Closed dnadlinger closed 5 years ago
(This wasn't caught before because the fitting tests weren't actually executed on the CI system.)
Hm, it worked when I ran them on my machine... I'll look into it.
Fails on my local macOS box too (SciPy 1.2.1) – can let you have a look tomorrow if unreproducible on your end. (You can also SSH into CircleCI's machines, I believe.)
Did they fail before the latest sinusoid commit? I coult imagine that violating the "fourier" limit might mess lombscargle around.
Turns out the test is working fine. Relaxing the "Fourier" limit makes the fitting fail the test.
As per the title; specifically
test_delayed_split_data_with_t_dead
fails deterministically: https://circleci.com/gh/OxfordIonTrapGroup/oitg/84