Fixes in pyx files

[fchapoton]

This is fixing a few issues about division in various cython files.

All the fixes were found using the command cython -a -3 *.pyx.

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.

[github-actions]

Documentation preview for this PR (built with commit 83f075e; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[fchapoton]

@kedlaya : could you please have a look at the 2 failing doctests in hypergeometric motives ? I wonder if the old results were correct.

[roed314]

I just checked in Magma, which agrees with Sage's old results (note that there's a t <-> 1/t difference in convention, thus the 1/2):

H := HypergeometricData([[3],[4]]);
EulerFactor(H,1/2,7);
7*$.1^2 - 3*$.1 + 1

Indeed, 21 is too big since it needs to be a weight 1 Weil polynomial.

[roed314]

I haven't looked at the code in detail, but I wonder why you even want to change the signature here. It doesn't seem to relate to the C division convention.

[fchapoton]

Thanks. So my changes are not good. I will have to investigate. Probably I will undo the changes here about hypergeometric motives.

[fchapoton]

I wanted to fix the warning displayed by "cython -a -3 hypergeometric_misc.pyx".

    cdef int gl, j, k, l, v, gv
    cdef long long i, q1, w, w1, w2, q2, r, r1
    cdef bint flip, use_longlongs

    q1 = p ** f - 1
                ^
------------------------------------------------------------

sage/modular/hypergeometric_misc.pyx:42:16: Cannot assign type 'double' to 'long long'

[kedlaya]

Off the top of my head I am guessing that changing p to unsigned long long is changing the behavior of % down the line. This is of course fixable, let me have a look.

[fchapoton]

I have tried to make a smaller change in hgm, and this seems to work. Not sure that it appeases "cython -a -3" buit this is maybe not important

[kedlaya]

I have tried to make a smaller change in hgm, and this seems to work. Not sure that it appeases "cython -a -3" buit this is maybe not important

On my end changing only f to be unsigned int seems to fix the warning without disturbing the doctest (which makes sense as the issue is with raising an integer to a possibly negative integer power).

[fchapoton]

Thanks. So this is now ready for review.

[fchapoton]

please review, if you have time

[dcoudert]

otherwise LGTM.

[fchapoton]

merci ! Voilà !

[dcoudert]

LGTM.