eclib: suppress / delete files left by doctests and check, fix building in parallel

[qed777]

From John Palmieri:

When I run doctests on the file ell_rational_field.py, I end up with a
small file called PRIMES in the current directory. This shouldn't
happen: running doctests shouldn't produce files in a non-temporary
directory.
[...]

This is a follow-up to #7575. Please see comment 24+ for some progress.

A new spkg is available at

• http://sage.math.washington.edu/home/mpatel/trac/8357/eclib-20080310.p10.spkg

It includes the PRIMES, 1, and make check fixes mentioned below and parallel build fixes (cf. #8306).

Note to release manager: Merge only the spkg.

Component: elliptic curves

Author: John Cremona, Mitesh Patel

Reviewer: Minh Van Nguyen

Merged: sage-4.3.4.alpha1

Issue created by migration from https://trac.sagemath.org/ticket/8357

[JohnCremona]

comment:1

The quickest quick fix would be just to comment out line 372 in src/procs/marith.cc (in eclib).

Anything else requires thought and testing, which I don't have time for i nthe next several days.

[qed777]

Description changed:

--- 
+++ 
@@ -4,41 +4,19 @@
 When I run doctests on the file ell_rational_field.py, I end up with a
 small file called PRIMES in the current directory. This shouldn't
 happen: running doctests shouldn't produce files in a non-temporary
-directory.  However, I can't figure out how this file gets there.  In
-particular, if I delete the following doctests from the rank method,
-then the file is not produced:
-
-{{{
-diff -r 23241bd151e3 sage/schemes/elliptic_curves/
-ell_rational_field.py
---- a/sage/schemes/elliptic_curves/ell_rational_field.py        Thu
-Feb 18 14:25:25 2010 -0800
-+++ b/sage/schemes/elliptic_curves/ell_rational_field.py        Wed
-Feb 24 20:37:09 2010 -0800
-@@ -1679,14 +1679,6 @@
-             sage: E.minimal_model().rank()
-             1
-
--        A large example where mwrank doesn't determine the result
-with certainty::
--
--            sage: EllipticCurve([1,0,0,0,37455]).rank(proof=False)
--            0
--            sage: EllipticCurve([1,0,0,0,37455]).rank(proof=True)
--            Traceback (most recent call last):
--            ...
--            RuntimeError: Rank not provably correct.
-         """
-         if proof is None:
-             from sage.structure.proof.proof import get_flag
-
-}}}
-
-If I run these by hand from the command line, though, I don't see the
-file PRIMES anywhere.  Can someone familiar with the elliptic curves
-code open a ticket, chase this down, and fix it?
+directory.
+[...]
 ```
 
 This is a follow-up to #7575.  Please see [comment 24+](https://github.com/sagemath/sage/issues/7575#comment:24) for some progress.
 
+A new spkg is available at
 
+* http://sage.math.washington.edu/home/mpatel/trac/8357/eclib-20080310.p10.spkg
+
+It includes
+
+* The quick fix mentioned below [attachment: trac_8357-suppress_PRIMES.patch](https://github.com/sagemath/sage-prod/files/10648255/trac_8357-suppress_PRIMES.patch.gz)
+* Parallel build fixes from #8306: [attachment: trac_8357-eclib_makefiles.patch](https://github.com/sagemath/sage-prod/files/10648253/trac_8357-eclib_makefiles.patch.gz), [attachment: trac_8357-spkg.patch](https://github.com/sagemath/sage-prod/files/10648254/trac_8357-spkg.patch.gz)
+
+**Note to release manager:** Merge only the spkg.

[qed777]

Author: John Cremona, Mitesh Patel

[qed777]

comment:3

There's still a timing(?) problem (log).

[qed777]

Attachment: trac_8357-eclib_makefiles.patch.gz

Tweak MAKEFILEs for parallel builds. eclib src repo.

[qed777]

comment:4

Replying to @qed777:

There's still a timing(?) problem (log).

I think the problem is that make sometimes attempts to build a target in progs before it's done building a requisite shared library, e.g., install_so. I've updated the "makefiles" patch and spkg with a fix that I'm testing now.

[JohnCremona]

comment:5

Replying to @qed777:

Replying to @qed777:

There's still a timing(?) problem (log).

I think the problem is that make sometimes attempts to build a target in progs before it's done building a requisite shared library, e.g., install_so. I've updated the "makefiles" patch and spkg with a fix that I'm testing now.

That sounds likely. We (I) should probably tidy this up; there are several useful executables which are built, but the only one actually used by and accessible from Sage after the build is mwrank. So i should change the targets so that the other progs are only built when doing a make check.