When playing a video with mpv, RSIbreak thinks the user is active. This results in RSIbreak interrupting the video with a popup. This happens when running the video fullscreen and in window on X windows. A workaround would be to make a script that suspends RSIbreak for the duration of the video via DBus. But that does not cover the situation where a video is playing an the user continues to use mouse and keyboard.
The combination VLC and RSIbreak is fine. So mpv is behaving in a way that lets RSIbreak think there is activity. Easy way to test: play a video not full-screen and after 40 seconds without keyboard or mouse activity (the default) the activity icon should become fully purple instead of 3/4, 1/2 or 1/4 purple.
This is either a bug/feature in mpv (sending events every so often to make sure the screensaver doesn't trigger) or in kidletime, rsibreak just asks kidletime how long the user has been idle, and when mpv is running kidletime says that the user isn't idle. I'd suggest you asking the mpv developers and if they don't think they're doing something fishy we can reassign this to kidletime and then it probably won't ever get fixed but at least will be assigned to the proper product :D
The bug report at mpv is here. https://github.com/mpv-player/mpv/issues/5702 I agree that this might require work in kidletime instead of rsibreak.
according to the investigation in the mpv bug, the issue is that its screensaver/suspend prevention "punches through" to the idle time detection. there isn't really much to be done about that as long as it uses legacy apis - nowadays, it should be using the dbus inhibitor api. but according to my test it still doesn't.
(In reply to Oswald Buddenhagen from comment #4) > according to the investigation in the mpv bug, the issue is that its > screensaver/suspend prevention "punches through" to the idle time detection. > there isn't really much to be done about that as long as it uses legacy apis > - nowadays, it should be using the dbus inhibitor api. but according to my > test it still doesn't. It should be fixed if the Monitoring phase of rsibreak used the "short input" filter just as the resting phase does. I'm not sure whether it wouldn't filter out too much activity