Closed hannorein closed 1 year ago
Ah, that's not good, it sounds like undefined behaviour somewhere. I'll see if I can track down the problem. What OS are you seeing the failing test on?
That's on MacOS.
I got to the bottom of this one, I'd left a call to pow() in the code for the step size update. I have replaced it with ias15 version of x^1/7. I have tested for bitwise energy violation matching for the longterm integration test is the unit test file (1e6 orbits for inner solar system) using a ubuntu desktop and an M1 macbook.
I have opened a pull request for the fix but wasn't sure if I was meant to link it to this issue somehow.
Closed in #685
@PeterBartram: One TES unit test fails on my computer:
However, it passes on the CI. That is a bit worrying because they results of the test should be the same, regardless where they are run. I don't think this is related, but FWIW, I just fixed a bug in the
rebound_integrator_tes
ctypes structure definition (what is actually a double was listed as a pointer to a double).