* refs/heads/master:
Because cwm warps the pointer during a client move (to stay within the client), there's a window of time where an expose or enternotify event will get generated for a lower client; use a hammer and drain events after keyboard move/resize, until such a time that cwm doesn't warp the pointer. Behavior noticed by kn.
Fix regression from r1.107 (lost a return); kettenis@ pointed out the high potential for a use-after-free (true!) where kn@ ran into the regression using an app that actually exercised the XGrabPointer() failure path.
Add application section
command uses execvp(3) not execve(2)
Plug a memory leak in log_debug(); OK okan@
Handle _NET_WM_NAME changes.
* refs/heads/master: (23 commits)
Check the atom type on propertynotify before iterating.
use screen_find() for xrandr crtc changes
Find the managed screen from the parent window for client_current().
Print window id in hex; while here, remove unnecessary newline.
Similar to keypress event, fetch the screen from the event root window in the buttonpress handler; bail if we don't manage the screen. Allows us to find the current client based on the screen/event root.
extend verbose logging for key/button events
[keypress event] turns out we've been checking the wrong window for a matching client thus always falling back to client_current(); while the current client is problaby right in most cases, use event's subwindow (not window) to find the client. Bail early if this event came to us from a screen we don't manage. This is result of us grabing all keybindings off the root window instead of selectively.
add parans for readibility
Teach client_current() to use a screen to find the current client instead of iterating over all (fallback if no screen provided for now). Initially convert trivial uses of client_current().
check cc->gc directly
zip extra lines
gc clientq inside groups, instead use the better maintained one per-screen
shuffle deck chairs: rename group actions to match intent for clarity
same thing as screen_find()
Separate out the menu window from the client resize/move geom window; in each case, create and destroy on-demand. Isolate more menu specific code.
fix a few misplaced (and misnamed) ewmh root window functions
_NET_WORKAREA needs ngroups, so screen_update_geometry() needs to come after conf_group().
simplify xftcolor config
Tie group number and name together during config.
Move the group index (desktop number) check to the only 2 callers that require checking due to ewmh.
...
client thus always falling back to client_current(); while the current client
is problaby right in most cases, use event's subwindow (not window) to find the
client. Bail early if this event came to us from a screen we don't manage.
This is result of us grabing all keybindings off the root window instead of
selectively.
* refs/heads/master:
Add 'group-close-[n]' action to close all windows within specified group.
simplify screen 'area' usage for initial client placement
restore order from before r1.248 (vtile/vtile containment changes).
Rename internal functions to delinate between client remove, delete and xproto delete; 'window-close' is now the proper action, but 'window-delete' as an alias will remain until more interesting changes require breaking configs.
Limit vtile/htile actions to clients fully within the screen of master client.
fix missing includes
* refs/heads/master:
Stop asking for events (NoEventMask) from menu window once done with the menu (we don't destroy it, only unmap).
Allow 'transientfor' clients to inherit group and bwidth either during init or via property notify events. Previously only the flags were set but nothing was in the path to apply said flags and/or bwidth. Required slight of re-orgnaization of client_init.
merge from base, from sashan@:
Use the original client border width to adjust initial placement of clients containing {P,US}Position requests where they are explicitly set to 'ignore' in cwmrc(5); clients are unaware that their border will be altered (removed in this case) when calcuating position and thus end up a factor of their original border width off once mapped by cwm(1). cwm(1) will essentially shift the client to the edge if the original request's position and border match.
via property notify events. Previously only the flags were set but nothing was
in the path to apply said flags and/or bwidth. Required slight of re-orgnaization
of client_init.
* refs/heads/master: (28 commits)
Use screen's saved view instead of re-querying the server.
Slightly expand and expose verbose debugging.
add debugging for x events
Add a simple debug logging mechanism.
Simplification; use asprintf where appropriate now.
Use func attributes where appropriate.
Fix wins comparison declaration since it's unsigned from XQueryTree().
Generate name_to_func[] in a clean and readable fashion.
Shrink tier[] by one after removing matchname in r1.55.
If the requested group number is invalid, bail but don't kill cwm.
Quick fix: exit after a failed execvp in u_spawn instead; previously we did in u_exec, but the introduction of re-exec'ing the previous invocation of cwm if 'exec_wm' failed missed the 'exec' failing path. Will likely split out as a proper fix.
Only exec the fallback when in CWM_EXEC_WM state.
Typo, from Julien Steinhauser.
Convert menu-exec-wm from an abritrary exec menu, into a config-based menu from which one may configure (wm <name> <path_and_args>) (and choose) specific window managers to replace the running one. 'wm cwm cwm' is included by default.
As done for buttonrelease, work specific un-cycling and un-highlighting actions into the keyrelease event, only performing what's actually needed for each; should result in much fewer events against keyreleases. No intended behaviour change.
Merge group_toggle_membership_leave into the buttonrelease event and only do border work for a group/ungroup action.
add helper function client_show to bring together like actions for unhide/raise
Add support for re-exec'ing with SIGHUP; equivalent to the already built-in 'restart' function.
Use poll and XNextEvent to replace XNextEvent blocking inside the x11 event handler.
zap stray that snuck in
...
into the keyrelease event, only performing what's actually needed for each;
should result in much fewer events against keyreleases. No intended behaviour
change.
Additionally, like we do for group membership, grab the keyboard only when
required for cycling.
* refs/heads/master:
drop obsolete comment
Alter callbacks to take a struct instead of a growing number of arguments; greatly simplifies upcoming work.
Ensure clients stay within the viewable bounds on placement, even with empty borders; based on a patch from Vadim Vygonets.
Clean up, unify and accurately calculate edge distance with client move/resize actions, so as to not lose windows off the edge.
Switch bwidth type; unfortunately X11 is inconsistent.
* refs/heads/master: (34 commits)
Make it clear these are flags.
Remove duplicate check that strsubmatch() already does; while here, fix a comment.
Sprinkle __func__ in appropriate error messages.
Get rid of 'matchname'; it's too surprising to have the menu change during client search as different potential str matches are cycled through. If there's interest, the only string that doesn't exist in the listing is the window's class - that can be added of course, but it makes the line too long imho.
clean up search_match_client(); no behaviour change
Refactor callbacks to take a void * so as to not try and generalize into client_ctx in keypress and buttonpress event handlers; pass appropriate *ctx's based on context.
remove another unused proto
Rename 2 kbfunc to match closer to what they do
Add an argument to the callbacks to pass the xevent context, button or key press. This allows to remove a few hacks to duplicate functions only for behaviour changes; now differing behaviours are pushed down to the callback. Also will allow for previously unavailable actions to be bind-able down the road.
Check the ptr bounds in the new client during cycling, since not all actions do ptrsave, such as restoring client geometry; adapted from a diff by Vadim Vygonets.
More accurate to say 'toggle', rather than 'select', for group[n]/nogroup.
Add CM-a for 'nogroup' (CM-0 stays for now); update manpage to reflect.
Stash wmname into conf.
When removing xrandr regions, ensure clients are within the bounds of the screen; adapted from an ancient diff from Sviatoslav Chagaev. Things in this area will likely change, but put this in so it works now and serves as a reminder.
Calculate client nameqlen in client_setname(), the only place it's needed/used.
Turn CALMWM_NGROUPS define into variable, ngroups.
Start simplifying menu code; and in turn, remove a cursor no longer needed.
Defaults are split between defines and conf_init(); normalize these, as well as give 'sticky' groups its own variable.
For both kb and mouse move, it is possible to grab a client and move it completely off the screen/region; instead, if the pointer is outside of the client bounds, warp the pointer to the closest edge before moving.
client_ptrwarp should not deal with unhiding or raising clients (non ptr requests); most callers do this already - deal with the few that do not. client_ptrwarp becomes a simple wrapper (setpos) but it will be expanded.
...
client_ctx in keypress and buttonpress event handlers; pass appropriate *ctx's
based on context.
While here, limit some globals, replace defines with appropriate variables and
fix some naming.
key press. This allows to remove a few hacks to duplicate functions only
for behaviour changes; now differing behaviours are pushed down to the
callback. Also will allow for previously unavailable actions to be bind-able
down the road.
the screen; adapted from an ancient diff from Sviatoslav Chagaev. Things
in this area will likely change, but put this in so it works now and
serves as a reminder.
* refs/heads/master:
During init, query screen for _NET_ACTIVE_WINDOW and set that client as active; while we already look at what's under the pointer, use this information first, then look under the pointer (saving that round-trip). This restores the active state to a client after restart even if the pointer is not above it (and of course the pointer is not above another client).
active; while we already look at what's under the pointer, use this
information first, then look under the pointer (saving that round-trip).
This restores the active state to a client after restart even if the
pointer is not above it (and of course the pointer is not above another
client).
* refs/heads/master:
Some clients fail to setup hints at all, so initalize for them; fallout from r1.218 switching to malloc - clearly missed this case.
Fix-up a few simple uses of client_current(): check CLIENT_ACTIVE flag instead of relying on curcc.
init label
Limit mouse resize to hints within the client; matches kbd resize behaviour.
Switch to just malloc since we need initialize most everything anyway.
change 'sticky' to 'stick' to toggle client stickiness (seems the default binding worked for everyone for a long time!); conflict with group sticky found by Ali Farzanrad - thanks!
Simplify group_holds_only_hidden(); from Vadim Vygonets.
Simplify toggling flags; from Vadim Vygonets.
Do not draw borders on ignored clients when returning from fullscreen; from Vadim Vygonets.
Remove redundant minimum client size adjustment (minw and minh are always positive since r1.214); from Vadim Vygonets.
* refs/heads/master:
more client vs screen context differences
If a client sets hints, honor them for kb resize requests, just like we do for mouse based resize requests.
Move kb pointer movement out of the kbfunc_client_moveresize since it's got nothing to do with clients, thus doing flags work causes lots of waste and almost useless jumpy pointer movements; while here, split out move and resize since they share almost no code, just like mouse client move/resize; factor out amount and factor. Still wonder why this is here, but it works now.
pledge "stdio rpath proc exec" cwm before main event loop, after init/setup - mostly for menu building.
Partial revert of replacing screen_area() with region_find(); until a fix for a regression is found; this bug has been around for a long time it seems, but this change exposed it. Likely need to track clients in to and out of regions.
Use position on root to figure out region.
Start cleaning up name vs function differences; replace magic numbers.
Clean up unused defines.
Extend region to include both view and work areas; switch to region_find() which no longer needs to recalculate gap each time a client (or menu) is created or altered. If no RandR, fall back to display dimensions while building regions instead of during execution.
* refs/heads/master: (21 commits)
Lost fix from r1.112; add comment.
Mechanical change: group->gc
Add consistent checks against NULL.
Move client cycle grab/ungrab into a more relevant place; while here, update comments about why we need to grab/ungrab the keyboard.
Re-add lost chunk in group_cycle from r1.113.
Further simplify _NET_WM_DESKTOP handling using new group_assign().
oops; remove left over debug print
Allowing sending a valid 'nogroup' (0) group_ctx to group_assign() (since we init all groups), though assigning the client's group to NULL for 'sticky'; use this simplification in a few places (others to follow).
Split out sticky mode checks and the restoring of a client's group and _NET_WM_DESKTOP from the config-based auto-grouping; no (intentional) behavior changes. Needed for further work in cleaning up this area.
Implement _NET_CLIENT_LIST_STACKING (from Thomas Admin), but bottom-to-top order, as per spec (notified Thomas as well).
Don't allow freeze operations on fullscreen (consistent with what fullscreen does).
Sort _NET_WM_STATE Atoms like the spec.
Move CLIENT_STICKY logic from client hide/unhide to group hide/unhide; rationale being that clients should be able to hide/unhide independently of group switching.
Add Xkb modifier to ignore mask; from Alexander Polakov.
Fix whitespace.
Add client freeze extension to _NET_WM_STATE Atom, allowing flag to persist. As usual with new Atoms, requires X restart.
_NET_WM_STATE_STICKY implies only sticky at the group/desktop level, not position and size; based on discussion with a few.
Instead of special casing the 'term' and 'lock' commands, go back to keeping them hidden; showing them has apparently caused confusion/angst.
Leave command list order from .cwmrc alone; remove sort.
Bring group and client cycle closer together.
...
client_find valid resources as needed, relieving the need for
screen_find to ungracefully handle invalid root windows. Removes a long
standing XXX. Should theoretically allow XClientMessage handling on
more than one X screen. Alter callers of screen_find to handle
failures.
