Bug 26922

Thanks for reporting this. this isn't really a bug though I've heard that it might be a broken feature (ahh, the wording, beautiful, isn't it?).

X has a principle of an implicit passive grab where the pointer device is grabbed when a button press event was delivered to a client (until the release event). In xinput's case, test-xi2 only registers on master devices for raw events. Once the grab is on the pointer, these events only go to the grabbing client.

If you change the code to register for raw events on XIAllDevices instead, you'll see the events from the physical devices. Likewise, with a second pointer you'll see the events for the second pointer even when the first pointer has a button down.

So, not a bug though I'm quite willing to admit that delivering raw events only if there's no grab on the device may prove to be a bad choice in the future.

*** Bug 26921 has been marked as a duplicate of this bug. ***

Unfortunately I'm already using XIAllDevices with the same exact results.

Created attachment 34197 [details]
Sample test prorgam

Sample test program that demonstrates the issue.

(In reply to comment #4)
> Created an attachment (id=34197) [details]
> Sample test prorgam
> 
> Sample test program that demonstrates the issue.

ok, I tried it and changed the device id to my mouse. that worked. then changed it to XIAllDevices and that gave me three events - press from the MD pointer and press + release from the slave. That's what I expect since my WM puts a button 1 grab on the MD pointer.

if you comment out line 63 (eventmask.deviceid = 7;), you should get the same events.

I see. This all seems to be broken. If the master device supposed to be an aggregate of slaves but it's not...

At least should spell this out somewhere in documentation.

reopening to keep track of it. This should probably be fixed in the protocol eventually.

Summary: raw events in XI 2.0 are only sent if no grab is active. If a grab activates, raw events are only sent to the grabbing client.
The implicit passive grab activated on a ButtonPress delivery causes asymmetric event delivery. RawEvents are processed first, hence the raw button press is sent, then the grab activates at the core event delivery and the raw button release is discarded - the grab deactivates only after the core event is delivered.

This needs to be fixed either by rearranging the raw event order for button releases or by handling the whole thing in a better manner inside the server.

Mass closure: This bug has been untouched for more than six years, and is not obviously still valid. Please file a new report if you continue to experience issues with a current server.