using even fancier pytest fixtures to help localize errors #101
Conversation
…actly the same numerical results as local
|
My last one had the real crux lines commented out accidentally, and when I uncommented them I broke the build on master, because it turns out the cloud runs are numerically just a little different than my local runs, so a couple ended up exceeding the limits. Merging this to fix master. This one also makes the pytest fixtures more sophisticated so we can better localize these kinds of errors in the future. That necessitated getting more sophisticated about how the code does plotting too, but the end result is somewhat cleaner than before. 🎉 |
| [(-25, -25), (2, 2), (0, 0), (2, 2)], | ||
| [(-25, -25), (3, 3), (0, 0), (3, 3)]], | ||
| iterated_first_order: [[(-7, -7), (-10, -11), (0, -1), (0, 0)], | ||
| iterated_first_order: [[(-7, -7), (-9, -10), (0, -1), (0, 0)], |
There was a problem hiding this comment.
These are what I had to loosen. Unfortunately I couldn't figure out which among the sea of numbers I had to increment, so now I mark.parametrize over the test functions too.
1 participant
I broke the build, which won't do. Trying to fix it.