issue9035.3.patch has switched to the new memory management API and has
tweaked the tests slightly for robustness.
This approach does introduce a behavioural change: the root of a SUBSTed
drive (essentially a symlink into the Dos namespace) will raise an
OSError because GetVolumePathName returns error 87: invalid parameter.
So os.path.ismount("F:\\") will fail where F: is the result of running,
eg, "SUBST F: C:\temp".
I think the simplest thing is to special-case drive roots (which are
always mount points) and then to apply the new GetVolumePathName logic.