Actually there are two different errors we are hunting here --- when
you sent me two backtraces about 12 hours ago there was one for each.

One is the apparent referenceing of uninitialized memory, somwhere
around line 900 of qcurves/sieve_search.cc (I'm being imprecise since
I have been adding dozens of output statements to that file trying to
track down what is happening).  That's the thing I have been trying to
find today.

The other one seems to only surface with gcc 4.3 since it does not
happen to me: it's in procs/template.h where I need a function
vector_union<NTL::ZZ> to be instantiated.  So it's a template function
problem.

John

On 05/12/2007, mabshoff
<[EMAIL PROTECTED]> wrote:
>
>
>
> On Dec 5, 7:46 pm, "John Cremona" <[EMAIL PROTECTED]> wrote:
> > > > mwrank compiles with gcc 4.2, so that has never been a problem. I am
> > > > not sure who fixed it, but the changes should have certainly flown
> >
> > If you mean, who fixed my source to compile ok under 4.2 then that was me.
>
> Ok, I wasn't sure and I don't think that mwrank contains an hg repo.
> So my bad.
>
> >
> > > > back into your tree. With the removal of mwrank.spkg that issue will
> > > > go away.
> >
> > Good!
> >
> >
> >
> > > > I am working on getting mwrank working with the cremona.spkg version,
> > > > i.e. the removal of the original mwrank.spkg. I will also send you a
> > > > patch for the gcc 4.3 fixes shortly. I will do some build testing on
> > > > my end to make sure I haven't broken anything and then merge it into
> > > > 2.9.
> >
> > Thanks.
> >
> > > > > John
> >
> > > > I had a look at trac and currently the following tickets related to
> > > > mwrank are open:
> >
> > > > #1058   the mwrank interface barfs on bad input
> >
> > This is a triviality.  mwrank_elliptic curves need to be initialized
> > with 5 integers, period.  Currently Sage just tried to coerce the 5
> > things to integers, whcih some stupid results (e.g. 1.2 gets rounded
> > to 1 with meaningless results).
> >
> > A more interesting issue is whether to allow mwrank_elliptic curves to
> > be initialised with rational (non-integral) data, since this makes
> > perfect mathematical sense and users regularly try it.  This could be
> > solved in two ways: either by me in the C++ code -- which is not
> > trivial but might be good to do for the stand-alone mwrank program
> > too.  Or by Sage converting to a minimal model (easy) and then
> > changing back the points afterwards to the original model (also easy).
> >  Both of these should be done as part of the plan to implement
> > model-changing maps (i.e. [u,r,s,t] transformations for those who know
> > what those are) which I rashly thought I would do months ago (see
> > ticket #740).
> >
> > > > #1233   mwrank wrapper causes crashes and error on non-minimal curves
> >
> > This is a much easier variant of the previous one, which I could
> > easily fix in the mwrank source code.  (Currently the stand-alone
> > mwrank allowed non-minimal input since it minimises and only passes a
> > minimal model to the 2-descent library functions.  The library
> > functions expect a minimal curve.  But I could easily do that
> > minimization within the library functions.)
> >
> > > > #1248   mwrank wrapper defect
> >
> > This was Paul Z's observation and seems to be a wrapper problem, so
> > "anyone" could try to track that down.
> >
>
> Any volunteers to do this?
>
> > > > #1256   mwrank*.spkg now redundant, included in cremona*.spkg
> >
> > In hand
> >
> > > > #1402   mwrank: "Invalid read" while Searching for Points
> >
> > This is the thing I am having such fun with this afternoon.....
> >
> > > > #1403   mwrank has termination issues
> >
> > i.e. mwrank crashed rather than stopping cleanly when reachinf EOF on
> > input.  Should be an easy fix to qrank/getcurve.cc, so I'll do it.
> >
>
> Thanks for the input, I overlooked that you had already added info to
> some of the tickets (like #1058, #1233). I added your comments to the
> tickets where there were none so far.
>
> We really need to get the email notification for trac going, so I will
> not miss those updates any more. Robert, one the quarter is over and
> you find some time could you try to get this working? I saw that you
> had some comments on the smtp configuration for trac.
>
> >
> >
> > > > I will take care of #1256 now. If you have some time can you please
> > > > give an assessment of the other tickets and tell us which ones you
> > > > consider to be in your realm of your code base and which ones in the
> > > > realm of Sage?
> >
> > > And just to add to that, feel free to say "please just send me what you
> > > think are fixes for everything, and I'll just have a look".  I can 
> > > personally
> > > think of numerous people reading this email right now who could easily
> > > fix things like #1058 -- e.g., I know Justin Walker could easily make 
> > > short
> > > work of it... if he knew for a fact he wasn't stepping on your toes by
> > > creating a fix.
> >
> > Thanks for tolerating my rants -- must be having a bad day, and not
> > helped by valgrind....
>
> I know the feeling, very often I feel the need to bang my head on my
> desk while trying to find the one elusive bug that doesn't want to be
> squashed. Hunting bugs in code that one didn't write makes it much
> more difficult on top of that. I will recompile mwrank with gcc 4.3
> using less optimization to see if the problem goes away. It might be a
> miscompilation of the code.
>
> >
> > John
> >
>
> Cheers,
>
> Michael
>
> > --
> > John Cremona
> >
>


-- 
John Cremona

--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://sage.scipy.org/sage/ and http://modular.math.washington.edu/sage/
-~----------~----~----~----~------~----~------~--~---

Reply via email to