Constrain mouse inside shadow window

[stdedos]

Is your feature request related to a problem? Please describe.
Somewhat related to #3388, I'd like to be able to constrain the mouse pointer within the Shadow window.

Again related to #3388, it could be released on a specific shortcut - VM software have already figured this out

[totaam]

Gdk.pointer_grab used to do this in GTK2 (pygtk):

import pygtk
pygtk.require("2.0")
import gobject, gtk

def release_me():
    gtk.main_quit()
    return True

def on_destroy(widget):
    gtk.main_quit()

def on_enter(widget, event):
    if event.type == gtk.gdk.ENTER_NOTIFY:
        cursor = gtk.gdk.Cursor(gtk.gdk.PIRATE)     # Pirates!!
        gtk.gdk.pointer_grab(widget.window, False, gtk.gdk.POINTER_MOTION_MASK,
                             widget.window, cursor, long(event.time))
        gobject.timeout_add(15000, release_me)
    else:
        # leaving: won't happen
        gtk.gdk.pointer_ungrab(long(event.time))
    return False

def on_move(widget, event):
    print "move to:", event.x, event.y

def main():
    w = gtk.Window()
    w.set_title("Grabby")

    w.connect('destroy', on_destroy)
    w.connect('enter-notify-event', on_enter)
    w.connect('motion-notify-event', on_move)

    w.show_all()

    gtk.main()

    gdk.pointer_ungrab(long(event.time))

if __name__ == '__main__':
    main()

But this doesn't seem to work with GTK3 anymore?

[totaam]

Actually, I just remembered that we already have this ClipCursor workaround: #1229 (comment) - f0301e1.

But I've just tried it and it doesn't seem to confine the pointer, only trap focus / input.
-d grab doesn't show anything useful so I have no idea why it stopped working.
Tested using --key-shortcut=Shift+F2:toggle_pointer_grab on MS Windows 10.

[totaam]

Minimal reproducer

#!/usr/bin/env python3

import gi
from xpra.util import AdHocStruct
gi.require_version('Gtk', '3.0')
from gi.repository import Gtk, GLib
from xpra.platform.win32.gui import pointer_grab

w = Gtk.Window(type=Gtk.WindowType.TOPLEVEL)
w._client = AdHocStruct()
w.show()
GLib.timeout_add(1000, pointer_grab, w)
Gtk.main()

[totaam]

Just found an old note in #1229 : virtualbox's mouse integration interferes with pointer grabs, turn it off for testing
So perhaps everything still works and I'm just testing it wrong!

@stdedos please try toggle_pointer_grab.

[stdedos]

It works as expected.

However ... at least, you should auto-disengage it on Alt+Tab.
Try it: enable_pointer_grab, verify confinement, Alt+Tab, Alt+Tab back, verify confinement: FAIL.
Then, it's not enough to just toggle_pointer_grab once (i.e. re-enable it); instead, you have to do toggle_pointer_grab twice (once to disable it, once to re-enable it).

... Which will also MAY be more complicated to judge if you need to auto-disengage_pointer_grab, if you implement #3388 (then Alt+Tab is directed to the server, instead of client interpreting it).

[totaam]

Try it: ..

I just don't have any hardware available that runs MS Windows native so I can't try it!
(but I'll take your word for it)

instead, you have to do toggle_pointer_grab twice

The first commit above should fix that (totally untested), the second one is not strictly needed.

We track and manage keyboard grabs and pointer grabs separately and I'm not sure that makes any sense but I don't have to courage to change it, especially if I can't test the only platform that actually honours it properly.

[stdedos]

Again, take the VM example (Virtualbox):

"Simple keys" can just pass on as-is. If you'd like to grab the mouse (because I like the constraint), I can. If I need to capture the keyboard (#3388 eg for Alt+Tab), I can

[totaam]

@stdedos is the double-grab issue fixed? Can we close this ticket?

[stdedos]

Yes, it's all good now! Thanks 😄