Added option to manually set messages encoding for pytest plugin

[capgelka]

I had a problem on teamcity agent - automatically detected encoding is cp1252 which is wrong and I got encoding error. So I added option to allow manually setting encoding.

[shalupov]

Could you provide steps to reproduce the problem?

I've actually run your example with one modification:

# coding=utf-8
import pytest

def test_cp():
    assert 'Ф'.encode('utf8') != b'\xd0\xa4'

python 3.3 (3.5 also ok) on Windows 10 (russian oem encoding):

sys.stdout.encoding: cp866
============================= test session starts =============================
platform win32 -- Python 3.3.5, pytest-2.9.1, py-1.4.31, pluggy-0.3.1
rootdir: C:\work\teamcity-messages, inifile:
plugins: teamcity-messages-1.17
collected 1 items

encoding_test.py ##teamcity[testStarted timestamp='2016-05-21T15:16:58.413' captureStandardOutput='false' flowId='tests.guinea-pigs.pytest.encoding_test.test_c
p' name='tests.guinea-pigs.pytest.encoding_test.test_cp']
F##teamcity[testFailed timestamp='2016-05-21T15:16:58.425' details='def test_cp():|n>       assert |'Ф|'.encode(|'utf8|') != b|'\xd0\xa4|'|nE       assert b|'\
xd0\xa4|' != b|'\xd0\xa4|'|nE        +  where b|'\xd0\xa4|' = <built-in method encode of str object at 0x00000000032F3C90>(|'utf8|')|nE        +    where <buil
t-in method encode of str object at 0x00000000032F3C90> = |'Ф|'.encode|n|nencoding_test.py:5: AssertionError' flowId='tests.guinea-pigs.pytest.encoding_test.te
st_cp' message='tests\guinea-pigs\pytest\encoding_test.py:3 (test_cp)' name='tests.guinea-pigs.pytest.encoding_test.test_cp']
##teamcity[testFinished timestamp='2016-05-21T15:16:58.427' duration='0' flowId='tests.guinea-pigs.pytest.encoding_test.test_cp' name='tests.guinea-pigs.pytest
.encoding_test.test_cp']


================================== FAILURES ===================================
___________________________________ test_cp ___________________________________

    def test_cp():
>       assert 'Ф'.encode('utf8') != b'\xd0\xa4'
E       assert b'\xd0\xa4' != b'\xd0\xa4'
E        +  where b'\xd0\xa4' = <built-in method encode of str object at 0x00000000032F3C90>('utf8')
E        +    where <built-in method encode of str object at 0x00000000032F3C90> = 'Ф'.encode

encoding_test.py:5: AssertionError
========================== 1 failed in 0.03 seconds ===========================

[capgelka]

Unfortunately I can't. The problem with wrong encoding was only on teamcity agent, not my local machine. If I had known what make windows to use cp1252 by default, I would have changed windows setting to solve it.

Tests from ccf6bba commit just to check if new option was applied. I added because thought it would be bad pull request with new option not covered by tests. I am sorry if it broken CI, on my local machine new tests passed.
In both new tests internal pytest test should not passed, so your example if it was for [https://github.com//pull/100/commits/ccf6bba7292f3b597875e34c8e8cae2f9222a194#diff-87d3cd8521c2572c31dc48328dfe9448R69]? not for test without manually set encoding failed as expected.
Maybe such kind of integration test with failed internal test is not good. But I still have no idea how one can better test this option.

[shalupov]

As far as I know it's about OEM encoding (aka codepage for non unicode programs), try https://superuser.com/questions/497857/how-to-change-language-for-non-unicode-programs-in-windows-8

I'll think about the integration test