2004-11-15 Matthias Clasen <mclasen@redhat.com>
* gtk/gtkselection.c: Reduce the clipboard timeout from 300
seconds to 30 seconds. (#155428, Billy Biggs)
Mon Nov 15 19:02:07 2004 Manish Singh <yosh@gimp.org>
* gdk/abicheck.sh
* gtk/abicheck.sh: don't hardcode lengths for cut, instead split on
the third field.
Mon Nov 15 18:58:22 2004 Manish Singh <yosh@gimp.org>
* gdk-pixbuf-private.h: _gdk_pixbuf_lock now returns a gboolean.
* abicheck.sh: don't hardcode lengths for cut, instead split on
the third field.
2004-11-15 Matthias Clasen <mclasen@redhat.com>
* gdk-pixbuf-animation.c:
* gdk-pixbuf-loader.c:
* gdk-pixbuf-io.c: Be extra careful when locking and handle
the case that the module may initialize the thread system.
* gdk-pixbuf-io.[hc] (_gdk_pixbuf_lock): Return whether the
lock was actually taken.
Mon Nov 15 14:25:21 2004 Jonathan Blandford <jrb@redhat.com>
* gtk/gtktreeviewcolumn.c (_gtk_tree_view_column_cell_set_dirty):
Patch from Gavin Romig-Koch <gavin@redhat.com> to no longer reset
the resized_width when marking a column dirty.
2004-11-12 Matthias Clasen <mclasen@redhat.com>
* gdk-pixbuf-io.c (gdk_pixbuf_io_init): Don't use
_gdk_pixbuf_load_module_unlocked() in the !USE_GMODULE case.
2004-11-12 Matthias Clasen <mclasen@redhat.com>
Make GDK+ compile with X11R5 (#148032)
* gdk/x11/xsettings-client.c (xsettings_client_new):
* gdk/x11/gdkproperty-x11.c (_gdk_x11_precache_atoms): Only
use XInternAtoms if it is available.
* gdk/x11/gdkasync.c: Include XIproto.h if necessary.
* configure.in: Check for XInternAtoms and X headers needed for
xReply.
2004-11-12 Matthias Clasen <mclasen@redhat.com>
Changes to make gdk-pixbuf threadsafe (#157310, #157306,
Colin Walters):
* gdk-pixbuf-io.h (enum GdkPixbufFormatFlags): Add
GDK_PIXBUF_FORMAT_THREADSAFE to indicate that an image loader
is threadsafe.
* gdk-pixbuf-io.c (get_file_formats, _gdk_pixbuf_load_module):
Use a lock to make initialization of global data structures
threadsafe.
* gdk-pixbuf-private.h:
* gdk-pixbuf-io.c (_gdk_pixbuf_lock, _gdk_pixbuf_unlock):
Auxiliary functions which use another lock to protect
threadunsafe image loaders.
* gdk-pixbuf-io.c (gdk_pixbuf_real_save):
(save_to_callback_with_tmp_file):
(gdk_pixbuf_real_save_to_callback):
(gdk_pixbuf_new_from_xpm_data):
(_gdk_pixbuf_generic_image_load):
* gdk-pixbuf-animation.c (gdk_pixbuf_animation_new_from_file):
* gdk-pixbuf-loader.c (gdk_pixbuf_loader_load_module):
(gdk_pixbuf_loader_close):
(gdk_pixbuf_loader_finalize):
Use _gdk_pixbuf_lock() and _gdk_pixbuf_unlock().
* io-ani.c, io-bmp.c, io-gif.c, io-ico.c:
* io-jpeg.c, io-pcx.c, io-png.c, io-pnm.c:
* io-ras.c, io-tga.c, io-wbmp.c, io-xbm.c:
* io-xpm.c: Mark as threadsafe.
* io-tiff.c: Remove pointless locking, mark as
threadunsafe.
2004-11-11 Matthias Clasen <mclasen@redhat.com>
* gtk/gtkfilefilter.c: Don't try to use xdgmime on
Windows where we don't build it. (#157944, J. Ali Harlow)
2004-11-11 Matthias Clasen <mclasen@redhat.com>
* gtk/gtkstatusbar.c (gtk_statusbar_size_allocate): Also
reserve space for the resize grip if the grip_window is not
realized yet. (#158010, Christian Persch)
(gtk_statusbar_button_press): Only start a resize if the
button press comes from the grip_window. (#158011, Christian Persch)
2004-11-11 Tor Lillqvist <tml@iki.fi>
Fix for #137551, by Robert Ögren:
* gdk/win32/gdkevents-win32.c (generate_focus_event): New function.
(gdk_keyboard_grab, gdk_keyboard_ungrab): Generate focus change
events.
(gdk_event_translate): Check for keyboard grabs and not pointer
grabs when handling WM_{SET,KILL}FOCUS. Use generate_focus_event().
2004-11-11 Matthias Clasen <mclasen@redhat.com>
* gtk/gtkuimanager.c (update_node): Remove two more places
where we were unconditionally setting "visible", causing
unnecessary notifications.
* gtk/gtkclipboard.c (gtk_clipboard_get_for_display):
(gtk_clipboard_get): Clarify the docs. (#156638, Torsten
Schoenfeld)
(gtk_clipboard_finalize): Warn if the clipboard is still in
its displays clipboard list.
2004-11-11 Matthias Clasen <mclasen@redhat.com>
Considerably speed up GtkUIManager, by changing the semantics
of the dirty flag. It is now propagated up the tree, and
update_node() doesn't descend into clean subtrees. (#143668,
Dave Neary, Michael Natterer, Sven Neumann, fix proposed
by Soeren Sandmann)
* gtk/gtkuimanager.c (node_prepend_ui_reference): Only prepend
a new node reference if the merge_id is different. Take a GNode,
so we can walk up the tree, adjust all callers.
(node_remove_ui_reference): Only mark a node dirty if the first
ui reference is removed. Take a GNode here as well for
consistency.
(update_node): Don't descend into clean subtrees.
(mark_node_dirty): New function to mark a node and its
predecessors dirty.
