|
2 | 2 | NumPy 1.14.4 Release Notes |
3 | 3 | ========================== |
4 | 4 |
|
5 | | -This is a bugfix release for bugs reported following the 1.14.3 release: |
| 5 | +This is a bugfix release for bugs reported following the 1.14.3 release. The |
| 6 | +most significa
10000
nt fixes are: |
6 | 7 |
|
7 | | -* float128 values on ppc systems now print more correctly. The whitespace |
8 | | - padding is still incorrect and is to be fixed in numpy 1.15. Numpy still |
9 | | - fails some printing-related (and other) unit tests on ppc systems. |
10 | | -* float values printed in the python2 interactive shell had extra quote marks |
| 8 | +* fixes for compiler instruction reordering that resulted in NaN's not being |
| 9 | + properly propagated in `np.max` and `np.min`, |
| 10 | + |
| 11 | +* fixes for bus faults on SPARC and older ARM due to incorrect alignment |
| 12 | + checks. |
| 13 | + |
| 14 | +There are also improvements to printing of long doubles on PPC platforms. All |
| 15 | +is not yet perfect on that platform, the whitespace padding is still incorrect |
| 16 | +and is to be fixed in numpy 1.15, consequently NumPy still fails some |
| 17 | +printing-related (and other) unit tests on ppc systems. However, the printed |
| 18 | +values are now correct. |
| 19 | + |
| 20 | +Note that NumPy will error on import if it detects incorrect float32 `dot` |
| 21 | +results. This problem has been seen on the Mac when working in the Anaconda |
| 22 | +enviroment and is due to a subtle interaction between MKL and PyQt5. It is not |
| 23 | +strictly a NumPy problem, but it is best that users be aware of it. See the |
| 24 | +gh-8577 NumPy issue for more information. |
11 | 25 |
|
12 | 26 | The Python versions supported in this release are 2.7 and 3.4 - 3.6. The Python |
13 | 27 | 3.6 wheels available from PIP are built with Python 3.6.2 and should be |
14 | 28 | compatible with all previous versions of Python 3.6. The source releases were |
15 | | -cythonized with Cython 0.28.2. |
| 29 | +cythonized with Cython 0.28.2 and should work for the upcoming Python 3.7. |
16 | 30 |
|
17 | 31 | Contributors |
18 | 32 | ============ |
| 33 | + |
| 34 | +A total of 7 people contributed to this release. People with a "+" by their |
| 35 | +names contributed a patch for the first time. |
| 36 | + |
| 37 | +* Allan Haldane |
| 38 | +* Charles Harris |
| 39 | +* Marten van Kerkwijk |
| 40 | +* Matti Picus |
| 41 | +* Pauli Virtanen |
| 42 | +* Ryan Soklaski + |
| 43 | +* Sebastian Berg |
| 44 | + |
| 45 | +Pull requests merged |
| 46 | +==================== |
| 47 | + |
| 48 | +A total of 11 pull requests were merged for this release. |
| 49 | + |
| 50 | +* `#11104 <https://github.com/numpy/numpy/pull/11104>`__: BUG: str of DOUBLE_DOUBLE format wrong on ppc64 |
| 51 | +* `#11170 <https://github.com/numpy/numpy/pull/11170>`__: TST: linalg: add regression test for gh-8577 |
| 52 | +* `#11174 <https://github.com/numpy/numpy/pull/11174>`__: MAINT: add sanity-checks to be run at import time |
| 53 | +* `#11181 <https://github.com/numpy/numpy/pull/11181>`__: BUG: void dtype setup checked offset not actual pointer for alignment |
| 54 | +* `#11194 <https://github.com/numpy/numpy/pull/11194>`__: BUG: Python2 doubles don't print correctly in interactive shell. |
| 55 | +* `#11198 <https://github.com/numpy/numpy/pull/11198>`__: BUG: optimizing compilers can reorder call to npy_get_floatstatus |
| 56 | +* `#11199 <https://github.com/numpy/numpy/pull/11199>`__: BUG: reduce using SSE only warns if inside SSE loop |
| 57 | +* `#11203 <https://github.com/numpy/numpy/pull/11203>`__: BUG: Bytes delimiter/comments in genfromtxt should be decoded |
| 58 | +* `#11211 <https://github.com/numpy/numpy/pull/11211>`__: BUG: Fix reference count/memory leak exposed by better testing |
| 59 | +* `#11219 <https://github.com/numpy/numpy/pull/11219>`__: BUG: Fixes einsum broadcasting bug when optimize=True |
| 60 | +* `#11251 <https://github.com/numpy/numpy/pull/11251>`__: DOC: Document 1.14.4 release. |
0 commit comments