This was not a problem before, because we used to unmap the windows.
But now that we use the proper iconified state, the window is still mapped and we still get damage events.
There is nothing wrong with sending window updates, that's how the window managers can show a window preview in the taskbar for example, but we shouldn't be doing it at full speed.
I have a patch which re-uses the go_idle
/ no_idle
methods - but it isn't 100% safe: if the source was idle, we can end up resuming a window-source that is still iconified.
this works - just needs guarding against no_idle called when still iconified
r9542 does this - will backport.
I have also spotted some issues with wm-state updates: "iconic" is an attribute of WindowModel
but we handle state changes in the superclass.. which can cause errors.
"iconic" fixes:
IconicState
for OR windows (safer)
this ticket has been moved to: https://github.com/Xpra-org/xpra/issues/874