Permalink
Commits on Feb 20, 2011
  1. TST: explicit file close to avoid ResourceWarnings under Python 3.2.

    rgommers committed Feb 20, 2011
    Thanks to Christoph Gohlke.
    
    (backport of r7147)
  2. BUG: remove ASCII form feed character that causes a problem with Py 3…

    rgommers committed Feb 20, 2011
    ….2rc3.
    
    Thanks to Christoph Gohlke for testing and the patch.
    
    (backport of r7156)
  3. BUG: special: fix zbiry prototype to match the Fortran code (#1233)

    rgommers committed Feb 20, 2011
    (backport of r7144)
Commits on Feb 19, 2011
  1. BUG: spatial/interpnd: never deallocate memory in a different Cython …

    rgommers committed Feb 19, 2011
    …module
    
    Different Cython modules end up as different DLL/.so files, and
    apparently on some platforms freeing memory allocated in a different DLL
    is not allowed. This seems to be true at least on Windows 7 + 32-bit
    Python.
  2. REL: set released=False

    rgommers committed Feb 19, 2011
Commits on Feb 12, 2011
Commits on Feb 10, 2011
  1. TST: Use np.seterr in test generators in Python2.4 compatible way

    rgommers committed Feb 10, 2011
    The seterr should be called in the test routine itself and not in the
    test generator.
    
    (backport of r7123)
  2. BUG: io.idl: Fix readsav() on Python 2.4

    rgommers committed Feb 10, 2011
    (backport of r7122)
  3. TST: optimize: add some leeway for additional FP rounding error

    rgommers committed Feb 10, 2011
    For instance, if arrays happen to be not aligned on 16-byte boundaries
    when using MKL, this produces additional (insignificant) rounding error.
    The tests should allow for this.
    
    (backport of r7121)
  4. BUG: Fix ndimage type checks, part of ticket #1724.

    rgommers committed Feb 10, 2011
    (backport of r7120)
  5. TST: io: close file handle before unlinking, for win32 (#1379)

    rgommers committed Feb 10, 2011
    Thanks to C. Gohlke for the patch.
    
    (backport of r7109)
Commits on Jan 31, 2011
  1. TST: fftpack: add tests checking fft routine overwrite behavior

    pv committed Jan 31, 2011
    (backport of r7111)
  2. BUG: fftpack: fix data overwriting in FFT routines (#1353)

    pv committed Jan 31, 2011
    The _fix_shape helper routine in some cases returns slices of the input
    array, which does not create copies. In those cases, one should not set
    overwrite_x=1.
    
    (backport of r7110)
Commits on Jan 30, 2011
  1. BUG: fix for build issue with MSCV 9, ifort and MKL. Patch by Christo…

    rgommers committed Jan 30, 2011
    …ph Gohkle. Closes #1376.
    
    (backport of r7103)
  2. TST: fix pilutil.imresize test. Patch by Mark Wiebe.

    rgommers committed Jan 30, 2011
    (backport of r7092)
Commits on Jan 29, 2011
  1. BUG (backport r7095): sparse: fix crash on deref of std::vector of si…

    pv committed Jan 29, 2011
    …ze 0 (fixes #1210)
    
    (cherry picked from r7095)
Commits on Jan 28, 2011
Commits on Jan 25, 2011
  1. BUG: Add back 2002 codata values.

    rgommers committed Jan 25, 2011
    (backport of r7085)
Commits on Jan 23, 2011
  1. TST: relax test precision for new stats.expect tests a little. Was fa…

    rgommers committed Jan 23, 2011
    …iling on win32.
    
    (backport of r7080)
  2. DEP: Remove old behavior of correlate/convolve.

    rgommers committed Jan 23, 2011
    (backport of r7075)
  3. GEN: sparse: re-swig sparsetools with swig 2.0.1 + PyCapsule hack

    rgommers committed Jan 23, 2011
    The files were swig-ed with Swig 2.0.1 with the patch from
    http://sourceforge.net/tracker/?func=detail&aid=3047039&group_id=1645&atid=301645
    
    (backport of r7071)
  4. BUG: Fix a call in maxentropy to removed sparse method.

    rgommers committed Jan 23, 2011
    (backport of r7072)
Commits on Jan 20, 2011
  1. TST: Avoid ComplexWarning in QR and Schur decomposition.

    rgommers committed Jan 20, 2011
    lwork returned from a workspace query has the same dtype as the
    principal matrix, but when we call the computational routine it
    expects an integer value.
    
    This also fixes a potential bug in schur(), where lwork keyword was
    ignored.
    
    Thanks to Fabian Pedregosa.
    
    (backport of r7069)