#6396 closed defect (fixed)
[with patch, positive review] primes_of_degree_one is broken for relative extensions
Reported by: | David Loeffler | Owned by: | William Stein |
---|---|---|---|
Priority: | major | Milestone: | sage-4.1.1 |
Component: | number theory | Keywords: | |
Cc: | ncalexan | Merged in: | sage-4.1.1.alpha0 |
Authors: | David Loeffler | Reviewers: | Nick Alexander, Minh Van Nguyen |
Report Upstream: | N/A | Work issues: | |
Branch: | Commit: | ||
Dependencies: | Stopgaps: |
Description
This is kind of irritating:
sage: N.<a,b> = NumberField([x^2 + 1, x^2 - 5]) sage: N.primes_of_degree_one_list(10) [Fractional ideal (1), Fractional ideal (1), Fractional ideal (1), Fractional ideal (1), Fractional ideal (1), Fractional ideal (1), Fractional ideal (1), Fractional ideal (1), Fractional ideal (1), Fractional ideal (1)]
Attachments (2)
Change History (15)
comment:1 Changed 13 years ago by
Cc: | ncalexan added |
---|
comment:2 Changed 13 years ago by
Summary: | primes_of_degree_one is broken for relative extensions → [with path, positive review] primes_of_degree_one is broken for relative extensions |
---|
This looks good to me, and I am cautiously optimistic that the method will select the same generator on all architectures. Apply.
comment:3 Changed 13 years ago by
Authors: | → David Loeffler |
---|---|
Milestone: | → sage-4.1 |
Reviewers: | → Nick Alexander |
comment:4 Changed 13 years ago by
Summary: | [with path, positive review] primes_of_degree_one is broken for relative extensions → [with patch, positive review] primes_of_degree_one is broken for relative extensions |
---|
comment:5 Changed 13 years ago by
Summary: | [with patch, positive review] primes_of_degree_one is broken for relative extensions → [with patch, needs work] primes_of_degree_one is broken for relative extensions |
---|
Doctest failures when applied to 4.1.alpha1
sage -t -long devel/sage/sage/rings/number_field/small_primes_of_degree_one.py ********************************************************************** File "/space/boothby/sage-4.1.alpha1/devel/sage-main/sage/rings/number_field/small_primes_of_degree_one.py", line 204: sage: N.primes_of_degree_one_list(10) Expected: [Fractional ideal ((-1/2*b + 1/2)*a + 2), Fractional ideal (-b*a + 1/2*b + 1/2), Fractional ideal ((1/2*b + 3/2)*a - b), Fractional ideal ((-1/2*b - 3/2)*a + b - 1), Fractional ideal (-b*a - b + 1), Fractional ideal (3*a + 1/2*b - 1/2), Fractional ideal ((-3/2*b + 1/2)*a + 1/2*b - 1/2), Fractional ideal ((-1/2*b - 5/2)*a - b + 1), Fractional ideal (2*a - 3/2*b - 1/2), Fractional ideal (3*a + 1/2*b + 5/2)] Got: [Fractional ideal (2*a + 1/2*b - 1/2), Fractional ideal ((-1/2*b - 1/2)*a - b), Fractional ideal (b*a + 1/2*b + 3/2), Fractional ideal ((-1/2*b - 3/2)*a + b - 1), Fractional ideal ((-b + 1)*a + b), Fractional ideal (3*a + 1/2*b - 1/2), Fractional ideal ((1/2*b - 1/2)*a + 3/2*b - 1/2), Fractional ideal ((-1/2*b - 5/2)*a - b + 1), Fractional ideal (2*a - 3/2*b - 1/2), Fractional ideal (3*a + 1/2*b + 5/2)] ********************************************************************** 1 items had failures: 1 of 8 in __main__.example_5 ***Test Failed*** 1 failures. For whitespace errors, see the file /space/boothby/sage-4.1.alpha1/tmp/.doctest_small_primes_of_degree_one.py [3.2 s]
comment:6 Changed 13 years ago by
Aargh! Every damn thing we do with ideals seems to return different answers on different platforms -- I've never understood why this is and it's fantastically annoying. The ideals are the same ones, of course, but their string representations are different because Pari picks generators in a totally unpredictable way.
I'll fix this when I get around to it (probably after SD16 is over)
David
Changed 13 years ago by
Attachment: | trac_6396-deg1primes.patch added |
---|
comment:7 Changed 13 years ago by
Can you try this new patch and see if it works better for you? It works on my machine, but then so did the last one, and there is no sage.math binary of 4.0.2 available AFAIK.
comment:8 Changed 13 years ago by
Summary: | [with patch, needs work] primes_of_degree_one is broken for relative extensions → [with new patch, needs review] primes_of_degree_one is broken for relative extensions |
---|
comment:9 Changed 13 years ago by
BTW: having built myself a 4.1.alpha1 on sage.math overnight, I've checked that it passes doctests there too. Can I just reinstate the positive review now, or does it need to be re-reviewed?
comment:10 Changed 13 years ago by
Summary: | [with new patch, needs review] primes_of_degree_one is broken for relative extensions → [with new patch, positive review] primes_of_degree_one is broken for relative extensions |
---|
Hello? I'm marking this as "positive review" so it comes to the attention of the release managers for 4.1.1, who can decide as they see fit what to do given the slightly ambiguous status. All that's needed is for someone to check that the doctests pass on both 64-bit and 32-bit.
(I really would prefer it if this patch didn't hang around bitrotting forever -- this is necessary for a major project I'm working on, which seems to have exposed a remarkable number of bugs; see also #6457, #6458, #6462 and #6463.)
Changed 13 years ago by
Attachment: | trac_6396-reviewer.patch added |
---|
reviewer patch; fix minor docstring formatting
comment:11 Changed 13 years ago by
Reviewers: | Nick Alexander → Nick Alexander, Minh Van Nguyen |
---|---|
Summary: | [with new patch, positive review] primes_of_degree_one is broken for relative extensions → [with patch, positive review] primes_of_degree_one is broken for relative extensions |
The patch trac_6396-reviewer.patch
makes some adjustment to the docstring introduced by trac_6396-deg1primes.patch
, and fixes some typos therein. All tests passed on sage.math (a 64-bit machine) and on my 32-bit Debian Lenny machine. Just to let people know, both patches have been merged in sage-4.1.1-alpha0. I can't close this ticket because I don't have the privilege to do so. Sorry, folks :-(
comment:12 Changed 13 years ago by
Merged in: | → sage-4.1.1.alpha0 |
---|---|
Resolution: | → fixed |
Status: | new → closed |
Here's a patch. Turns out that the bug was due to using the wrong generator. I've set it to use
absolute_generator
rather thangen
, and patched absolute number fields soabsolute_generator
is an alias forgen
in that case. The patch also ReSTifies small_primes_of_degree_one.py and adds it to the reference manual.(Nick, since he wrote the original code.)
David