Tests fail on Windows

[aknuds1]

I noticed there are currently various test failures on Windows after pulling your latest changes into my fork, @harvimt.

[harvimt]

2 steps forward one step back.

Can you roll back to a working commit and just cherry-pick the actual
commit makes the changes that fix the tests on linux?

(since those changes are almost entirely in _unix.py it shouldn't affect
windows)

Then it's probably some of the other changes I made (probably adding more
tests)

On Mon, Dec 15, 2014 at 8:26 AM, Arve Knudsen notifications@github.com
wrote:

I noticed there are currently various test failures on Windows after
pulling your latest changes into my fork.

—
Reply to this email directly or view it on GitHub
#12.

[harvimt]

I'm trying to setup vagrant/ansible to make testing on windows easy (at
least on my end).
(and maybe eventually vagrant + ansible + ptest-xdist so all the output is
in one command

On Mon, Dec 15, 2014 at 4:09 PM, Mark Harviston infinull@gmail.com wrote:

2 steps forward one step back.

Can you roll back to a working commit and just cherry-pick the actual
commit makes the changes that fix the tests on linux?

(since those changes are almost entirely in _unix.py it shouldn't affect
windows)

Then it's probably some of the other changes I made (probably adding more
tests)

On Mon, Dec 15, 2014 at 8:26 AM, Arve Knudsen notifications@github.com
wrote:

I noticed there are currently various test failures on Windows after
pulling your latest changes into my fork.

—
Reply to this email directly or view it on GitHub
#12.

[harvimt]

Tests that are failing (PySide on win2012R2x64)

• test_can_read_subprocess and test_can_communicate_subprocess - problem with test and carriage returns vs new lines easy fix, patch coming
• test_regression_bug13 - Event loop stopped before Future completed

Fixing the test_can_read/communicate_subprocess test fixed the test_regression_bug13 test, this seems to indicate that something is fragile w/ regards to teardown (or something) (the exceptions caused some teardown code to not run, thus affecting the other test further down the line)