mwrank crashing

[sagetrac-dmharvey]

I'm not sure if this is a bug in SAGE or mwrank or if I'm just getting unlucky....

sage: E = EllipticCurve([3, -15675])

sage: E.rank()
---------------------------------------------------------------------------
<type 'exceptions.RuntimeError'>          Traceback (most recent call last)

/home/dmharvey/<ipython console> in <module>()

/home/was/sage/local/lib/python2.5/site-packages/sage/schemes/elliptic_curves/ell_rational_field.py in rank(self, use_database, verbose, only_use_mwrank, algorithm)
    839             elif algorithm == 'mwrank_shell':
    840                 misc.verbose("using mwrank shell")
--> 841                 X = self.mwrank()
    842                 if not 'The rank and full Mordell-Weil basis have been determined unconditionally' in X:
    843                     raise RuntimeError, '%s\nRank not provably correct (maybe try rank with only_use_mwrank=False).'%X

/home/was/sage/local/lib/python2.5/site-packages/sage/schemes/elliptic_curves/ell_rational_field.py in mwrank(self, options)
    142             from sage.interfaces.all import Mwrank
    143             mwrank = Mwrank(options=options)
--> 144         return mwrank(self.a_invariants())
    145 
    146     def conductor(self, algorithm="pari"):

/home/was/sage/local/lib/python2.5/site-packages/sage/interfaces/mwrank.py in __call__(self, cmd)
     68 
     69     def __call__(self, cmd):
---> 70         return self.eval(cmd)
     71 
     72     def console(self):

/home/was/sage/local/lib/python2.5/site-packages/sage/interfaces/expect.py in eval(self, code, strip)
    463         code = code.strip()
    464         try:
--> 465             return '\n'.join([self._eval_line(L) for L in code.split('\n') if L != ''])
    466         except KeyboardInterrupt:
    467             self._keyboard_interrupt()

/home/was/sage/local/lib/python2.5/site-packages/sage/interfaces/expect.py in _eval_line(self, line, allow_use_file, wait_for_prompt)
    426                         return ''
    427                     raise RuntimeError, "%s\n%s crashed executing %s"%(msg,
--> 428                                                    self, line)
    429                 out = E.before
    430             else:

<type 'exceptions.RuntimeError'>: End Of File (EOF) in read_nonblocking(). Exception style platform.
<pexpect.spawn instance at 0x2aaabad74d40>
version: 2.0 ($Revision: 1.151 $)
command: /home/was/sage/local/bin/mwrank
args: ['/home/was/sage/local/bin/mwrank']
patterns:
    Enter curve: 
buffer (last 100 chars): 
before (last 100 chars): [0, 0, 0, 3, -15675]

after: <class 'pexpect.EOF'>
match: None
match_index: None
exitstatus: None
flag_eof: 1
pid: 20884
child_fd: 3
timeout: None
delimiter: <class 'pexpect.EOF'>
logfile: None
maxread: 10000
searchwindowsize: None
delaybeforesend: 0
Mwrank crashed executing [0, 0, 0, 3, -15675]

Component: algebraic geometry

Issue created by migration from https://trac.sagemath.org/ticket/201

[ncalexan]

comment:1

As of 2.1.5, it appears to be an error in mwrank, or at least sage's build of mwrank:

leeto:~/Devel/sage-alpha8/local/bin nalexand$ ./mwrankProgram mwrank: uses 2-descent (via 2-isogeny if possible) to
determine the rank of an elliptic curve E over Q, and list a
set of points which generate E(Q) modulo 2E(Q).
and finally saturate to obtain generating points on the curve.
For more details see the file mwrank.doc.
For details of algorithms see the author's book.

Please acknowledge use of this program in published work, 
and send problems to John.Cremona@nottingham.ac.uk.

Version compiled on Dec  8 2006 at 12:52:17 by GCC 4.0.1 (Apple Computer, Inc. build 5247)
using base arithmetic option NTL_ALL (NTL bigints and multiprecision floating point)
Using NTL multiprecision floating point with 15 decimal places.
Enter curve: [0,0,0,3, -15675]

Curve [0,0,0,3,-15675] :        Basic pair: I=-9, J=423225
disc=-179119403541
2-adic index bound = 2
2-adic index = 2
Two (I,J) pairs
Looking for quartics with I = -9, J = 423225
Looking for Type 3 quartics:
Trying positive a from 1 up to 14 (square a first...)
(1,1,-72,275,-364)      --nontrivial...(x:y:z) = (1 : 1 : 0)
Point = [386:2489:8]
        height = 5.56883271776712
Rank of B=im(eps) increases to 1
Trying positive a from 1 up to 14 (...then non-square a)
Trying negative a from -1 down to -21
Finished looking for Type 3 quartics.
Looking for quartics with I = -144, J = 27086400
Looking for Type 3 quartics:
Trying positive a from 1 up to 57 (square a first...)
(9,-4,-600,2784,-3644)  --nontrivial...(x:y:z) = (1 : 3 : 0)
Point = [2703:26837:27]
        height = 6.49337125992467
Doubling global 2-adic index to 2
global 2-adic index is equal to local index
so we abort the search for large quartics
Rank of B=im(eps) increases to 2
Exiting search for large quartics after finding enough globally soluble ones.
Mordell rank contribution from B=im(eps) = 2
Selmer  rank contribution from B=im(eps) = 2
Sha     rank contribution from B=im(eps) = 0
Mordell rank contribution from A=ker(eps) = 0
Selmer  rank contribution from A=ker(eps) = 0
Sha     rank contribution from A=ker(eps) = 0
Rank = 2

Regulator (before saturation) = 16.9955132982626
Searching for points (bound = 10)...done
Regulator (after searching) = 16.9955132982626
Saturating (bound = 100)...RR: division by zero
Abort trap

[williamstein]

comment:2

On 8/17/07, John Cremona <john.cremona@gmail.com> wrote:
> It's a simple precision problem -- even running mwrank with precision
> 20 ("-p20") sorts it.  This happens in the code to find the real roots
> of a real cubic, and it's the classical problem where subtracting two
> close reals gives a bad answer (in this case it gave 0 in a situation
> where theory said that the number cannot be zero).
>
> Testing this using sage I found that the command
> mwrank_set_precision() does not work:
>
> sage: mwrank_set_precision(30)
> ---------------------------------------------------------------------------
> <type 'exceptions.ImportError'>           Traceback (most recent call last)
>
> /home/jec/gp/<ipython console> in <module>()
>
> /home/jec/sage-2.7/local/lib/python2.5/site-packages/sage/libs/mwrank/interface.py
> in set_precision(n)
>      27         n -- long
>      28     """
> ---> 29     from sage.libs.mwrank.mwrank import _set_precision #
> import here to save time
>      30     _set_precision(n)
>      31
>
> <type 'exceptions.ImportError'>: cannot import name _set_precision
>
> -- can you fix that?  [Also, while preparing this email:  version() returns
> sage: version()

Yes, I'll add this to the bug list.

 http://www.sagemath.org:9002/sage_trac/ticket/434

We're organizing big "bug squashing days" now (this first is tomorrow).

>  'SAGE Version 2.7.3, Release Date: 2007-08-02'
>
> although this is 2.8 (according to the banner at the top), though it's
> in a directory called sage-2.7 since that was where I last did a build
> from scratch;  since then I have been upgrade()-ing.

Do "sage -br" (and possibly hg_sage.pull(); hg_sage.merge() from in SAGE).
That the above says 2.7.3 means that your the SAGE Python library is still
at version 2.7.3.  Some sort of interactive confirmation might be needed to
merge the new version of SAGE with your changes.

> Finally, the attached file realroots.cc includes a patch which
> improves the stability of the offending function (so that mwrank -p15
> now works OK on that curve).  And options.h for the email address.
>
> Thanks for the bug report!

Excellent, thanks!

[williamstein]

Attachment: options.h.gz

[williamstein]

comment:3

Attachment: realroots.cc.gz

[sagetrac-mabshoff]

comment:5

Hello,

valgrind did reveal some issues:

==5927== Mismatched free() / delete / delete []
==5927==    at 0x4A05130: operator delete(void*) (vg_replace_malloc.c:244)
==5927==    by 0x4CCA293: vec_l::init(long) (vec.cc:62)
==5927==    by 0x4CD40B2: echmodp(mat_l const&, vec_l&, vec_l&, long&, long&, long) (mat.cc:1366)
==5927==    by 0x4D74CA4: saturator::nextq() (saturate.cc:155)
==5927==    by 0x4D753AA: saturator::test_saturation(int, int) (saturate.cc:90)
==5927==    by 0x4D75483: saturator::do_saturation(int, int) (saturate.cc:236)
==5927==    by 0x4D7614C: saturator::do_saturation_upto(int, int) (saturate.cc:268)
==5927==    by 0x4D3190E: mw::process(Point const&, int) (mwprocs.cc:382)
==5927==    by 0x4D37640: mw::process(NTL::ZZ const&, NTL::ZZ const&, NTL::ZZ const&, int) (mwprocs.cc:275)
==5927==    by 0x4D37888: mw::process(NTL::ZZ const&, NTL::ZZ const&, NTL::ZZ const&) (mwprocs.cc:258)
==5927==    by 0x4D43254: qsieve::check_point(unsigned long, long, long, long*, int) (sieve_search.cc:994)
==5927==    by 0x4D43890: qsieve::sift0(long, long, long, int) (sieve_search.cc:905)
==5927==  Address 0x713F1B8 is 0 bytes after a block of size 0 alloc'd
==5927==    at 0x4A05CB9: operator new[](unsigned long) (vg_replace_malloc.c:199)
==5927==    by 0x4CCA5D7: vec_l::vec_l(long) (vec.cc:36)
==5927==    by 0x4D74C1C: saturator::nextq() (saturate.cc:153)
==5927==    by 0x4D753AA: saturator::test_saturation(int, int) (saturate.cc:90)
==5927==    by 0x4D75483: saturator::do_saturation(int, int) (saturate.cc:236)
==5927==    by 0x4D7614C: saturator::do_saturation_upto(int, int) (saturate.cc:268)
==5927==    by 0x4D3190E: mw::process(Point const&, int) (mwprocs.cc:382)
==5927==    by 0x4D37640: mw::process(NTL::ZZ const&, NTL::ZZ const&, NTL::ZZ const&, int) (mwprocs.cc:275)
==5927==    by 0x4D37888: mw::process(NTL::ZZ const&, NTL::ZZ const&, NTL::ZZ const&) (mwprocs.cc:258)
==5927==    by 0x4D43254: qsieve::check_point(unsigned long, long, long, long*, int) (sieve_search.cc:994)
==5927==    by 0x4D43890: qsieve::sift0(long, long, long, int) (sieve_search.cc:905)
==5927==    by 0x4D43E68: qsieve::sift(long) (sieve_search.cc:829)

[sagetrac-mabshoff]

comment:6

Okay, with the two files from above and a patch a things are working better.

For the patch see http://fsmath.mathematik.uni-dortmund.de/~mabshoff/patches/mwrank-fix-delete-operators-in-vec.cc.patch

Valgrind still complains:

Rank = 2
Searching for points (bound = 8)...==7030== Conditional jump or move depends on uninitialised value(s)
==7030==    at 0x4D48750: qsieve::sift0(long, long, long, int) (sieve_search.cc:903)
==7030==    by 0x4D48D54: qsieve::sift(long) (sieve_search.cc:829)
==7030==    by 0x4D4B9B3: qsieve::search() (sieve_search.cc:1119)
==7030==    by 0x4D3BD93: mw::search(NTL::RR, int, int) (mwprocs.cc:724)
==7030==    by 0x4E036F9: two_descent::saturate(long) (descent.cc:93)
==7030==    by 0x401D7E: main (mwrank.cc:91)
==7030==
==7030== Conditional jump or move depends on uninitialised value(s)
==7030==    at 0x4D47D95: qsieve::check_point(unsigned long, long, long, long*, int) (sieve_search.cc:942)
==7030==    by 0x4D4877C: qsieve::sift0(long, long, long, int) (sieve_search.cc:905)
==7030==    by 0x4D48D54: qsieve::sift(long) (sieve_search.cc:829)
==7030==    by 0x4D4B9B3: qsieve::search() (sieve_search.cc:1119)
==7030==    by 0x4D3BD93: mw::search(NTL::RR, int, int) (mwprocs.cc:724)
==7030==    by 0x4E036F9: two_descent::saturate(long) (descent.cc:93)
==7030==    by 0x401D7E: main (mwrank.cc:91)
==7030==
==7030== Conditional jump or move depends on uninitialised value(s)
==7030==    at 0x4D47D9F: qsieve::check_point(unsigned long, long, long, long*, int) (sieve_search.cc:951)
==7030==    by 0x4D4877C: qsieve::sift0(long, long, long, int) (sieve_search.cc:905)
==7030==    by 0x4D48D54: qsieve::sift(long) (sieve_search.cc:829)
==7030==    by 0x4D4B9B3: qsieve::search() (sieve_search.cc:1119)
==7030==    by 0x4D3BD93: mw::search(NTL::RR, int, int) (mwprocs.cc:724)
==7030==    by 0x4E036F9: two_descent::saturate(long) (descent.cc:93)
==7030==    by 0x401D7E: main (mwrank.cc:91)
done:
  found points of rank 2

and

==7311== Conditional jump or move depends on uninitialised value(s)
==7311==    at 0x4CF4F5F: Curvedata::minimalize() (curvedata.cc:133)
==7311==    by 0x4CF92F4: Curvedata::Curvedata(Curvedata const&, int) (curvedata.cc:81)
==7311==    by 0x4CFE99F: CurveRed::CurveRed(Curvedata const&) (curvered.cc:121)
==7311==    by 0x4D4C7C8: annihilators(Curvedata const&, long) (htconst.cc:1676)
==7311==    by 0x4D60180: CurveHeightConst::CurveHeightConst(Curvedata const&) (htconst.cc:1445)
==7311==    by 0x4D77343: index_bound(Curvedata*, std::vector<Point, std::allocator<Point> >&, int, int) (saturate.cc:509)
==7311==    by 0x4D79ABF: saturator::saturate(std::vector<long, std::allocator<long> >&, NTL::ZZ&, long, int, int, int) (saturate.cc:359)
==7311==    by 0x4D320DA: mw::saturate(NTL::ZZ&, std::vector<long, std::allocator<long> >&, long, int) (mwprocs.cc:645)
==7311==    by 0x4E03AE4: two_descent::saturate(long) (descent.cc:137)
==7311==    by 0x401D7E: main (mwrank.cc:91)
==7311==
==7311== Conditional jump or move depends on uninitialised value(s)
==7311==    at 0x4CF4F5F: Curvedata::minimalize() (curvedata.cc:133)
==7311==    by 0x4CF92F4: Curvedata::Curvedata(Curvedata const&, int) (curvedata.cc:81)
==7311==    by 0x4CFE99F: CurveRed::CurveRed(Curvedata const&) (curvered.cc:121)
==7311==    by 0x4D76CD2: tamagawa_primes(Curvedata const&) (saturate.cc:532)
==7311==    by 0x4D79C88: saturator::saturate(std::vector<long, std::allocator<long> >&, NTL::ZZ&, long, int, int, int) (saturate.cc:390)
==7311==    by 0x4D320DA: mw::saturate(NTL::ZZ&, std::vector<long, std::allocator<long> >&, long, int) (mwprocs.cc:645)
==7311==    by 0x4E03AE4: two_descent::saturate(long) (descent.cc:137)
==7311==    by 0x401D7E: main (mwrank.cc:91)

Cheers,

Michael

[williamstein]

comment:7

fixed! for sage-2.8.2.