Comment 59 for bug 1295267

I think I found a reason why this happens.
I left xrandr running in a loop, logging everything and then I forced both displays into standby with xset.

After checking the logs this is what I noticed:

DP-0 connected primary 3840x2160+0+0 (normal left inverted right x axis y axis) 597mm x 336mm
DVI-D-0 connected 1920x1080+3840+0 (normal left inverted right x axis y axis) 510mm x 290mm
DP-0 connected primary 3840x2160+0+0 (normal left inverted right x axis y axis) 597mm x 336mm
DVI-D-0 connected 1920x1080+3840+0 (normal left inverted right x axis y axis) 510mm x 290mm
DP-0 disconnected (normal left inverted right x axis y axis)
DVI-D-0 connected primary 1920x1080+0+0 (normal left inverted right x axis y axis) 510mm x 290mm
DP-0 connected primary 3840x2160+0+0 (normal left inverted right x axis y axis) 597mm x 336mm
DVI-D-0 connected 1920x1080+3840+0 (normal left inverted right x axis y axis) 510mm x 290mm
DP-0 connected primary 3840x2160+0+0 (normal left inverted right x axis y axis) 597mm x 336mm
DVI-D-0 connected 1920x1080+3840+0 (normal left inverted right x axis y axis) 510mm x 290mm

When displays go to low power mode my 4K screen is briefly turned off and then back on. When 4K screen is turned off, the 1080p becomes primary and and all the windows are moved to it, when screen is connected back and windows are restored everything is put in the wrong place.

This seems like a wrong implementation of DPMS with the screen manufacturer, not much can be done about this, but WM could go around this.

(strangely enough, same thing happens if I just turn off my HiDPI screen, I don't know how I didn't remember this before)

Which brings up a separate problem - handling of windows when one screen is turned off.

WM should remember position of windows when one screen is disconnected. Windows should be then placed on the _visible_ area on the remaining screen. When previous screen is connected again if windows were not moved, they should go back to their original positions.