Handle EAGAIN for channel and sftp writes correctly and return rc, by…

…tes_written tuple so clients can resume from last written byte on next call to write.

Updated docstrings for channel write functions, tests, travis cfg, flake8 configuration for sftp.

.travis.yml

@@ -22,7 +22,6 @@ install:
   - python setup.py build_ext --inplace
   - eval "$(ssh-agent -s)"
 script:
-  - export LD_LIBRARY_PATH=/usr/local/lib/x86_64-linux-gnu
   - nosetests
   - flake8 ssh2
   # Test source distribution builds

Changelog.rst

@@ -1,6 +1,17 @@
 Change Log
 =============
 
+0.17.0
++++++++
+
+Changes
+--------
+
+* ``SFTPHandle.write`` function changed to return tuple of ``return_code, bytes_written`` for non-blocking applications to be able to handle partial writes within an SFTP write resulting from a blocked socket.
+* ``Channel.write*`` functions changed to return tuple of ``return_code, bytes_written`` as above.
+
+Behaviour in blocking mode has not changed. Non-blocking applications will now need to handle these functions returning a tuple and resume writes from last written offset of given data.
+
 0.16.0
 +++++++