I have an ATMega32u4 as an HID.
It takes commands on an interrupt endpoint.
It provides a realtime status out on an interrupt endpoint with a bInterval of 16.
It works perfectly when connected to a Windows app or an Android app.
No complaints there.
To double check polling on the in endpoint (and how bInterval got rounded),
I added some test code to the ATMega to toggle a test point when the in endpoint got loaded.
I saw what I expected: bInterval=10, poll=8, bInterval=16, poll=16.
But... On the Android when the the Android app was not running there was no polling.
On the Windows when the Windows app was not running there was still polling going on.
I don't have a Linux app, but I connected it to a RPi and polling went on when I cat /dev/hidraw0 and goes off when I ^C.
Ok, this does not really have anything to do with the USB standard exactly.
It's more of a question of whether a host should leave USB devices alone if the user is not using them.
Has anybody seen this sort of stuff or have an explanation?
I guess that MS wants to keep HID input buffers full in case somebody might want to use it.
