2004-10-04 Elijah Newren <newren@math.utah.edu>
Fix a variety of focus race conditions in all focus modes, or at
least make them harder to trigger (fixes#152000)
* src/core.[ch] (meta_core_user_lower_and_unfocus): add a
timestamp parameter; pass it along to
meta_workspace_focus_default_window
* src/display.[ch] (meta_display_get_current_time_roundtrip): new
function
* src/display.c (event_callback): pass a timestamp to the
meta_workspace_activate and meta_workspace_focus_default_window
function calls
* src/frames.c (meta_frames_button_press_event): pass a timestamp
to meta_core_user_lower_and_unfocus
* src/keybindings.c (handle_activate_workspace): pass a timestamp
to meta_workspace_activate, (process_workspace_switch_grab): pass
a timestamp to meta_workspace_focus_default_window and
meta_workspace_activate, (handle_toggle_desktop): pass a timestamp
to meta_workspace_focus_default_window,
(do_handle_move_to_workspace): pass a timestamp to
meta_workspace_activate_with_focus, (handle_workspace_switch):
meta_workspace_activate
* src/screen.c (meta_screen_new): pass a timestamp to
meta_workspace_activate
* src/window.c (meta_window_free): pass a timestamp to
meta_workspace_focus_default_window, (idle_calc_showing): don't
increment the focus sentinel here, (meta_window_minimize): pass a
timestamp to meta_workspace_focus_default_window,
(meta_window_client_message), pass a timestamp to
meta_workspace_focus_default_window
* src/workspace.h (meta_workspace_activate): add timestamp
parameter, (meta_workspace_activate_with_focus): add timestamp
parameter, (meta_workspace_focus_default_window): add timestamp
parameter
* src/workspace.c (meta_workspace_focus_mru_window): make this
function take a timestamp and use it for meta_window_focus or
XSetInputFocus, (meta_workspace_activate_with_focus): make this
function take a timestamp and pass it along to meta_window_focus
and meta_workspace_focus_default_window,
(meta_workspace_activate): make this function take a timestamp and
pass it to meta_workspace_activate_with_focus),
(meta_workspace_focus_default_window): make this function take a
timestamp, warn if its 0 but try to handle that case sanely, and
pass the timestamp on to meta_window_focus or
meta_workspace_focus_mru_window or XSetInputFocus
2004-09-22 Elijah Newren <newren@math.utah.edu>
* src/keybindings.c (process_workspace_switch_grab): Focus the
default window after the user dismisses the workspace switcher
popup (fixes#123803; note that an alternate fix was made
independently by David Baron for sloppy and mouse focus users)
2004-09-17 Elijah Newren <newren@math.utah.edu>
* src/workspace.c (meta_workspace_focus_mru_window): Don't focus a
window that is minimized (fixes#147947)
2004-09-15 Elijah Newren <newren@math.utah.edu>
Patch from Ken Harris in #135786 to focus a new default window
when lowering via middle-click on the frame.
* src/core.[hc], src/frames.c: rename meta_core_user_lower to
meta_core_user_lower_and_unfocus
* src/core.c (meta_core_user_lower_and_unfocus): if in
click-to-focus mode then also move the window to the back of the
mru list and focus the new default window for the active workspace
2004-09-15 Elijah Newren <newren@math.utah.edu>
Focus the no_focus_window if no suitable window is in the mru list
(should fix the almost contrived extra issue found in #147475)
* doc/how-to-get-focus-right.txt: We no longer need to lie about
only focusing panels upon explicit request.
* src/workspace.c: (meta_workspace_focus_top_window): removed this
function--it was more code than needed and was unreliable anyway,
(meta_workspace_focus_mru_window): if a suitable window isn't in
the mru list, focus the no_focus_window instead of calling
focus_top_window.
2004-09-15 Elijah Newren <newren@math.utah.edu>
Prevent focus inconsistencies by only providing one focus method
(fixes#151990)
* src/screen.c (meta_screen_show_desktop): remove call to
meta_workspace_focus_top_window (it was merely focusing a window
that was going to be hidden anyway, and likely the one that
already had focus)
* src/workspace.[hc]: remove meta_workspace_focus_mru_window and
meta_workspace_focus_top_window from workspace.h, make them static
functions in workspace.c
2004-09-15 Elijah Newren <newren@math.utah.edu>
Remove race condition for focus window choice on window close
followed by rapid mouse movement in sloppy and mouse focus modes
(fixes#152000)
* src/window.c (meta_window_free): Don't increment the focus
sentinel for windows being freed, (idle_calc_showing): don't
increment the focus sentinel for windows being minimized
2004-09-15 Elijah Newren <newren@math.utah.edu>
Fix unwanted loss of focus to the mouse window when using keynav
(fixes#101190)
* src/display.c (event_callback): Ignore EnterNotify events with
xcrossing.mode of either NotifyGrab or NotifyUngrab
2004-09-15 Elijah Newren <newren@math.utah.edu>
Focus correct window after minimizing via the tasklist (fixes
#128200; see also #107681)
* src/display.h (struct _MetaDisplay): track the
previously_focused_window
* src/display.c (meta_display_open): initialize
previously_focused_window
* src/window.c (meta_window_free): clear the
previously_focused_window if it's being freed,
(meta_window_client_message): if we get a request to minimize the
previously_focused_window and the focus_window is a dock or the
desktop, focus the default window, (meta_window_notify_focus):
update the previously_focused_window
2004-09-07 Elijah Newren <newren@math.utah.edu>
Add a new write-up on making window focus consistent (see #152004)
* doc/how-to-get-focus-right.txt: New document
* rationales.txt: Remove references to focus bugs, instead point
to doc/how-to-get-focus-right.txt
2004-08-29 Elijah Newren <newren@math.utah.edu>
* src/window.c (window_takes_focus_on_map): Disable
focus-stealing-prevention for now; there are still some issues and
hard code freeze is tomorrow...so this will have to wait until
Gnome 2.10.
2004-08-27 Havoc Pennington <hp@redhat.com>
* src/compositor.c (meta_compositor_new): disable NameWindowPixmap
stuff always for now, it seemed kind of busted
(paint_screen): don't grab the server during repaint, adds to the
speed, though only slightly.
* src/frames.c (meta_frames_set_window_background): factor out all
the set_background stuff to one function; disable setting
background to transparent, because it breaks existing themes. We
need to add a flag in the theme XML file to say "start me with a
transparent background"
2004-08-27 Elijah Newren <newren@math.utah.edu>
Prevent an assertion failure that can occur after increasing the
number of workspaces; also fix a warning and stacking order when a
window is denied focus (fixes#150615)
* src/window.c (meta_window_stack_just_below): the position of the
window should be set equal to that of the one we want to be below,
not 1 lower than that number
* src/workspace.c (maybe_add_to_list): new function to add
on_all_workspace windows to an mru_list, (meta_workspace_new):
call maybe_add_to_list for all windows on the screen in order to
initialize the mru_list
2004-08-26 Havoc Pennington <hp@redhat.com>
* src/frame.c: delete extra copy of find_argb_visual so things
compile
* src/compositor.c (HAS_NAME_WINDOW_PIXMAP): copy the
XCompositeNameWindowPixmap() stuff from xcompmgr, though I can't
say I really know what it's supposed to help with (painting the
window border?)
2004-08-26 Havoc Pennington <hp@redhat.com>
* src/frame.c, src/theme.c: couple of cosmetic tweaks from
resolving my local patch with the bugzilla patch from the 8-19
entry below
2004-08-26 Havoc Pennington <hp@redhat.com>
* configure.in: move the have_xrender variable initialization up
in the file since it can be set as part of composite check
2004-08-19 Havoc Pennington <hp@redhat.com>
Fixes from Rich Wareham
* src/display.h (struct _MetaDisplay): add render extension check
to the display
* src/display.c: check for render
* configure.in: don't build compositing manager by default, don't
want any nasty surprises; check for render separately from
compositing manager
* src/frame.c: use an ARGB visual when available for the window
frame, so we can be all cool-ass
2004-08-25 Elijah Newren <newren@math.utah.edu>
Make dialogs that Metacity shows follow focus-stealing-prevention
conventions. (fixes one issue in #149028; see comments 47-54)
* src/delete.c (delete_ping_reply_func,
delete_ping_timeout_func): Make callback functions take a
timestamp arg, (delete_ping_timeout_func): pass the timestamp to
metacity-dialog
* src/display.c (meta_display_ping_timeout): add a timestamp to
the call to the ping_timeout_func, (meta_display_ping_window,
process_pong_message): add a timestamp to the call to the
ping_reply_func
* src/display.h (MetaWindowPingFunc typedef): add a timestamp to
this function typedef
* src/keybindings.c (error_on_command): require a timestamp and
pass the timestamp on to metacity-dialog, (handle_run_command):
pass a timestamp to error_on_command
* src/metacity-dialog.c (copy_of_gdk_x11_window_set_user_time):
new function (temporary; only for use while using gtk+-2.4),
(kill_window_question, warn_about_no_sm_support,
error_about_command): make these functions take a timestamp and
call copy_of_gdk_x11_window_set_user_time, (main): require the
first two args to the program to be "--timestamp <timestamp>"
* src/session.c (warn_about_lame_clients_and_finish_inter): pass a
timestamp of 0 to metacity-dialog to prevent focus (it's a popup
not generated by and kind of user request).
