Differences between revisions 12 and 13
Revision 12 as of 2012-12-14 01:07:52
Size: 4016
Editor: ?TobiasHansen
Comment: SAGE uses Sympy 0.7.1 now, which is in Debian (see http://www.sagemath.org/packages/standard/)
Revision 13 as of 2012-12-14 01:29:41
Size: 3837
Editor: ?TobiasHansen
Comment: New scipy and numpy situation. The mentioned patch is in Debian.
Deletions are marked like this. Additions are marked like this.
Line 7: Line 7:
 * Scipy. Sage does not yet support scipy 0.8 because it depends on numpy-1.4 and "Sage is not going there yet". Strangely, Debian sid currently has scipy 0.7.2 (good), but includes numpy 1.4.1 (ut-oh). So that will probably create issues.
  * Scipy in Sage is reported to require this patch: http://projects.scipy.org/scipy/changeset/5790

  * pexpect issue (Debian pkg: python-pexpect): Debian ships version 2.3, but Sage requires version 2.0 (though one of the necessary patches is already in 2.4). Plotting in the notebook is likely to break with newer versions that are in Debian; and >2.0 is reported to be slow.
 * Numpy: Sage uses Numpy 1.5, Debian currently ships Numpy 1.6. Don't know if this is a problem.
  * pexpect issue (Debian pkg: python-pexpect): Debian ships version 2.4, but Sage requires version 2.0 (though one of the necessary patches is already in 2.4). Plotting in the notebook is likely to break with newer versions that are in Debian; and >2.0 is reported to be slow.
Line 23: Line 22:
  * Scipy has version 0.10.1 in unstable, while sage uses 0.11.0.

Some people within DebianScience would like to package again Sage for Debian. An old version of Sage got packaged long time ago by Tim Abbott, but, because of lack of maintainance, it became totally unusable and was removed from Debian.

Here is the current list of known or suspected issues:

  • flint&flintqs: sage upstream would like to get rid of flintqs since some recent versions of flint ship with a quadratic sieve (see this ticket, but it hasn't been done yet. The following mail of William Hart tells more about the state of the quadratic sieve code in both flint and flint2

  • opencdk / gnutls. William Stein thinks it is only needed to "provide a secure SSL mode for the Sage notebook". So we can probably use the Debian version of gnutls, but that will require assessment.
  • Numpy: Sage uses Numpy 1.5, Debian currently ships Numpy 1.6. Don't know if this is a problem.
    • pexpect issue (Debian pkg: python-pexpect): Debian ships version 2.4, but Sage requires version 2.0 (though one of the necessary patches is already in 2.4). Plotting in the notebook is likely to break with newer versions that are in Debian; and >2.0 is reported to be slow.

  • ecl is OK unless compiled with Unicode support (http://github.com/cschwan/sage-on-gentoo/issues/closed#issue/2). Debian may do that, so it could be an issue.

  • extcode: a miscellany. Appears to have Tim's debian subdirectory for building the package, jsMath, and tons of other stuff. Most (all?) of it is not a problem (jsMath is already in Debian, but which version is included in Sage is not clear to me), but some things may require finding upstream and building a Debian package. More work needed.
  • Unpackaged dependencies:
  • Dependencies that need to be updated:
    • libflint Bug#592349

    • libfplll0 Bug#592354

    • gfan Bug#592425

    • lcalc Bug#592426

    • m4ri Bug#592429

    • linbox doesn't compile with gcc 4.7 Bug#667264, and the current debian package is an rc, which doesn't have everything needed for sage (?NonZeroRandIter in class ?LinBox::Modular<float>).

    • zn_poly has version 0.8-1.1 in unstable, while sage uses 0.9.
    • Scipy has version 0.10.1 in unstable, while sage uses 0.11.0.

Here are some recommendations for building Sage that may save time/thought:

The old, buggy sagemath Debian package source code is available via git: git://git.debian.org/git/debian-science/packages/sagemath.git and at http://snapshot.debian.org/package/sagemath/3.0.5dfsg-5.1/