figure out how to test ismount function

[StefanKarpinski]

See #11279. It's unclear how to reasonably test this, especially cross-platform.

[garrison]

Of possible interest: test_macpath.py, test_ntpath.py, and test_posixpath.py

[lobingera]

@StefanKarpinski: Do you mean, how to create a known mounted filesystem? I have not tested this, but it should be possible to put a gzipped iso filesystem into a few hundered bytes. And many linux systems have -ro, user options for /media/cd or similar.
At least the test could be build isaw that automatic regression testing is completing (for UNIX-oid systems).

[nbaum]

Does it help that / will always be a mount point? (On filesystems which have a /, at least.)

Edit: No. I see that ismount has a special case for /, so that only checks one path through the function.

[simonster]

On Linux I'm not sure we can do better than finding a mounted file system (if there is one) and testing against that, since it may not be possible to mount anything new without root access. On OS X we could create a disk image. No idea about Windows.

[malmaud]

@nbaum / will indeed always be a mountpoint and ismount will return true on it.

[yuyichao]

FWIW

1. The root directory is a special case and you are not really testing all the cases.
2. The logic is actually wrong in some cases since / might not be a mount point in a chroot environment.

Not sure how much does any of these matters....

[yuyichao]

And it turns out that ismount also doesn't handle bind mount... The device can be the same for a bind mount...

Edit: and having the same divice across mount point is probably not specific to bind mount since you can mount a device multiple times.... (Haven't tested though)

[yuyichao]

P.S. Git is probably using a similar logic..

[yuyichao]

since it may not be possible to mount anything new without root access

I think fuse doesn't need root access. (But it does require external tools...)