Closed noblem closed 7 years ago
I started to appraoch this by shrinking the test suite down from "mirror everything" to "mirror just 3-5 samples for 2-3 cohorts". This will allow us to checking in baseline regression data for test comparison.
I shrunk it further to just 2 patient cases for 4 cohorts, so that the initial download takes 5-10 mins. Next will be to MD5sum the files and compare to our test baselines
This is done: the suite of regression tests is now quite comprehensive (smoke test, mirror, dice, loadfile and report) and compared against a series of baseline file lists and MD5 sums in tests/baseline.
Running "make test" issues a considerable amount of mirror/dice/etc activity. But inspection of the Makefile leads me to believe that the artifacts of this are not validated against a stored set of baseline test output.