The bindings looks sane to me. Looking at the API I'm thinking, that this is basicly a OO model. This might be an optional enhancement (not finished, but should give the idea):

matthiasblaesing@ba02d6b

Great idea. I'll update. One thing to note, the docs specify return types for both the *_ref() and *_unref() functions: ref returns the unmodified object and unref "always returns null". I do agree with changing the unref return to void (since testing shows it doesn't actually return null as advertised) but I think the ref version should return this;.

Hmm, I expect the "double-free" is me trying to release parent reference. This doc is confusing but I suppose it means "don't unref it"...

On success, udev_device_get_parent() and udev_device_get_parent_with_subsystem_devtype() return a pointer to the parent device. No additional reference to this device is acquired, but the child device owns a reference to such a parent device.

Sorry for the 8 or 9 force pushes. The good news is my own appveyor setup is pretty strict on javadocs. The bad news is that it was killing the build too quickly for me to see the errors in the logs and I kept finding and fixing a few at a time. It should be good to go now.

Merged - thank you