Open alexjbest opened 4 years ago
Yeah, that does not surprise me. During the Tamagawa tests, I found out that it is really not safe to test for equality in Qp, as the last digits of a root do not have to be correct. So you apply Frobenius to a root, and you get something that is not in the list of roots. To solve this, I changed all equalities into the checks for an error of half the precision. Now it is required the roots to lie at least so far apart from each other. Maybe this is a bit much, and can probably be improved by using ball arithmetic instead of just elements.
When we did the ICERM demo, this example used to work
now it needs more prec