Upgrade to flint-3.3.1

[vbraun]

Straightforward upgrade to the latest release

Apparently Singluar is not yet compatible with it

[singular-4.4.1] [spkg-install] /usr/bin/bash ./libtool  --tag=CXX   --mode=compile g++ -std=gnu++11 -DHAVE_CONFIG_H -I.  -I./include -I./include -I. -DFACTORY_BUILDING_DLL -I/home/vbraun/Sage/git/local/var/tmp/sage/build/singular-4.4.1/src -I/home/vbraun/Sage/git/local/var/tmp/sage/build/singular-4.4.1/src  -I/home/vbraun/Sage/git/local/include/  -I/home/vbraun/Sage/git/local/include -I/home/vbraun/Sage/git/local/include -DSING_NDEBUG -DOM_NDEBUG   -g -O2 -pipe -fno-common -g0 -O3 -Wno-unused-function -Wno-trigraphs -Wno-unused-parameter -Wunknown-pragmas -Wno-unused-variable -fomit-frame-pointer -fwrapv -fvisibility=default -finline-functions -fno-exceptions -fno-threadsafe-statics -fno-enforce-eh-specs -fconserve-space -funroll-loops -fno-rtti -c -o FLINTconvert.lo FLINTconvert.cc
[singular-4.4.1] [spkg-install] libtool: compile:  g++ -std=gnu++11 -DHAVE_CONFIG_H -I. -I./include -I./include -I. -DFACTORY_BUILDING_DLL -I/home/vbraun/Sage/git/local/var/tmp/sage/build/singular-4.4.1/src -I/home/vbraun/Sage/git/local/var/tmp/sage/build/singular-4.4.1/src -I/home/vbraun/Sage/git/local/include/ -I/home/vbraun/Sage/git/local/include -I/home/vbraun/Sage/git/local/include -DSING_NDEBUG -DOM_NDEBUG -g -O2 -pipe -fno-common -g0 -O3 -Wno-unused-function -Wno-trigraphs -Wno-unused-parameter -Wunknown-pragmas -Wno-unused-variable -fomit-frame-pointer -fwrapv -fvisibility=default -finline-functions -fno-exceptions -fno-threadsafe-statics -fno-enforce-eh-specs -fconserve-space -funroll-loops -fno-rtti -c FLINTconvert.cc  -fPIC -DPIC -o .libs/FLINTconvert.o
[singular-4.4.1] [spkg-install] FLINTconvert.cc: In function 'void convertFacCFMatrix2Fq_nmod_mat_t(fq_nmod_mat_struct*, const fq_nmod_ctx_struct*, const CFMatrix&)':
[singular-4.4.1] [spkg-install] FLINTconvert.cc:655:36: error: 'struct fq_nmod_mat_struct' has no member named 'rows'
[singular-4.4.1] [spkg-install]   655 |       convertFacCF2nmod_poly_t (M->rows[i-1]+j-1, m (i,j));
[singular-4.4.1] [spkg-install]       |                                    ^~~~
[singular-4.4.1] [spkg-install] make[11]: *** [Makefile:1274: FLINTconvert.lo] Error 1
[singular-4.4.1] [spkg-install] make[10]: *** [Makefile:1379: all-recursive] Error 1
[singular-4.4.1] [spkg-install] make[9]: *** [Makefile:1041: all] Error 2
[singular-4.4.1] [spkg-install] make[8]: *** [Makefile:624: all-recursive] Error 1
[singular-4.4.1] [spkg-install] make[7]: *** [Makefile:475: all] Error 2
[singular-4.4.1] [spkg-install] *********************************************************************************************************************************************************
[singular-4.4.1] [spkg-install] Error building singular-4.4.1
[singular-4.4.1] [spkg-install] *********************************************************************************************************************************************************

See Singular/Singular#1279

Fixes #40253

[antonio-rojas]

eclib needs patching too (builds fine but crashes at runtime) JohnCremona/eclib#87

[github-actions]

Documentation preview for this PR (built with commit 048c572; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[vbraun]

This PR now works for me, but doesn't fix #40253 when using gcc 15

[JohnCremona]

eclib needs patching too (builds fine but crashes at runtime) JohnCremona/eclib#87

I will shortly be merging @antonio-rojas 's eclib patch there, so either that patch needs to applied in Sage, or I canmake a new eclib release which Sage will need to use and make as a prerequisite for this PR. Sorry about this, there was a backwards-incompatible change in FLINT.

[vbraun]

The eclib patch is already in this PR, fyi

[enriqueartal]

This PR now works for me, but doesn't fix #40253 when using gcc 15

Any idea why? I got the same result, built but segfault.

[vbraun]

We've narrowed the segfault down to flint miscompiling with gcc 15, see flintlib/flint#2340 for details

[enriqueartal]

It works for me!

[vbraun]

I've added the -fno-strict-aliasing to the cflags, but I still get

**********************************************************************
File "src/sage/schemes/elliptic_curves/gp_simon.py", line 84, in sage.schemes.elliptic_curves.gp_simon.?
Failed example:
    E.simon_two_descent()                             # long time
Expected:
    (0, 2, [])
Got:
    (2, 2, [(-121/64*w - 2719/64 : -15555/512*w - 186181/512 : 1)])
**********************************************************************

Though not sure if its related to flint, might be another gcc 15 issue. Afaik thats just using pari so no flint involved...

[antonio-rojas]

I've added the -fno-strict-aliasing to the cflags, but I still get

**********************************************************************
File "src/sage/schemes/elliptic_curves/gp_simon.py", line 84, in sage.schemes.elliptic_curves.gp_simon.?
Failed example:
    E.simon_two_descent()                             # long time
Expected:
    (0, 2, [])
Got:
    (2, 2, [(-121/64*w - 2719/64 : -15555/512*w - 186181/512 : 1)])
**********************************************************************

Though not sure if its related to flint, might be another gcc 15 issue...

That's a random issue, I've beed getting it occasionally for a long time

[vbraun]

Ah true, using different random seed "fixes" the two descent ;)

So this PR is ready for review.

[antonio-rojas]

there's now en eclib release including the patch, and also a 3.3.1 flint bugfix release.

[vbraun]

I've updated the patch for flint-3.3.1, though it still segfaults without -fno-strict-aliasing

The new eclib release doesn't build with gcc-14, I've opened JohnCremona/eclib#88

[vbraun]

I've upgraded the PR to eclib 20250627, now eclib does not need the flint compatibility patch any more

It would be nice to get this reviewed so we can make a release that actually works on Fedora....

[enriqueartal]

I tried this in two computers. In the second one, fresh compilation, everything is fine, only one test did not pass (as in #38544 (comment)); the compilation failed in maxima using system ecl, it worked with the spkg.
In the first computer some tests took very long time and caused crashes after stopping, e.g. src/sage/combinat/designs/incidence_structures.py.

[vbraun]

I've bumped the min version in build/pkgs/eclib/spkg-configure.m4 so a working eclib is required

Still needs review...

[JohnCremona]

I see no problems, but I'm not sure if I can give a positive review without doing a build from scratch on a machine without lots of versions of FLINT and eclib already installed -- which I do not have. If you don't think that matters, please say so.

[enriqueartal]

I built with the last change. Some tests failed with ./sage -t -a but they passed when run independently (Fedora 42)

[vbraun]

I ran it on the buildbot and tests pass there...

[vbraun]

Just a reminder that this PR is necessary to pass tests on Fedora 42... please review...

[enriqueartal]

From my experience it builts in my installations of Fedora 42. Concerning the tests, there were some failures with ./sage - t -a, but all failed tests passed when they were run on their own. It seems fine for me but I do not know if something else is needed for a positive review.

[vbraun]

this PR should really be uncontroversial, its just upstream updates and collecting a patch that upstream provided to work