-*- text -*- * Use the new xcb-ewmh for the EWMH hints. * Bug: We grab MODKEY all the time! We can grab it only when we start tabbing instead and release it when tabbing is complete. * Flag (-i, perhaps?) to allow windows to be unmapped. Add unmap patch from Christian. * Key to move to previous and next workspace. David Jacobs has a patch for mod4+shift + I/O. https://gist.github.com/1478365 * Key to move windows to other workspaces, perhaps mod4+Shift + 0..9 and mod4+shift + I/O. * Full-screen windows that are resized from client should be kept full-screen. * When moving windows between physical screens, try to place the window on roughly the same place on a new screen, if possible. * When a window tries to map itself outside the physical screens, always map it to the screen closest to the coordinates it asked for instead of always mapping on the first screen in the list. * Feature: We need to continue dragging for move and resize until *both* MODKEY and mouse button has been released. Will be much nicer with trackball. * Move to first window (or middle of screen) on another screen with MODKEY + . and , instead of moving windows to new screen. Shifted these keys will move window to new screen? * Handle new modes on physical outputs. What do we have to do? * A separate workspace list for every physical output. * Allow hexadecimal colour values on command line. * Grab specific keys Current code grabs modifier keys and any other keypress. We need to specifically grab the keycodes we need. xbindkeys and possibly other programs don't work. Partially done. Now grabs all our keys unshifted as well as shifted. * Feedback window We may need to tell the user the new geometry somehow. Is this what I want? The feedback window can also show what workspace we just changed to. Can we cooperate with a stand-alone program? Write our status to a root hint or named pipe than can be picked up by another program? * Virtual screens/workspaces Partially done. Still needed: - A window might be on one, *several* or all virtual screens. We already have a way of fixing a window on all screens (MODKEY-f), but we need a way of saying "stick on this workspace". Perhaps something like MODKEY-a , where is 1--9 for virtual screens. Better ways? Note that this seems to be mildly incompatible with the EWMH _NET_WM_DESKTOP hint we're currently using: We will only be able to save one of the desktops used. * Hide windows Instead of iconifying, hide them à la 9wm. Even if we use a key to hide them, this probably means we have to have a menu to get them back. Perhaps use an external program somehow? Needs to talk to mcwm anyway. Unix socket? * Menu We might need a menu for hidden windows (see above). Since I'm probably implementing menu windows anyway, perhaps I should add a menu with basic window functions, like 9wm and twm. This way, one might use the window manager without keyboard, if necessary. Not much work if I have to do the menu anyway... But also chords? * Chords Maybe have mouse button chords to do move and resize? Configurable if on or off? Idea from Christian Neukirchen. * Snap to border and screen edge, which favours the edge. * Gaps on borders for docks, status windows et cetera. Keep space reserved for Conky, dzen2 et cetera. Also respect EWMH hints _NET_WM_TYPE_DOCK and _NET_WM_TYPE_DESKTOP. * Flag to disable that mcwm won't allow windows to move off the screen? * Use xcb-event's event handlers? * Configurable keys. * Configuration file. * Handle Urgency hint Some windows might need attention and marks this with an urgency hint (for instance, urxvt can generate such a hint when receiving the BEL character). Do we want to handle it? How do we tell the user? Can this be done with some stand-alone program instead? * Support sensible portions of ICCCM and EWMH EWMH hints that tells applications about active workspace and focused window... And a few others. * Code cleaning - Obivous cleanup: The event switch is way too big. - The states are known everywhere. A tight state machine would be nicer. - Dispatch table for key bindings instead of keysym->enum->case? - Use bitfields instead of extra lists for virtual screens? * Resize behaviour Better resize behaviour. We want not to accidentally change the vertical size when changing horizontal size and vice versa.