[MERGE][0.15] Small updates for win32 support

[John Arbash Meinel]

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

Alexander Belchenko wrote:
> John Arbash Meinel ?8H5B:
>>> The attached patch has 2 fixes for win32. Not all tests pass, but at
>>> least more of them pass than before.
> 
> Your patch introduce new fails during selftest like this:
> 
> ERROR: #1106 test_revert.TestRevert.test_preserve_execute
>     <WorkingTree4 of C:/Temp/jam/test0000.tmp/1K/01106/work/tree> is not locked
> 
> And there is still some LockError exists.
> Gzipped log attached.
> 
> [µ]
> 

Well it seems a little weird to call WorkingTree.is_executable() when
you don't have a lock on the tree.

However, the old code paths called id2path() which *does* have a
@needs_read_lock on it, So I can just decorate is_executable() with
@needs_read_lock.

I'm running the test suite right now. I'm pretty sure there will be some
failing tests, but it should be better than it is has been.

There are other things we need to deal with, though. One thing Wouter
Van Heyst stumbled upon, is that 'bzr status' isn't updating the
dirstate with new values. This is because it can't grab a write lock on
the file (since it already has a read lock).

I don't know of a way to upgrade an os shared lock to an exclusive lock
without unlocking in the middle. But it is something I'm going to look into.

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFF8eJHJdeBCYSNAAMRAr9VAKCQo9FoZemGk9+1ANmdIDGKxKg7hQCgqQzD
83QgoaGZ6mHJZ6uRRzNOPDs=
=ffds
-----END PGP SIGNATURE-----