giac does not start

[charpent]

Starting with 7.5beta2, sage seems to be unable to start giac. See giac-error.txt

Now compiling giac with --disable-lapack to avoid unwanted links with system libraries

update to giac 1.2.2-103. The new source built with spkg-source to put in upstream/ can be downloaded at: ​http://webusers.imj-prg.fr/~frederic.han/xcas/sage/giac-1.2.2.103.tar.bz2

[charpent]

Session log demonstrating that giac desn't start

[charpent]

Log file of sage -i -c giac starting from sage 7.5.beta2 + database_gap 4.8.3 + dot2tex 2.9.0p1 + fricas 1.2.7 (serial build)

[charpent]

Log of a new, fresh attempt to install giac on a Corei7 + 16 GB RAM running Debian testing. giac installs fine but fails its test suite.

This is a special case : the very same package installed fine this afternoon on a smaller machine (i686 + 7 GB RAM + Debian (64 bits) testing). It might be hardware-related or a corner case.

[frederichan]

From your log, it looks related to this: ../src/icas: symbol lookup error: /usr/lib/libblas.so.3: undefined symbol: gotoblas

What gcc did you used to build giac?

[frederichan]

About your initial question with

sum(1/((x+1)*(2*x-1)),x,0,+infinity)

I can notice a change between giac 1.2.2.95 and 1.2.2.101.

about the link PB, I think that the safest solution is to build giac with lapack disabled because with my config of sage 7.5beta2 giac turned out to be linked to not only to the system libblas but also to the system liblapack. I guess that we won't lose interesting things for sage.

[frederichan]

Last 10 new commits:

​b613d7d	Merge branch 'public/giacpyGB' of git://trac.sagemath.org/sage into develop
​eb69ec5	Merge branch 'master' of git://trac.sagemath.org/sage into develop
​5e4cb99	Merge branch 'public/giacpyGB' of git://trac.sagemath.org/sage into develop
​d17e348	Merge branch 'develop' of git://trac.sagemath.org/sage into develop
​ff5a753	Merge branch 'develop' of git://github.com/sagemath/sage into develop
​3ddf1ca	Merge branch 'develop' of git://github.com/sagemath/sage into develop
​b2d156f	Merge branch 'develop' of git://trac.sagemath.org/sage into develop
​428ab26	Merge branch 'develop' of git://trac.sagemath.org/sage into develop
​0b6ebcd	Merge branch 'develop' of git://github.com/sagemath/sage into develop
​8754023	update giac to 1.2.2.103; --disable-lapack to avoid links toward system libs

[charpent]

• Builds OK
• Passes its own testsuite
• Ptestlong has three failures :

  ----------------------------------------------------------------------
  sage -t --long --warn-long 100.3 src/sage/rings/polynomial/multi_polynomial_ideal.py  # 1 doctest failed
  sage -t --long --warn-long 100.3 src/sage/matrix/matrix_space.py  # Bad exit: 2
  sage -t --long --warn-long 100.3 src/sage/homology/simplicial_complex.py  # 1 doctest failed
  ----------------------------------------------------------------------
  

The latter two are transient (pass when ran standalone), the first one is permanent :

charpent@asus16-ec:/usr/local/sage-7$ sage -t --long --warn-long 100.3 src/sage/rings/polynomial/multi_polynomial_ideal.py
Running doctests with ID 2016-11-17-09-17-43-cce569e2.
Git branch: t/21860/giacblas
Using --optional=database_gap,giac,giacpy_sage,mpir,python2,sage
Doctesting 1 file.
sage -t --long --warn-long 100.3 src/sage/rings/polynomial/multi_polynomial_ideal.py
**********************************************************************
File "src/sage/rings/polynomial/multi_polynomial_ideal.py", line 3533, in sage.rings.polynomial.multi_polynomial_ideal.NCPolynomialIdeal.groebner_basis
Failed example:
    ideal(J.transformed_basis()).change_ring(P).interreduced_basis()  # optional - giacpy_sage
Expected:
    [a - 60*c^3 + 158/7*c^2 + 8/7*c - 1, b + 30*c^3 - 79/7*c^2 + 3/7*c, c^4 - 10/21*c^3 + 1/84*c^2 + 1/84*c]
Got:
    [7*a - 420*c^3 + 158*c^2 + 8*c - 7, 7*b + 210*c^3 - 79*c^2 + 3*c, 84*c^4 - 40*c^3 + c^2 + c]
**********************************************************************
1 item had failures:
   1 of  67 in sage.rings.polynomial.multi_polynomial_ideal.NCPolynomialIdeal.groebner_basis
    [723 tests, 1 failure, 9.10 s]
----------------------------------------------------------------------
sage -t --long --warn-long 100.3 src/sage/rings/polynomial/multi_polynomial_ideal.py  # 1 doctest failed
----------------------------------------------------------------------
Total time for all tests: 9.2 seconds
    cpu time: 15.5 seconds
    cumulative wall time: 9.1 seconds

The review is positve for giac, negative for giacpy_sage. Therefore :

• created #21994, which depends on this ;
• positive_review

[rws]

Confirm there are no errors with make ptestlong and --optional=ccache,gdb,giac,mpir,python2,sage. I didn't install/test giacpy.

[charpent]

Replying to rws:

Confirm there are no errors with make ptestlong and --optional=ccache,gdb,giac,mpir,python2,sage. I didn't install/test giacpy.

I have installed neither ccache nor gdb. But I have database_gap.

./sage -tp 8 --optional=database_gap,giac,mpir,python2,sage -a | tee logs/ptestlong-21860V2.log

gives one failure :

----------------------------------------------------------------------
sage -t --warn-long 91.6 src/sage/homology/simplicial_complex.py  # 1 doctest failed
----------------------------------------------------------------------

which is transient :

charpent@asus16-ec:/usr/local/sage-7$ ./sage -t --warn-long 91.6 src/sage/homology/simplicial_complex.py  
Running doctests with ID 2016-11-17-12-14-36-35f0d332.
Git branch: t/21860/giacblas
Using --optional=database_gap,giac,giacpy_sage,mpir,python2,sage
Doctesting 1 file.
sage -t --warn-long 91.6 src/sage/homology/simplicial_complex.py
    [582 tests, 2.77 s]
----------------------------------------------------------------------
All tests passed!
----------------------------------------------------------------------
Total time for all tests: 2.9 seconds
    cpu time: 2.6 seconds
    cumulative wall time: 2.8 seconds

Does this answer your question ?

[charpent]

One more (irrelevant ?) data point :

On another, smaller machine, rebuilding Sage while forcing the use of system's atlas/blas/lapack via a suitable SAGE_ATLAS_LIB choice allows to install and test successfully the previous version of giac and giacpy_sage.

[vbraun]

Reviewer name should be a real name, not trac account name

[charpent]

Wups ! Corrected.

Not the first time that happens to me... Maybe that should be hinted at in the ticket form legend ?

[tmonteil]

Out of curiosity, will disabling lapack slow down giac ? If yes, would there be a possibility to enable it when possible ?

[parisse]

Disabling lapack will not slow down giac, unless you are doing approx numeric computations with large matrices (n>1000).

[tmonteil]

Thanks fot the info. I plan to make Xcas available from Sage in the next Sage Debian Live release, this is why i was asking.