Hack around Python 2 ASCII encoding bug / incompatibility

[PiDelport]

For context, see previous PRs #4 and #7.

This expands the test coverage to run against ASCII, adds a workaround for PythonCharmers/python-future#256 (which testing against ASCII runs into), and enables the hack for Python 2 ASCII (see comments).

[codecov]

Codecov Report

Merging #8 into master will decrease coverage by 69.63%.
The diff coverage is 38.09%.

@@             Coverage Diff             @@
##           master       #8       +/-   ##
===========================================
- Coverage   95.83%   26.19%   -69.64%     
===========================================
  Files           5        5               
  Lines         168      668      +500     
  Branches       26      112       +86     
===========================================
+ Hits          161      175       +14     
- Misses          4      488      +484     
- Partials        3        5        +2

Impacted Files	Coverage Δ
tests/test_os.py	100% <ø> (+14.28%)	⬆️
tests/test_extra.py	100% <100%> (ø)	⬆️
src/backports/os.py	14.13% <7.14%> (-79.46%)	⬇️
tests/test_helpers.py	95.83% <0%> (-4.17%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 848c092...f69d58f. Read the comment docs.

[pombredanne]

Excellent! I wonder if vstinner/misc@a5f90a0#diff-b500f48c9778753dc97ebc453352f351R80 was not also about this....
FWIW I also stumbled on this other implementation https://github.com/chromium/caterpillar/blob/master/src/surrogateescape.py but I am not sure it works entirely the same.

I confirm that this now works correctly work with your fix using a degenerated locale!
Thanks ++
@sschuberth FYI.

[PiDelport]

Hmm, I'm not sure what's up with the test coverage reporting. Something seems to be going wrong with the collection. :/

Running a coverage report locally gives the expected coverage, so I'm going to assume something with the CI integration is wonky; I'll go ahead and merge this, and worry about the CI reporting later.

[pombredanne]

FWIW the coverage issue may be about how codecov merges reports from multiple runs in the CI matrix...

[pombredanne]

And you even pushed this to Pypi... https://pypi.python.org/pypi/backports.os/0.1.1 Thank you very much!