p-adic height gives incorrect precision

[sagetrac-dmharvey]

If I ask for precision 10, I get precision 9:

sage: E = EllipticCurve("37a")
sage: P = E.gens()[0]
sage: h = E.padic_height(5, 10)
sage: h(P)
4*5 + 3*5^2 + 3*5^3 + 4*5^4 + 4*5^5 + 5^6 + 4*5^8 + O(5^9)

It didn't use to behave this way; it probably happened accidentally as a consequence of some changes chris wuthrich made, and it appears that the doctests were modified to make this behaviour the "correct" one. This should be fixed, because in the large prime case it ends up wasting a lot of time computing extra digits in intermediate steps.

Component: algebraic geometry

Issue created by migration from https://trac.sagemath.org/ticket/635

[sagetrac-dmharvey]

comment:3

Attachment: patch-635.hg.gz

Attached patch fixes the precision problem, but seems to introduce some weird segfault-like issues when running long doctests. I have no idea why this is happening because I'm only touching high-level python code. My guess is that those issues are independent of this one.

[sagetrac-dmharvey]

comment:4

I've tried the same patch against 2.8.5. Pretty much the same thing happens: the ordinary doctests (on ell_rational_field.py) are all fine, but long doctests produce intermittent segfaults... on some invocations everything is fine, and sometimes it segfaults. I have tried debugging with gdb, but either (a) the problem doesn't occur, or (b) I get an empty stack trace.

I'm marking this as [with patch], because I think this patch is okay, and it's just exposing some other low-level bug somewhere else. If someone can track down the crasher that would be great.

[williamstein]

comment:5

Changed to 2.8.7, since it will hopefully be in David Roe's patch already.