[merge] win32- all tests pass but 1

[Aaron Bentley]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

John Arbash Meinel wrote:
> Aaron Bentley wrote:
> It may not even be required for the caller to close() the file handle,
> as long as _KnitData doesn't hold it open.

Since finalizer invocation isn't deterministic, I think that's a race
condition.

> These are the 'KnownFailures'. I would be okay with creating a new
> exception class, like TestSkippe, maybe TestKnownFailed, or something
> like that.
> I return here only because a plain return in setUp doesn't do anything.

Fine, I won't insist on doing a TestSkipped.  Clearly it means something
different to you than it means to me.

>>> This code looks like it could fail because it's joining a url and a
>>> local filesystem path.  Am I wrong about that?
> 
> It could fail if the working tree was a non ascii path.
> I could easily call local_path_to_url() instead, but I was thinking we
> were actually asserting that you can use the cwd() in a simple file://
> url. I'd be happy to switch to a real url if you prefer.

I don't think it's right to assert you can use the cwd in a simple file
URL, because it's easy to imagine situations where people are in
non-ascii urls, and don't have any control over that.

> I'd be okay either way with the TestSkipped stuff. I was just trying to
> avoid Skip when we really mean
> NotGoingToRunBecauseItDoesntWorkAndItIsOldAndUnsupported.

But it's actually a case of NotGoingToRunBecausePlatformSucks.

Okay, I can live with returning for now.  So you have my +1 if you
switch the cwd thing to a real URL.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFEpiEx0F+nu1YWqI0RAmtMAJ4oXiZJDoJ9J0yvHzLi1Z46Lfd9BQCgiaZH
s7z7iuFQJyPJlK+DZR8/snQ=
=MPV5
-----END PGP SIGNATURE-----