Permalink
Commits on Dec 28, 2012
  1. REL: Release 1.7.0rc1

    certik committed Dec 28, 2012
  2. Update the C API version number

    certik committed Dec 28, 2012
  3. FIX: remove a spurious check in get_cast_transfer_function

    At least, I hope it's spurious. Certainly it's creating a spurious
    error message, is unexpected by callers (one of whom specifically
    makes the *opposite* check before calling get_cast_transfer_function),
    and even if it is a useful check for some reason I can't see, it
    certainly doesn't belong in this function (which is otherwise just
    taking care of byte-swapping and alignment issues and doesn't know
    anything about dtypes). And worst case, we'll have turned an exception
    into a crash; even if I'm wrong, this shouldn't cause any code to go
    from working to not working, just from broken to slightly-more-broken.
    
    Test and original diagnosis by @cgohlke.
    
    Fixes gh-2798.
    njsmith committed with certik Dec 28, 2012
  4. Merge pull request #2861 from certik/backport2858

    Backport #2858
    certik committed Dec 28, 2012
  5. DOC: Add a blank line to multiline comments

    Multiline comments should begin with a blank line. This patch fixes it.
    certik committed Dec 27, 2012
  6. DOC: Thoroughly document datetime issues

    After understanding the issues with datetime (see gh-568), I have documented my
    understanding as comments in the code, so that the next time there are problems
    with datetime on some platform, one can quickly figure out what is going on
    from the code + comments.
    certik committed Dec 27, 2012
Commits on Dec 27, 2012
  1. Merge pull request #2857 from certik/backport2856

    FIX: Use the NPY_TIME_T macro everywhere
    certik committed Dec 27, 2012
  2. FIX: Use the NPY_TIME_T macro everywhere

    Previously, two (critical) parts of the code used `time_t` instead of
    `NPY_TIME_T`. Due to the fact, that most of the time `NPY_TIME_T` was equal to
    `time_t`, this bug didn't show up. But in mingw, `NPY_TIME_T` is actually equal
    to `__time64_t` and then this causes 64 bit integers to be cast into 32 bit
    integers (thus becoming negative), which causes localtime() to fail in mingw.
    
    Fixes gh-568.
    certik committed Dec 27, 2012
  3. Merge pull request #2737 from certik/fix438

    DOC: document the change in .base
    njsmith committed Dec 27, 2012
Commits on Dec 26, 2012
  1. DOC: document the change in .base

    Fixes gh-438. Based on Nathan's comments in gh-2737.
    certik committed Nov 14, 2012
  2. Merge pull request #2854 from certik/backport2800

    Backports #2800
    certik committed Dec 26, 2012
  3. TST: Add test for in1d ravelling

    The behavior of in1d is not really defined here, but it should
    be at least consistent over different execution branches. This is
    what it has been for most usages.
    seberg committed with certik Dec 16, 2012
  4. BUG: Fix regression for in1d with non-array input

    There was a regression introduced by the speed improvement in
    commit 6441c2a. This fixes it, and generally ravels the arrays for
    np.in1d. However it can be argued that at least the first array should
    not be ravelled in the future.
    
    Fixes "Issue gh-2755"
    seberg committed with certik Dec 8, 2012
  5. Merge pull request #2851 from certik/travis_fix2

    TST: Replace Python 3.1 test with 3.3.
    certik committed Dec 26, 2012
Commits on Dec 21, 2012
  1. Merge pull request #2797 from ContinuumIO/typestring_fix

    Fix invalid typestring size
    njsmith committed Dec 21, 2012
  2. Add fix to unit test; remove debug logging

    Add check to see if 12 bytes is a valid size for a long double
    jayvius committed Dec 21, 2012
  3. Refactor unit tests and add DECREF

    - Refactor unit tests for invalid dtypes so that each test is on a separate line. This will make it easier to figure out which typestring is failing the unit test.
    - Add Py_DECREF for temp variable
    jayvius committed Dec 17, 2012
  4. Fix invalid typestring size

    Revert to pre numpy 1.7 behavior where invalid typestring size was ignored and isplay deprecate warning. This warning should eventually be changed to an error in future numpy versions.
    jayvius committed Dec 7, 2012
  5. TST: Replace Python 3.1 test with 3.3.

    Travis does not support 3.1 anymore, but supports 3.3. See:
    
    https://groups.google.com/d/topic/travis-ci/02iRu6kmwY8/discussion
    certik committed Dec 21, 2012
  6. Merge pull request #2849 from certik/backport2836

    Fix unicode tests for flexible string representation on Python 3.3
    certik committed Dec 21, 2012
Commits on Dec 20, 2012
Commits on Dec 17, 2012
  1. Merge pull request #2842 from certik/backport459

    Backport459
    rgommers committed Dec 17, 2012
Commits on Dec 16, 2012
  1. Merge pull request #2841 from certik/backport433

    MAINT: update Intel Fortran compiler flags for the 21st century.
    certik committed Dec 16, 2012
  2. MAINT: update Intel Fortran compiler flags for the 21st century.

    Thanks to Intel for this contribution.
    Note that there's also an MKL application note which recommends these flags:
    http://software.intel.com/en-us/articles/numpyscipy-with-intel-mkl
    rgommers committed with certik Sep 9, 2012
  3. Merge pull request #2839 from certik/backport476

    [FIX] preserve memory order in np.copy()
    certik committed Dec 16, 2012
  4. [FIX] preserve memory order in np.copy()

    This switches us back to the behaviour seen in numpy 1.6 and earlier,
    which it turns out that scikit-learn (and probably others) relied on.
    njsmith committed with certik Oct 1, 2012
  5. MAINT: Update and cleanup 1.7.0-notes.rst.

    This backports 7f1f57d into the release
    branch.
    certik committed Dec 16, 2012
  6. Merge pull request #2837 from certik/backport2831

    FIX: Fixes the segfault in PyArray_FromScalar
    certik committed Dec 16, 2012