sagemath: update to 9.8

[tornaria]

This PR includes

• update brial to 1.2.12 (it has to be built in sync because of boost)
• update singular to 4.3.1.p3 (it has to be built in sync because an output difference)

Changes in sagemath package:

• fix edge case of python 3.11 integer conversion on 32 bit
• support tachyon 0.99.5
• support singular 4.3.1p3
• fix tests for giac 1.9.0.35
• fix tests for numpy 1.24
• skip unstable tests in klyachko.py
• remove zn_poly dependency

Testing the changes

• I tested the changes in this PR: YES

[tornaria]

I'm doing system testing of this PR together with #40222, #41210, #41572, #41972, #41973, #42016, #42043, #42044.

[tornaria]

So x86_64 (both glibc and musl) failed with this:

sage -t --warn-long 30.0 --random-seed=0 /builddir/sagemath-9.7/pkgs/sagemath-standard/build/lib.linux-x86_64-cpython-311/sage/interfaces/expect.py
**********************************************************************
File "/builddir/sagemath-9.7/pkgs/sagemath-standard/build/lib.linux-x86_64-cpython-311/sage/interfaces/expect.py", line 937, in sage.interfaces.expect.Expect._eval_line
Failed example:
    singular.interrupt()
Expected:
    True
Got:
    False
**********************************************************************
File "/builddir/sagemath-9.7/pkgs/sagemath-standard/build/lib.linux-x86_64-cpython-311/sage/interfaces/expect.py", line 943, in sage.interfaces.expect.Expect._eval_line
Failed example:
    singular('2+3')
Expected:
    Singular crashed -- automatically restarting.
    5
Got:
    5
**********************************************************************
1 item had failures:
   2 of  16 in sage.interfaces.expect.Expect._eval_line
    [101 tests, 2 failures, 23.40 s]

It has already happened before (#39876 (comment), #41085 (comment)).

This was already known upstream (https://trac.sagemath.org/ticket/33907) and it turns out I already fixed this (it's even merged upstream for 9.8!) but I forgot. This only happens under very heavy load (as here in CI, but not in my box even if I run 16 threads on 8 cores, it has to be way more loaded to trigger)

I'll push a new PR with the patch from https://git.sagemath.org/sage.git/commit?id=6f5c1c2fc8bcfb5e6555716d05ce70511795ffa1.

Meanwhile my local tests with all the above PR together are successful.

[tornaria]

Tested locally together with #41210, #41572, #41972, #41973, #42016, #42043, #42044, #42050 on x86_64 (glibc/musl) and i686 (glibc).

[tornaria]

Rebased without changes to restart CI.

[tornaria]

I had to make a change since the tarball was moved; a different dirname changes the sha256 of the tarball but I checked that the contents are identical -- other than the name of the top level dir which xbps-src renames anyway (that feature gets another 👍 ).

[tornaria]

@dkwo @leahneukirchen this is good to merge IMO, as are #41210, #41572, #41972, #41973, #42016, #42043, #42044, #42050. I tested locally all together on the three non-cross arches.

[dkwo]

On aarch64 (after rebuilding libtool and ntl), the following tests fail:

graphs/generators/{distance_regular.ptx, classical_geometries.py} # segfault
graphs/strongly_regular_db.pyx # segfault
combinat/matrices/hadamard_matrix.py # segfault
quadratic_forms/binary_qf.py # 2 failed
schemes/elliptic_curves/hom_velusqrt.py # 1 failed
graphs/generic_graph.py # timed out

[dkwo]

I did not test with the linked PR's though.

[tornaria]

On aarch64 (after rebuilding libtool and ntl), the following tests fail:

graphs/generators/{distance_regular.ptx, classical_geometries.py} # segfault
graphs/strongly_regular_db.pyx # segfault
combinat/matrices/hadamard_matrix.py # segfault
quadratic_forms/binary_qf.py # 2 failed
schemes/elliptic_curves/hom_velusqrt.py # 1 failed
graphs/generic_graph.py # timed out

Were these doctests passing before this PR?

[dkwo]

If I build from Void's master branch, it fails to start the tests, with ImportError: ../sage/rings/polynomial/polynomial_integer_dense_flint.cpython-311-aarch64-linux-gno.so: undefined symbol: fmpz_set_ZZ. Strangely, in this last case I had to first build maxima without sbcl, as we discussed in the past, while in building the current PR I did not have to do it.

[dkwo]

btw, flintlib currently fails checks on musl:

=> flintlib-2.9.0_1: running do_check ...
    CC   test_helpers.o
make[1]: Entering directory '/builddir/flintlib-2.9.0'
make[1]: Nothing to be done for 'shared'.
make[1]: Leaving directory '/builddir/flintlib-2.9.0'
make[1]: Entering directory '/builddir/flintlib-2.9.0'
make[1]: Nothing to be done for 'static'.
make[1]: Leaving directory '/builddir/flintlib-2.9.0'
make[1]: Entering directory '/builddir/flintlib-2.9.0'
mkdir -p build/test
    CC   build/test/t-add_ssaaaa
    CC   build/test/t-add_sssaaaaaa
    CC   build/test/t-add_ssssaaaaaaaa
    CC   build/test/t-byte_swap
    CC   build/test/t-count_leading_zeros
    CC   build/test/t-count_trailing_zeros
    CC   build/test/t-invert_limb
    CC   build/test/t-sdiv_qrnnd
    CC   build/test/t-smul_ppmm
    CC   build/test/t-sub_dddmmmsss
    CC   build/test/t-sub_ddmmss
    CC   build/test/t-udiv_qrnnd
    CC   build/test/t-udiv_qrnnd_preinv
    CC   build/test/t-umul_ppmm
make[1]: Leaving directory '/builddir/flintlib-2.9.0'
make[1]: Entering directory '/builddir/flintlib-2.9.0'
add_sssaaaaaa....add_ssaaaa....byte_swap....add_ssssaaaaaaaa....PASS
count_leading_zeros....PASS
count_trailing_zeros....PASS
invert_limb....PASS
sdiv_qrnnd....make[1]: *** [Makefile.subdirs:107: build/test/t-sdiv_qrnnd_RUN] Arithmetic exception
make[1]: *** Waiting for unfinished jobs....
PASS
PASS
PASS
make[1]: Leaving directory '/builddir/flintlib-2.9.0'
make: *** [Makefile:222: check] Error 2

[tornaria]

sagemath 9.8 is released. I'll work on that.

Meanwhile, it would be helpful if all the other package updates can be merged.

Re: aarch64, can we move that to a separate issue? I'm not ready to support sagemath on aarch64 yet.

[dkwo]

Re: aarch64, can we move that to a separate issue? I'm not ready to support sagemath on aarch64 yet.

That makes sense, sorry for pulluting this thread.

[tornaria]

We had a few failures because I patched sagemath to support singular 4.3.1p3, but I didn't upgrade singular (see #42235).

Just to make it easy, I include #42235 here.

[tornaria]

The only test failure in the previous run was a very slow test that triggered a timeout on i686. I fixed the slow test in sagemath/sage#35127 and include the fix here so the timeout doesn't trouble us again.

[tornaria]

@leahneukirchen this is IMO ready to merge.