Status of tests on Windows

[Konstantin Veretennicov]

On Thu, Mar 24, 2016 at 12:04 AM, Mads Kiilerich <mads at kiilerich.com> wrote:
> On 03/23/2016 02:11 PM, Konstantin Veretennicov wrote:
>
> On Wed, Mar 23, 2016 at 8:38 PM, Mads Kiilerich <mads at kiilerich.com> wrote:
>
> So what _is_ the status of testing on Windows?
>
> Well, at the moment it is "A few tests are failing regularly and a few
> are flaky".
>
> If interested in details, see 3 test runs for revision
> https://bitbucket.org/conservancy/kallithea/commits/b027fc1a0e85246677d59709a7e418b31f2173c6:
>
> https://ci.appveyor.com/project/kveretennicov/kallithea/build/65/tests
> https://ci.appveyor.com/project/kveretennicov/kallithea/build/66/tests
> https://ci.appveyor.com/project/kveretennicov/kallithea/build/67/tests
>
>
> I guess 7e67883ee300 was insufficient and test_checkout_branch shows that we
> also have to completely ignore stderr unless the exit code is non-zero, with
> something like:
>
> --- a/kallithea/lib/vcs/subprocessio.py
> +++ b/kallithea/lib/vcs/subprocessio.py
> @@ -367,3 +367,3 @@ class SubprocessIOChunker(object):
>          _returncode = _p.poll()
> -        if _returncode or (_returncode is None and bg_err.length):
> +        if _returncode:
>              try:
>
>
> Can you test that?

Test run looks pretty much the same to me with this change:
https://ci.appveyor.com/project/kveretennicov/kallithea-aej5n/build/3/tests

> For the mime types, I guess they come from the OS and the test thus has to
> be different on different platforms ... or we have to make it independent of
> the OS.

Yes, it relies on stdlib's mimetypes.guess_type(). On Windows it gives
"text/plain" for .py.

My understanding is that this "text/plain" comes from Windows
registry. This can be overridden
by early call to mimetypes.init([]). Then we have "text/x-python" on
Windows as well.
It's hard for me to tell, though, whether it will fix or break things...

--
Konstantin

>
> /Mads