If a test does not do that, outcomes could be unpredictable. For the most part, tearDown() functions are not implemented (or I left them in there with "pass"), but I use them to undo anything especially pathological I did in the setUp().
I added _undefine_solar_system() to Body, which wipes the internal cache; it's needed for debugging.
Similarly, I added _unload_all() to spicedb.py, which unloads all kernels. Again, needed for debugging.
Bugfix: the reference to AliasFrame in body.py didn't work quite as expected. I changed it to Frame.as_frame().
Bugfix:hosts/galileo/__init__.py was not loading the Galileo Frames kernel, which means I don't quite know how it worked before--apparently a leftover kernel load from somewhere else was hiding this bug.
There's a small discrepancy between some gold-master calculations on Windows vs. Mac/Linux. We have seen it before and it remains unexplained. Nevertheless, I needed to update the error threshold on some Galileo Europa tests to get this to pass for Windows. Several angles (center phase, orbital longitude) had discrepancies of 0.0012-0.0014 degrees, whereas the default error threshold was 0.001.

SETI / rms-oops

Host unit tests moved #131

Codecov Report