Increases android.py coverage from 24% to 36% and total coverage from
25% to 29%.
Note that this tests would have caught the glob issue reported in #1044.
This is a best effort approach trying to increase the coverage of that
module. Therefore some tests are not completely isolating all
dependencies via `mock.patch`-ing.
Subsequent pull request could cover more methods and focus on edge case.
This increases the overall coverage from 20% to 25% and
`buildozer/targets/android.py` coverage from 15% to `24%`.
* Cleaned up some variable references
* Improved how user build dir is accessed
* Minor code style improvements
* Fixed dist_dir calculation following p4a changes
Only if the p4a version used has the file `pythonforandroid/recommendations.py`, which contains the info we want, otherwise we will return the buildozer's one
Running `buildozer unknown_command` should show a meaningful error and
exit. Fixes regression introduced in 4936d31 and adds unit tests.
Also updates other tests `assert` keyword rather than `self.assert*` in
order to keep style consistent.
Last, minor `Dockerfile` documentation update as per recent @tshirtman
feedback.
- run setup.py on both Python2 and Python3
- runs simple test case for both Python2 and Python3
- checks the PEP8 style
- updates README.md adds Travis badge
