Release 3.5.2

Maintenance release to fix compatibility with Python 3.8

Get ready for a 3.5.1 release

Sometimes admitting the defeat is the wisest thing to do

Merge pull request #696 from PyTables/develop

Release 3.4.4

Merge pull request #681 from PyTables/develop

Release 3.4.3

Merge pull request #622 from tomkooij/v3.4.2

release 3.4.2

Release 3.4.1

Changes from 3.4.0 to 3.4.1
===========================

Bugs fixed
----------
 - Fixed bug in ptrepack

Release 3.4.0

Changes from 3.3.0 to 3.4.0
===========================

Improvements
------------

- Support for HDF5 v1.10.x (see :issue:`582`)
- Fix compatibility with the upcoming Python 2.7.13, 3.5.3 and 3.6 versions.
  See also :issue:`590`. Thanks to Yaroslav Halchenko
- Internal Blosc version updated to 1.11.3
- Gracefully handle cpuinfo failure. (PR #578)
  Thanks to Zbigniew Jdrzejewski-Szmek
- Update internal py-cpuinfo to 3.3.0. Thanks to Gustavo Serra Scalet.

Bugs fixed
----------

- Fix conversion of python 2 `long` type to `six.integer_types` in atom.py.
  See also :issue:`598`. Thanks to Kyle Keppler for reporting.
- Fix important bug in bitshuffle filter in internal Blosc on big-endian
  machines. See also :issue:`583`.
- Fix allow for long type in nextafter. (PR #587) Thanks to Yaroslav Halchenko.
- Fix unicode bug in group and tables names. :issue:`514`

Release 3.3.0

Change version number to 3.3.0.

Release notes:

Changes from 3.2.3.1 to 3.3
===========================

Improvements
------------

- Single codebase Python 2 and 3 support (PR #493).
- Internal Blosc version updated to 1.11.1 (closes :issue:`541`)
- Full BitShuffle support for new Blosc versions (>= 1.8).
- It is now possible to remove all rows from a table.
- It is now possible to read reference types by dereferencing them as
  numpy array of objects (closes :issue:`518` and :issue:`519`).
  Thanks to Ehsan Azar
- Get rid of the `-native` compile flag (closes :issue:`503`)
- The default number of threads to run numexpr (MAX_NUMEXPR_THREADS)
  internally has been raised from 2 to 4.  This is because we are in
  2016 and 4 core configurations are becoming common.
- In order to avoid locking issues when using PyTables concurrently in
  several process, MAX_BLOSC_THREADS has been set to 1 by default.  If
  you are running PyTables in one single process, you may want to
  experiment if higher values (like 2 or 4) bring better performance for
  you.

Bugs fixed
----------

- On python 3 try 'latin1' encoding before 'bytes' encoding during unpickling
  of node attributes pickled on python 2. Better fix for :issue:`560`.
- Fixed Windows 32 and 64-bit builds.

Fix numpy 1.9 compatibilty

numpy.random.randint(..., dtype='int64') is not supported in numpy 1.9.
Changed to .astype('int64')