Sage 9.1.beta3 released

[Volker Braun]

As always, you can get the latest beta version from the "develop" git branch. Alternatively, the self-contained source tarball is at http://www.sagemath.org/download-latest.html 

1d465c7e3c (tag: 9.1.beta3, trac/develop) Updated SageMath version to 9.1.beta3

e4607fdd24 Trac #29093: Don't “fix” the permissions of preexisting DOT_SAGE

798a8500f4 Trac #29085: CombinatorialPolyhedron: Expose `is_bounded`

c8d4b8a0c7 Trac #29056: numpy: make numpy find openblas from Homebrew

448b7f0853 Trac #28979: Remove period at the end of one-line summary of modules for consistency

577f006eeb Trac #12803: quotient of ideals throws up Singular bug

10fb04d6d3 Trac #29100: spkg-configure.m4 for boost and boost_cropped

b4956f6d77 Trac #29095: Fix sphinxify deprecation warnings with sphinx 2.2

aa4b0b2ff0 Trac #29094: Use rich_to_bool_sgn instead of rich_to_bool when comparing outputs of mpz_cmp

b38e091a96 Trac #29084: Make fflas_ffpack detect and use system openblas on Arch

e410c82f26 Trac #29073: `gale_transform` does not work for `RDF`

2c207ead2b Trac #29071: verify if cblas.pc and lapack.pc should be replaced by links to openblas.pc

ef9af32c09 Trac #29053: Add debian/fedora/arch/conda package information to build/pkgs, generate Dockerfiles and installation help; add tox.ini

b9e43fd0ac Trac #29028: bugs for binary quadratic forms whose discriminant is a square

4429daa651 Trac #29015: Fix frobenius power in non-monic cyclic covers

5d2c5b1be7 Trac #28953: spring cleanup in scripts

1a7f090bb5 Trac #28869: Isomorphisms of absolute fields can not be used in change_ring

4d21d4f613 Trac #28095: Add --enable-OPTIONALSPKG options to configure

94056f864b Trac #28054: invariant form cannot be specified for GO/SO over a finite field

6f626864aa Trac #27954: Integral curves over finite fields: closed points

79401117b9 Trac #26421: Polynomial roots cannot be calculated over a ring of Laurent polynomials

3373ddc9b1 Trac #26363: Polyhedron_normaliz.save

3b41e5cd3c Trac #26130: fflas_ffpack is not tested, tests fail

67759239ba Trac #25509: Provide a simple (and generic) way to create immutable vectors and matrices

3a2f088d88 Trac #29086: support latex \circ

e269e10d7d Trac #29078: Move `is_simple` and `is_simplicial` to `CombinatorialPolyhedron`

5abdaf4583 Trac #29077: get rid of six.iterkeys

1909f1681d Trac #29076: multivariate factorization in AA[x,y] broken in 9.0

da04c4fc09 Trac #29051: Repair use of system BLAS broken by #29025: remove numpy site.cfg [DEFAULT/ALL] sections

2cbd93e0b7 (tag: 9.1.beta2) Updated SageMath version to 9.1.beta2

[Matthias Köppe]

On Sunday, February 2, 2020 at 5:52:32 PM UTC-5, Volker Braun wrote:

As always, you can get the latest beta version from the "develop" git branch. Alternatively, the self-contained source tarball is at http://www.sagemath.org/download-latest.html 

1d465c7e3c (tag: 9.1.beta3, trac/develop) Updated SageMath version to 9.1.beta3 [...]

A partial build of 9.1.beta3 (packages cbc csdp fflas_ffpack gsl iml numpy r suitesparse cvxopt rpy2 ecl maxima) on a number of platforms had the following results.

- ubuntu-trusty minimal/standard ... build error for fflas_ffpack

- ubuntu-trusty minimal/standard OK

- ubuntu-bionic/eoan/focal minimal and debian-sid minimal ... build error for R

- ubuntu-bionic/eoan standard OK

- ubuntu-focal standard and debian-sid/bullseye standard ... build error for ecl

- debian-jessie minimal ... build error for gfortran

- debian-jessie/buster standard ... build error for fflas_ffpack

- debian-buster minimal ... build error for R and fflas_ffpack

- debian-bullseye minimal ... build error for R

- fedora-26/31/32 minimal and centos-7 minimal ... build error for gfortran

- fedora-26/28/31 standard OK

- fedora-32 standard ... build error for openblas

- centos-7 standard ... build error for fflas_ffpack

- centos-8 minimal ... build error for pcre, yasm

- centos-8 standard ... build error for pcre, gsl

- archlinux-latest minimal/standard ... build error for R

- conda-forge minimal/standard ... build error in python3

- ubuntu-bionic-i386 standard OK

- homebrew minimal/standard OK

Logs are at https://github.com/mkoeppe/sage/actions/runs/34012261 

These platforms are defined in SAGE_ROOT/tox.ini (see https://github.com/sagemath/sage/blob/develop/tox.ini; added in https://trac.sagemath.org/ticket/29053, merged in 9.1.beta3). See also https://trac.sagemath.org/ticket/29087 (GitHub actions workflow).

A ticket fixing the gfortran failures is at https://trac.sagemath.org/ticket/29089 .

R, fflas_ffpack, and ecl seem to require attention.

[Dima Pasechnik]

fflas_ffpack problems are manifesting themselves on old gcc, which do
not know about vx512, see
https://trac.sagemath.org/ticket/29102

symptoms:
2020-02-02T23:41:08.9253838Z libtool: compile: g++ -std=gnu++11
-DHAVE_CONFIG_H -I. -I../../.. -I../../.. -O2 -Wall -g
-I/sage/local/include -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2
-mavx -mavx2 -mavx512f -mavx512vl -mavx512dq -mfma -fabi-version=6
-I/sage/local/include -I/sage/local/include -fopenmp -c ffpack.C
-fPIC -DPIC -o .libs/ffpack.o
2020-02-02T23:41:08.9254140Z g++: error: unrecognized command line
option '-mavx512f'
2020-02-02T23:41:08.9254453Z g++: error: unrecognized command line
option '-mavx512vl'
2020-02-02T23:41:08.9254720Z g++: error: unrecognized command line
option '-mavx512dq'


ecl fails on new libffi, see https://trac.sagemath.org/ticket/29128

the R problem is due to a Fortan problem, already discussed here at length.
symptoms:
2020-02-03T00:13:19.7256378Z gfortran -fno-optimize-sibling-calls
-fpic -g -O2 -c dpbfa.f -o dpbfa.o
2020-02-03T00:13:19.7256491Z In file included from
../../src/include/R_ext/Applic.h:41,
2020-02-03T00:13:19.7256578Z from maxcol.c:37:
2020-02-03T00:13:19.7256895Z ../../src/include/R_ext/BLAS.h:103:55:
error: expected declaration specifiers or '...' before ')' token
2020-02-03T00:13:19.7257012Z const double *beta, double *y, const
int *incy FCLEN);
2020-02-03T00:13:19.7257118Z
^
2020-02-03T00:13:19.7257434Z ../../src/include/R_ext/BLAS.h:110:35:
error: expected declaration specifiers or '...' before ')' token
2020-02-03T00:13:19.7257533Z double *y, const int *incy FCLEN);
2020-02-03T00:13:19.7257629Z ^

[Eric Gourgoulhon]

On Ubuntu 18.04 running on a bi-Xeon E5-2623 (8 cores) + 16 GB RAM computer:

  - incremental build (with -j16) from 9.1.beta2: OK

  - make ptestlong reports 1 failure:

sage -t --long --warn-long 53.7 src/sage/groups/cubic_braid.py  # 1 doctest failed

Details in the attached log.

The test is passed when run standalone.

Eric.

[Harald Schilly]

some positive feedback: I wasn't able to build beta1 and beta2 for cocalc (that scipy error, not able to find libs), but beta3 works!

-- h

[Dima Pasechnik]

I've fixed the ECL/libffi problem on
https://trac.sagemath.org/ticket/21811 (needs review)

[Emmanuel Charpentier]

After upgrading Sage 9.2.beta2 to 9.2.beta3 on Deban testing running on core i5 + 8 GB RAM:

• ptestlong passes with no failures;
  
• ptestalllong fails when running findstat.py ; this kills findstat.py, octave.py, magma_free.py, src/doc/en/faq/index.rst, src/doc/en/prep/Logging-On.rst and the ptestallong process itself, apparently.
  

In other words, this is quite similar to what I reported for 9.1.beta2 on the same machine. Logs available if requested.

HTH,

[Emmanuel Charpentier]

On a similar setup^(corei7 + 16 GB RAM, Debian testing, no local proxy, gap_packages installed), upgrading 9.1.beta2 to 9.2.beta3 leads to:

• ptestlong passes with three permanent failures on gap_packages.py, glpk_backend.pyx and glpk/error.pyx;
• ptestallong fails after a crash on findstat.py, leading to the interruption of other running tests.

These results look identical to what has been already reported  for 9.1.beta2.

HTH,

[Justin C. Walker]

> On Feb 2, 2020, at 14:52 , Volker Braun <vbrau...@gmail.com> wrote:
>
> As always, you can get the latest beta version from the "develop" git branch. Alternatively, the self-contained source tarball is at http://www.sagemath.org/download-latest.html

Built w/o problems from fresh clones of the develop branch, on three macOS systems: 10.13.6 (2017 iMac Pro, 18-core Xeon W), 10.14.6 (2017 MBP, Quad-core Core i7), and 10.15.2 (2019 MBP, 8-core Core i9).

10.13.6: all tests passed!

10.14.6: all tests passed!

10.15.3: all tests passed!

Justin

--
Justin C. Walker, Curmudgeon-At-Large
Institute for the Absorption of Federal Funds
--------
If you're not confused,
You're not paying attention
--------

[Emmanuel Charpentier]

On the same setup, rebuilding 9.1.beta3 + Trac#21811 after make distclean gives the same results : the same three permanent failures after ptestlong, test crash during ptestalllong. I've checked that the latter is reproduced even after raising the number of open files (ulimit -n 4096, instead of the default 1024).

HTH, 

[Emmanuel Charpentier]

FWIT, still on the same machine, a serial make ptestlong terminates, giving three permanent and four transient failures:

Running doctests with ID 2020-02-09-11-16-06-10503c7a. Git branch:
t/21811/packages/ecl/ffi_abi_libffi33.patch Using
–optional=build,dochtml,dot2tex,external,fricas,gap_packages,giacpy_sage,libsemigroups,memlimit,sage
External software to be detected:
cplex,ffmpeg,graphviz,gurobi,imagemagick,internet,latex,macaulay2,magma,maple,mathematica,matlab,octave,pandoc,scilab

File	Result	P/T
src/sage/numerical/backends/glpk_backend.pyx	1 doctest failed	P
src/sage/libs/glpk/error.pyx	1 doctest failed	P
src/sage/databases/findstat.py	8 doctests failed	T
src/sage/plot/animate.py	7 doctests failed	T
src/sage/misc/latex.py	1 doctest failed	T
src/sage/combinat/tutorial.py	1 doctest failed	T
src/sage/tests/gap_packages.py	1 doctest failed	P

External software detected for doctesting:
ffmpeg,graphviz,imagemagick,internet,latex,mathematica,pandoc

This hints at a (concurrency ?) problem in the test infrastructure itseff...

HTH,