2007-04-11 Elijah Newren <newren gmail com>
Make sure apps have correct info about their coordinates, even on
unmap. Fixes temporary hang with libXt (XtVaSetValues setting x &
y coordinates). #399552.
* src/frame.c (meta_window_destroy_frame): Add a comment noting
that the current choice causes the need for a ConfigureNotify
event
* src/window.c (meta_window_free): Send a configure notify event
due to our XReparentWindow coordinate choices on withdrawal,
(unmaximize_window_before_freeing): no need to send a configure
notify from here since it is always done in meta_window_free new,
(send_configure_notify): have to special case the coordinates used
when withdrawing the window
svn path=/trunk/; revision=3192
2007-04-11 Thomas Thurman <thomas@thurman.org.uk>
Workaround for a gdk bug which dies with BadAlloc if you try
to allocate an insanely huge rectangle for an insanely huge
window. Fixes#399529.
svn path=/trunk/; revision=3191
2007-04-11 Elijah Newren <newren gmail com>
Advertise support of Above and Below operations (assuming the
proposed EWMH additions of _NET_WM_ACTION_(ABOVE|BELOW) will be
accepted, otherwise these changes will have to be modified). Part
of #115247.
* src/display.[ch] (meta_display_open, struct MetaDisplay):
* src/screen.c (set_wm_check_hint):
Add support for _NET_WM_ACTION_ABOVE and _NET_WM_ACTION_BELOW
* src/window.c (set_allowed_actions_hints):
add active_above and action_below
svn path=/trunk/; revision=3190
2007-04-10 Elijah Newren <newren gmail com>
* src/window.c (recalc_window_features): make sure to set
_NET_WM_ALLOWED_ACTIONS so that libwnck menus don't have sensitive
but ineffective menu items. The "On Top" item is now buggy, but
due to the fact that _NET_WM_ACTION_ABOVE is not yet defined in
the EWMH. Fixes#115247.
svn path=/trunk/; revision=3189
2007-04-09 Elijah Newren <newren gmail com>
Add support for _NET_MOVERESIZE_WINDOW. #344521.
* src/display.c (handle_net_moveresize_window, event_callback):
Remove handle_net_moveresize_window() and the call to it; this
code was highly buggy, though to be fair it was never tested and
had simply been put into the code in commented out form.
* src/screen.c (set_supported_hint):
add atom_net_moveresize_window
* src/window.[ch]:
(meta_window_configure_request, meta_window_move_resize_request):
Split out the moving/resize part of the configure request and put
it into meta_window_move_resize_request
(meta_window_client_message):
check for NET_MOVERESIZE_WINDOW messages and call
meta_window_move_resize_request() with the appropriate parameters
to handle them
(meta_window_move_resize_internal):
fix some of the big comment at this function -- it wasn't quite
right, use the passed in gravity instead of
window->size_hints.win_gravity when calling adjust_for_gravity()
to make sure the correct adjustments are used.
(meta_window_get_gravity_position,
meta_window_get_geometry, meta_window_move_resize_request):
add a gravity parameter to meta_window_get_gravity_position and
have it use that gravity instead of window->size_hints.win_gravity
svn path=/trunk/; revision=3186
2007-04-08 Elijah Newren <newren gmail com>
Remove grab_start_serial, which we expect to be an ancient attempt
to workaround sloppy/mouse focus bugs that have since been
correctly fixed. May fix some race conditions. May cause nasty
bugs in sloppy/mouse focus modes. We'll find out soon enough...
See #304430.
* src/display.c (event_callback):
remove event->xany.serial >= display->grab_start_serial in several
event callback handlers
* src/display.[ch] (struct _MetaDisplay, meta_display_begin_grab_op):
* src/keybindings.c (do_choose_window, handle_workspace_switch):
* src/frames.c (meta_frames_button_press_event):
* src/core.[ch] (meta_core_begin_grab_op):
* src/window.c (meta_window_client_message, meta_window_begin_grab_op):
don't require an event_serial to be passed to
meta_display_begin_grab_op () and don't record it anymore.
* src/ui.c (struct _EventFunc, filter_func,
meta_ui_get_last_event_serial)
* src/core.h (meta_ui_get_last_event_serial):
remove meta_ui_get_last_event_serial() function (don't ask me why
it was declared in core.h) and the last_even_serial field of
_EventFunc
svn path=/trunk/; revision=3182
2007-04-08 Elijah Newren <newren gmail com>
Fix move/resize events in relation to combinations of
ConfigureRequest and WM_NORMAL_HINTS change notifications (plus a
few code cleanups). Fixes#426519.
* src/window.c (meta_window_move_resize_now):
move to the user_rect position, not some weird combination of rect
and user_rect
* src/window.c (meta_window_configure_request):
set user_rect in response to ConfigureRequest events (after the
ConfigureRequest values have been constrained) and add a big
comment explaining this change, remove unused only_resize variable
and irrelevant huge FIXME comment about it
* src/window.[ch] (meta_window_get_client_root_coords):
new function
* src/display.c (meta_display_begin_grab_op):
* src/keybindings.c (process_keyboard_move_grab):
* src/window.c (meta_window_unmaximize,
meta_window_move_resize_internal, meta_window_begin_wireframe,
update_move, meta_window_refresh_resize_popup,
warp_grab_pointer)
combine multi-step client rect root coord setting into a single
function call to meta_window_get_client_root_coords()
svn path=/trunk/; revision=3181
2007-04-07 Elijah Newren <newren gmail com>
* src/prefs.[ch] (screen_bindings array,
META_KEYBINDING_SET_SPEW_MARK definition):
* src/keybindings.c (handle_spew_mark, screen_handlers array):
Add an (unbound by default) keybinding for setting spew marks in
verbose debugging logs. I'm not sure why this was ever removed;
I've wanted it so many times.
* HACKING:
valgrind wants --log-file not --logfile.
svn path=/trunk/; revision=3174
2007-04-05 Thomas Thurman <thomas@thurman.org.uk>
* src/theme_parser.c: if theme is invalid and therefore got
freed, don't attempt to read from it. Closes#423855.
svn path=/trunk/; revision=3172
2007-04-04 Elijah Newren <newren gmail com>
* src/window.c (meta_window_move_resize_internal): send synthetic
configurenotify events also in response to MapRequest events when
the window has a frame and the application specifies PPosition or
UPosition hints. I believe they are already sent for all other
cases. Should fix#322840. Fixes the testcase at least. :)
svn path=/trunk/; revision=3156
2004-04-04 Elijah Newren <newren gmail com>
Fix lots of little issues with min/max constraints and size
increment constraints. Fixes#329152, #418395, and possibly
others.
* src/window-props.c (meta_set_normal_hints):
Do more checking to make sure application specified constraints
are self-consistent, modifying the size_hints as necessary to
achieve self-consistency.
* src/constraints.c (setup_constraint_info): remove ugly
copy-pasto, (constrain_size_increments): be careful that fixing
violation of the constraints doesn't cause a violation of the
minimum size constraints.
* src/window.c (ensure_size_hints_satisfied): new function,
(meta_window_unmaximize, meta_window_unmake_fullscreen): the
saved_rect may no longer be valid (as in the case of #329152) so
call ensure_size_hints_satisfied to fix it up.
* doc/how-to-get-focus-right.txt: Some minor spacing and wording
fixes completely unrelated to the rest of this commit
svn path=/trunk/; revision=3155
2007-04-03 Elijah Newren <newren gmail com>
* src/window.c (meta_window_unmaximize):
Only use saved_rect for determining the position to unmaximize to
for the previously-maximized direction(s). Fixes#355497.
svn path=/trunk/; revision=3154
2007-04-03 Elijah Newren <newren gmail com>
* src/display.c (meta_display_update_active_window_hint):
_NET_ACTIVE_WINDOW is a single xwindow id, not two.
svn path=/trunk/; revision=3152
2007-04-03 Elijah Newren <newren gmail com>
Avoid some crashes when dragging windows partially offscreen.
Possible (or at least partial) fix for #353513.
* src/edge-resistance.c (apply_edge_resistance): be more careful
about calls to find_index_of_edge_near_position() returning
possibly invalid indices. Also, add a warning comment to
find_index_of_edge_near_position().
svn path=/trunk/; revision=3147
2007-04-02 Elijah Newren <newren gmail com>
Patch from Carlo Wood to do some miscellaneous code cleanups found
while working on #358311.
* src/constraints.c (do_screen_and_xinerama_relative_constraints):
nicer way of avoiding compilation warning
* src/boxes.c (meta_rectangle_clamp_to_fit_into_region,
meta_rectangle_clip_to_region, meta_rectangle_shove_into_region):
Much cleaner way of ignoring invalid boxes in comparisons
svn path=/trunk/; revision=3145
2007-04-02 Elijah Newren <newren gmail com>
Patch from Carlo Wood to fix handling of unidirectional
maximization and partial struts. #358311.
* src/constraints.c (constrain_maximization):
determine target size for unidirectionally maximized windows by
determining how far they can be maximized without hitting
orthogonal struts. Avoids weird "empty spaces".
* src/boxes.[ch] (meta_rectangle_expand_to_avoiding_struts):
new function
2007-04-02 Elijah Newren <newren gmail com>
Make the strut lists (stored in workspaces) record both the
rectangle and the side that the strut is on. Lots of code
cleanups relating to struts.
* src/boxes.h (struct MetaStrut):
new struct for struts
* src/window.[ch] (struct MetaStruts, struct MetaWindow,
meta_window_update_struts):
overhaul to make window's struts remember their side as well as
their rectangular location, and just use a list instead of several
copies of near-identical code for left/right/top/bottom (allowing
us to nuke MetaStruts struct as well)
* src/testboxes.c (new_meta_strut, get_strut_list):
* src/workspace.c (ensure_work_areas_validated):
* src/boxes.c (meta_rectangle_get_minimal_spanning_set_for_region,
meta_rectangle_expand_to_avoiding_struts,
get_disjoint_strut_rect_list_in_region, fix_up_edges,
meta_rectangle_find_onscreen_edges,
meta_rectangle_find_nonintersected_xinerama_edges):
modify to handle struts being rectangle + side instead of just rectangle
* src/workspace.c (ensure_work_areas_validated):
simplify strut list creation considerably given MetaWindow change,
modify work_area computations to take advantage of region
computations being done (makes the code shorter as well as more
robust against pathological cases).
* src/util.[ch] (meta_free_gslist_and_elements):
new convenience function
* src/common.h (enum MetaDirection):
* src/edge-resistance.c (movement_towards_edge):
* src/boxes.c (meta_rectangle_edge_aligns,
rectangle_and_edge_intersection, split_edge):
Add more MetaDirection fields for convenience
* src/boxes.h (enum FixedDirections):
* src/constraints.c (setup_constraint_info, place_window_if_needed):
add a FIXED_DIRECTION_NONE to the FixedDirections enum to make
code more clear
svn path=/trunk/; revision=3144
2007-01-27 Bruno Boaventura <brunobol@gnome.org>
* src/theme.c (kill_window_question): Fallback to NORMAL state after
checking for the middle button. Fixes bug #419043.
Patch from Benjamin Berg <benjamin@sipsolutions.net>.
svn path=/trunk/; revision=3140
2007-03-31 Elijah Newren <newren gmail com>
Clean up event mask handling and meta_create_offscreen_window, to
prevent nasty metacity/gdk interactions causing hangs. See #354213.
* src/screen.[ch] (meta_create_offscreen_window):
* src/display.c (meta_display_open):
* src/screen.c (meta_screen_new):
Add a valuemask parameter to meta_create_offscreen_window
* src/display.c (meta_display_open):
make it explicit that we can't rely on PropertyNotify events for
the leader_window due to nasty metacity/gdk interaction
* src/session.c (warn_about_lame_clients_and_finish_interact):
remove cut-and-paste code for timestamp pinging and just call
meta_display_get_current_time_roundtrip
svn path=/trunk/; revision=3137
2007-03-30 Elijah Newren <newren gmail com>
Add support for _NET_WM_USER_TIME_WINDOW in order to cut down on
context switches.
* src/display.c (meta_display_open):
* src/display.h (struct _MetaDisplay):
* src/screen.c (set_supported_hint):
new atom
* src/display.c (meta_display_open,
meta_display_get_current_time_roundtrip):
* src/display.h (struct _MetaDisplay):
create a dedicated timestamp pinging window instead of reusing
display->leader_window
* src/display.c (event_callback):
* src/window-props.c (reload_net_wm_user_time_window):
* src/window.c (meta_window_new_with_attrs, meta_window_free,
process_property_notify):
* src/window.h (struct _MetaWindow):
monitor property notify events on _NET_WM_USER_TIME_WINDOW windows too
* src/window-props.[ch]:
new meta_window_reload_propert(y|ies)_from_xwindow() functions
* src/window-props.[ch]
(init_net_wm_user_time_window, reload_net_wm_user_time_window,
meta_display_init_window_prop_hooks):
* src/window.c (meta_window_new_with_attrs):
new hooks to handle new atom
svn path=/trunk/; revision=3134
* clutter/cogl/Makefile.am:
* clutter/cogl/cogl.h:
* clutter/cogl/gles/Makefile.am:
* clutter/cogl/gles/cogl.c:
Begin poplulating cogl GLES code.
* configure.ac:
* clutter/egl/clutter-event-egl.c:
* clutter/egl/clutter-stage-egl.c:
* clutter/egl/clutter-stage-egl.h:
* clutter/egl/clutter-backend-egl.c:
* clutter/egl/clutter-backend-egl.h:
* clutter/egl/clutter-egl.h:
Add initial EGL/X backend work mostly ported from backend branch.
Builds but untested as yet.
* clutter/glx/clutter-stage-glx.c:
Only include XFixes Header if we have have it.
* clutter/clutter-behaviour.c: (clutter_behaviour_apply):
* clutter/clutter-behaviour.h:
Add clutter_behaviour_is_applied()
* clutter/cogl/Makefile.am:
* clutter/cogl/cogl.h:
* clutter/cogl/gles/Makefile.am:
* clutter/cogl/gles/cogl.c:
Begin poplulating cogl GLES code.
* configure.ac:
* clutter/egl/clutter-event-egl.c:
* clutter/egl/clutter-stage-egl.c:
* clutter/egl/clutter-stage-egl.h:
* clutter/egl/clutter-backend-egl.c:
* clutter/egl/clutter-backend-egl.h:
* clutter/egl/clutter-egl.h:
Add initial EGL/X backend work mostly ported from backend branch.
Builds but untested as yet.
* clutter/glx/clutter-stage-glx.c:
Only include XFixes Header if we have have it.
* clutter/clutter-behaviour.c: (clutter_behaviour_apply):
* clutter/clutter-behaviour.h:
Add clutter_behaviour_is_applied()
