GitLab

2000-11-03  Havoc Pennington  <hp@redhat.com>

        Get widgets working on the btree/buffer side. Display of them
	still doesn't work.

	* gtk/gtktextview.c: start implementing child widget stuff

	* gtk/gtktextiter.c (gtk_text_iter_get_child_anchor): new function

	* gtk/gtktextlayout.c: handle embedded widgets

	* gtk/gtktextdisplay.c: handle embedded widgets

	* gtk/gtktextchild.c: Implement all the child anchor goo

	* gtk/gtktextbuffer.c (gtk_text_buffer_create_child_anchor): New
	function

	* gtk/gtktextbtree.c: Add child anchor table to the btree struct
	(insert_pixbuf_or_widget_segment): abstract out common portions of
	creating a child anchor or a pixbuf segment.
	(gtk_text_btree_create_child_anchor): new function
	(gtk_text_btree_unregister_child_anchor): new function

	* gtk/gtkmarshal.list: added VOID:OBJECT,INT,INT for the
	allocate_child signal on GtkTextLayout

	* gtk/gtktextiter.c (gtk_text_iter_get_pixbuf): fix bogus return
	values
	(gtk_text_iter_get_child_anchor): new function

	* gtk/gtktextbuffer.c (gtk_text_buffer_real_changed): Add a
	default handler for the changed signal, which calls
	gtk_text_buffer_set_modified(), instead of just always emitting
	changed then calling set_modified() manually. I guess this is
	maybe more flexible. It seems logical.

Fri Nov  3 08:58:38 2000  Tim Janik  <timj@gtk.org>

        * gtk/gtkobject.c: _g_signal_handlers_destroy() is prefixed with '_'
        now.

        * gtk/gtksignal.c: use g_signal_connect_closure_by_id() now.

Thu Nov  2 14:10:51 2000  Owen Taylor  <otaylor@redhat.com>

	* gdk/x11/gdkdnd-x11.c: Fix a bunch of functions that should have
	been marked static.

2000-11-02  Havoc Pennington  <hp@redhat.com>

	* gtk/testgtk.c (create_labels): had a test backward

	* gdk/x11/gdkgc-x11.c (gdk_gc_copy): Copy the client-side GC
	fields, so the dest GC ends up with the correct clip origin,
	etc. Fixes a bug where colored labels didn't redraw
	properly.

2000-11-02  Alexander Larsson  <alexl@redhat.com>

	* gdk/linux-fb/gdkcolor-fb.c, gdk/linux-fb/gdkdnd-fb.c,
	gdk/linux-fb/gdkgc-fb.c, gdk/linux-fb/gdkpango-fb.c,
	gdk/linux-fb/gdkpixmap-fb.c, gdk/linux-fb/gdkwindow-fb.c:
	Add last argument to t_type_register_static call.

	* gdk/linux-fb/gdkdrawable-fb2.c:
	Add last argument to t_type_register_static call.
	export the get_image drawable virtual function.

	* gdk/linux-fb/gdkimage-fb.c:
	Add last argument to t_type_register_static call.
	rename gtk_image_get to _gdk_fb_get_image and
	export it for use by gdkdrawable-fb2.

	* gdk/linux-fb/gdkprivate-fb.h:
	export _gdk_fb_get_image.

	* gdk/linux-fb/gdkselection-fb.c:
	Add new api call stubs.

	* gtk/Makefile.am:
	Missing ( in linux-fb part.

	* modules/linux-fb/Makefile.am:
	Must link with gdk_pixbuf or get unresolved symbol.

2000-11-01  Havoc Pennington  <hp@pobox.com>

* gtk/gtktextbuffer.c (gtk_text_buffer_class_init): Use
GTK_TYPE_TEXT_ITER and BOXED marshallers

* gtk/gtktexttagtable.c (gtk_text_tag_table_class_init): ditto

* gtk/gtktexttag.c (gtk_text_tag_class_init): ditto

* gtk/gtk-boxed.defs: add GtkTextIter

* gtk/gtkmarshal.list: Add marshallers for text widget that
properly use OBJECT and BOXED

2000-11-01  Havoc Pennington  <hp@redhat.com>

	* gtk/testgtk.c (create_labels): Test set_markup_with_accel()

	* gtk/gtklabel.c (set_markup): Convert the unichar to a GDK keyval
	before returning it.

2000-11-01  Havoc Pennington  <hp@redhat.com>

	* gdk/gdkpixbuf-render.c (gdk_pixbuf_render_to_drawable_alpha):
	Clip the rectangle to be rendered to the drawable, since
	get_from_drawable() requires a region within the drawable

2000-11-01  Havoc Pennington  <hp@redhat.com>

	* gtk/gtkimage.c (gtk_image_expose): Um, don't randomly subtract 4
	from coordinates.

Wed Oct 25 20:10:57 2000  Owen Taylor  <otaylor@redhat.com>

	* gdk/gdkevents.c (gdk_event_copy): Only ref the window if
	it isn't NULL.

2000-11-01  Havoc Pennington  <hp@pobox.com>

* gdk-pixbuf.c (gdk_pixbuf_new_subpixbuf): New function to create
a pixbuf pointing to a subregion of another pixbuf.

2000-11-01  Havoc Pennington  <hp@pobox.com>

* gtk/gtklabel.c (gtk_label_set_attributes): Set a PangoAttrList
on a label
(gtk_label_finalize): unref the attr list if any.

* gtk/testgtk.c (create_get_image): close test on second click
(make_message_dialog): close dialog if it exists

* gdk/gdkpango.c (gdk_draw_layout): Handle rise attribute

* gdk-2.0.pc.in (Requires): Make it require gdk-pixbuf-2.0 not
gdk-pixbuf

* gtk/gtklabel.c (gtk_label_set_markup): new function to set label
from Pango markup format
(gtk_label_set_markup_with_accel): ditto but with accelerator
parsing

* gtk/gtkimage.c (gtk_image_expose): reformatting.

* gdk/gdkpixbuf-drawable.c (gdk_pixbuf_get_from_drawable): Hack to
reflect current state of GDK - use gdk_drawable_get_colormap, etc.
Check GDK_IS_WINDOW() not !GDK_IS_PIXMAP() to decide whether to
call gdk_window_get_origin().

* gdk/gdkpixbuf-render.c (gdk_pixbuf_render_to_drawable_alpha):
implement GDK_PIXBUF_ALPHA_FULL

Wed Nov  1 03:43:42 2000  Tim Janik  <timj@gtk.org>

        * gtk/gtksignal.c (gtk_signal_compat_matched): make use of
        g_signal_handlers_block_matched, g_signal_handlers_unblock_matched and
        g_signal_handlers_disconnect_matched to block/unblock and disconnect
        multiple handlers respectively, instead of only treating the
        first handler found (bug reported by owen).

Mon Oct 30 19:23:50 2000  Jonathan Blandford  <jrb@redhat.com>

	* gtk/gtktreeview.c (gtk_tree_view_expand_row): New function.

	* gtk/gtktreeview.c (gtk_tree_view_collapse_row): New function.

	* gtk/gtktreeview.c, gtk/gtktreemodel.c, gtk/gtktreeselection.c
	gtk/gtktreestore.h, gtk/gtktreeview.h, gtk/gtktreeviewcolumn.c,
	gtk/gtktreeviewcolumn.h: Fixed inline docs.

	* docs/reference/gtk/tree_widget.sgml: Proto intro to the tree
	widget.  Needs writing someday.

	* docs/reference/gtk/tmpl/gtktree*sgml: Added the tree widget to
	the docs.

2000-10-30  Havoc Pennington  <hp@redhat.com>

	* gtk/gtktextview.h, gtk/gtktextview.c:
	Rename delete_at_cursor delete_from_cursor, and
	move to move_cursor.

2000-10-30  Havoc Pennington  <hp@redhat.com>

	* gtktextbtree.c, gtktextbtree.h, gtktextbuffer.c,
	gtktextbuffer.h, gtktextchild.c, gtktextchild.h,
	gtktextchildprivate.h, gtktextdisplay.c, gtktextdisplay.h,
	gtktextiter.c, gtktextiter.h, gtktextiterprivate.h,
	gtktextlayout.c, gtktextlayout.h, gtktextmark.c, gtktextmark.h,
	gtktextmarkprivate.h, gtktextsegment.c, gtktextsegment.h,
	gtktexttag.c, gtktexttag.h, gtktexttagprivate.h,
	gtktexttagtable.c, gtktexttagtable.h, gtktexttypes.c,
	gtktexttypes.h, gtktextview.c, gtktextview.h:

	Massive reindentation and reformatting. Arg alignment and comments
	still need fixing in various places.

2000-10-30  Havoc Pennington  <hp@redhat.com>

	* gtk/gtktextview.h, gtk/gtktextview.c: Change run action signals
	to be named insert_at_cursor and delete_at_cursor so C++ can
	handle them.

2000-10-28  Tor Lillqvist  <tml@iki.fi>

	* gtk/gtkrc.c (gtk_rc_append_default_module_path): Make sure not
	to get a double directory separator in case the home directory
	ends with one. This would cause trouble especially on Windows,
	where the path will be confused with the syntax for an an UNC
	(network share) path \\server\share\path.
	(gtk_rc_add_initial_default_files): Ditto.

Sat Oct 28 02:33:09 2000  Robert Brady  <robert@suse.co.uk>

	* gtk/gtkmarshal.list: add new marshaller for GtkTextView.

	* gtk/gtktextview.c: Use correct marshaller.

Fri Oct 27 17:56:26 2000  Jonathan Blandford  <jrb@redhat.com>

	* gtk/gtkliststore.c: Not sure how this ever worked.  Fixed it to
	use new iter stuff, and added all the appropriate signals et al.

	* gtk/gtktreedatalist.h: s/gtk/_gtk to further reinforce that this
	code is internal to gtk only.

2000-10-28  Tor Lillqvist  <tml@iki.fi>

	* gtk/gtk.def
	* gtk/makefile.{mingw,msc}.in: Adjust to Tim's changes. Use
	glib-genmarshal etc.

2000-10-28  Tor Lillqvist  <tml@iki.fi>

	* gtk/makefile.{mingw,msc}.in: Adjust to Tim's changes. Use
	glib-genmarshal etc.

Fri Oct 27 20:15:00 2000  Tim Janik  <timj@gtk.org>

        * gtk/Makefile.am: added dependancies for testtextbuffer which broke
        parallel builds.

Fri Oct 27 18:51:44 2000  Tim Janik  <timj@gtk.org>

        * gtk/gtksignal.[hc]: adapt to GSignal signature changes.

2000-10-26  Havoc Pennington  <hp@redhat.com>

        * gtk/gtktextiter.c (my_strrstr): fix constness warning

        Fixed output of:
        nm -g -l .libs/libgtk-x11-1.3.so | grep ' T ' | grep -v gtk_ | grep -v GTK_

        * gtk/gtktextsegment.c (char_segment_new): rename
        (_char_segment_new_from_two_strings): rename
        (_toggle_segment_new): rename

        * gtk/gtkclipboard.c (ensure_clipboard_widget): make this static
        (make_clipboard_widget): make it static
        (text_get_func): static
        (text_clear_func): static

        * gtk/gtkcolorsel.c (adjustment_changed): make this function
        static
        (hex_changed): make it static
        (hsv_changed): make it static
        (opacity_entry_changed): make it static

        * gtk/gtktextbtree.c (change_node_toggle_count): rename with _gtk
        (toggle_segment_check_func): rename with _gtk

        * gtk/gtktextchild.c (_pixbuf_segment_new): rename with _gtk
        (_widget_segment_new): rename
        (_widget_segment_add): rename
        (_widget_segment_remove): rename
        (_widget_segment_ref): rename
        (_widget_segment_unref): rename

        * gtk/gtktextmark.c (_mark_segment_new): Rename
        _gtk_mark_segment_new

2000-10-26  Havoc Pennington  <hp@redhat.com>

	* gtk/gtkmarshal.list: add new marshallers used by the text widget

	* gtk/gtktextbuffer.c (gtk_text_buffer_class_init): fix marshaller
	types

	* gtk/gtktexttagtable.c (gtk_text_tag_table_class_init): fix
	marshaller types

	* gtk/gtktextlayout.h, gtk/gtktextlayout.c, gtk/gtktextdisplay.h,
	gtk/gtktextdisplay.c: We need to preserve Tk copyrights and
	license on these files.

	* gtk/gtktextiter.c (gtk_text_iter_backward_search): Make this
	work.

	* gtk/gtktextbtree.c (gtk_text_btree_new): init stamps to
	runtime random number different for each tree, instead of
	a constant I made up.

Thu Oct 26 07:36:16 2000  Tim Janik  <timj@gtk.org>

        * gtk/Makefile.am: buncha cleanups, kludged build sources
        dependencies with a Makefile: oldest-source-stamp dependancy
        for the moment. automake is really fighting us here, with
        intentionally not allowing user-defined rules or dependancies
        prior to it's -include $(DEP_FILES) statement. sucks rocks!

        * gtk/Makefile.am: added gtktextlayout.h and gtktextdisplay.h
        to public headers. tehy have a comment stating that they are
        "semi-public" aren't included by gtk.h and if enums popup there,
        they can be /*< skip >*/-ed.

        * gtk/gtktextlayout.h:
        * gtk/gtktextdisplay.h: indentation fixes, comment fixes,
        added LGPL... <SIGH>

Wed Oct 25 20:40:25 2000  Jonathan Blandford  <jrb@redhat.com>

	* gtk/gtktreemodel.h: Got rid of GtkTreeNode, and changed it to
	GtkTreeIter.  Added iterators everywhere.

	* gtk/gtktreeviewcolumn.c: Changed to use the iterators.

	* gtk/gtktreeviewselection.c: Changed to use the iterators.

	* gtk/gtktreestore.c: Changed to use the iterators.

	* gtk/gtkliststore.c: Commented out the code.  Will convert to
	iterators tomorrow.

	* gtk/gtkmodelsimple.c: Commented out the code.  Will convert to
	iterators tomorrow.

	* gtk/treestoretest.c: Changed to use iterators.

	* demos/testgtk/main.c: Moved to use the new iterator system.

Thu Oct 26 02:10:32 2000  Tim Janik  <timj@gtk.org>

        * gtk/Makefile.am: invoke maketypes.awk with LC_ALL=C to avoid
        localization misbehaviour.

        * gtk/gtk-boxed.defs: nuked "boxed" type GtkCTreeNode, since it's
        a useless type without copy/free.

        * gtk/gtkctree.c (gtk_ctree_class_init): use GTK_TYPE_POINTER instead
        of GTK_TYPE_CTREE_NODE.

Thu Oct 26 00:51:52 2000  Tim Janik  <timj@gtk.org>

        * gtk/gtksignal.c:
        (gtk_signal_emitv): use g_value_set_static_{string|boxed} where
        applicable.

        * gtk/gtk-boxed.defs (GtkSelectionData): specify
        gtk_selection_data_copy and gtk_selection_data_free
        do GtkSelectionData work as params in signals.

Wed Oct 25 20:47:41 2000  Tim Janik  <timj@gtk.org>

        * gtk/gdk-pixbuf-loader.c (gdk_pixbuf_loader_class_init): urg, removed
        implementation of gtk_marshal_VOID__INT_INT_INT_INT. if people do that,
        couldn't they at least give it a non-standard name?

        * gtk/gtktextlayout.c: arg! yet another implementation of
        gtk_marshal_VOID__INT_INT_INT_INT(), is this a conspiracy?

        * gtk/gtktextbuffer.c: gotcha! captured a vagabonding
        gtk_marshal_VOID__INT_POINTER_INT() implementation, braught it back
        home. now i know this _is_ a conspiracy.

        * gtk/gtkwidget.c (gtk_widget_class_init): marshaller fixups for
        ::state-changed.

        * gtk/gtkaccelgroup.c (gtk_accel_group_create_remove):
        (gtk_accel_group_create_add): marshaller signature fixups.

        * gtk/gtklistitem.c (gtk_list_item_class_init): signal creation fixups,
        pass in GTK_TYPE_SCROLL_TYPE instead of GTK_TYPE_ENUM.

        * gtk/gtkobject.[hc]: removed GTK_CONNECTED flag, it's not valid
        anymore.

Tue Oct 24 23:59:21 2000  Tim Janik  <timj@gtk.org>

        * docs/reference/Makefile.am: disabled SUBDIRS for the moment, since
        due to the signal system changes, it wouldn't build currently. to
        be fixed soon.

        * docs/Changes-2.0.txt: GtkSignal/GSignal updates.

        * gtk/gtkwidget.c: ::direction_changed takes an enum as argument,
        so it needs gtk_marshal_VOID__ENUM() instead of
        gtk_marshal_NONE__UINT().

        * gdk/gdk*.c: adapted type registration functions.

        * gtk/gtkbindings.c:
        * gtk/gtkaccelgroup.c: operate on GSignalQuery, GtkSignalQuery is
        gone.

        * gtk/gtkenums.h: define GtkSignalRunType in terms of GSignalType.

        * gtk/gtkobject.c:
        (gtk_object_destroy):
        (gtk_object_shutdown): fixed recursion guards. basically we have to
        catch the case where any of GObject.shutdown() or gtk_object_destroy()
        is called during ::destroy, and avoid recursion there.

        * gtk/gtktypeutils.c:
        * gtk/maketypes.awk: awk-script hackup to provide gtk_type_init() with
        boxed_copy/boxed_free. this needs a more general solution based on a
        publically installed code-generator utility.

        * gtk/gtktypeutils.[hc]: compat aliased GTK_TYPE_BOXED to G_TYPE_BOXED,
        glib's gobject has support for that now.
        define GtkSignalMarshaller in terms of GSignalCMarshaller.


Mon Oct 23 09:36:42 2000  Tim Janik  <timj@gtk.org>

        * gtk/gtksignal.[hc]:
        * gtk/gtkmarshal.[hc]:
        * gtk/Makefile.am: generate marshallers with glib-genmarshal and don't
        compile gtkmarshal.c on its own anymore, just include it in gtksignal.c.
        removed #include <gtkmarshal.h>s all over the place, gtksignal.h takes
        care of that.

        * *.c: marshaller name fixups.

        * gtk/gtkmarshal.list: added a comment briefing the format.

Sun Oct 22 23:14:39 2000  Tim Janik  <timj@gtk.org>

        * gtk/gtksignal.[hc]: nuked old implementation. we mostly have
        compatibility macros here now. more specifically, most of
        the API is preserved (yes, _most_, nonwithstanding the
        following exceptions listed, the API is stil lHUGE ;)
        things that got removed completely:
        GtkSignalQuery, gtk_signal_query(), gtk_signal_n_emissions(),
        gtk_signal_n_emissions_by_name(), gtk_signal_handlers_destroy(),
        gtk_signal_set_funcs(), gtk_signal_handler_pending_by_id(),
        gtk_signal_add_emission_hook(), gtk_signal_add_emission_hook_full(),
        gtk_signal_remove_emission_hook().
        non-functional functions variants:
        gtk_signal_add_emission_hook(), gtk_signal_remove_emission_hook().
        the GtkCallbackMarshal argument to gtk_signal_connect_full() is
        not supported anymore.
        (gtk_signal_compat_matched): new internal function to aid
        implementation of the compatibility macros, it  provides
        functionality to block/unblock/disconnect handlers based
        on func/data.

        * gtk/gtkenums.h: define GtkSignalRunType in terms of GSignalType,

        * *.c: adaptions to new type registration API signatures.


Fri Oct 20 15:26:33 2000  Tim Janik  <timj@gtk.org>

        * gtk/gtktypeutils.[hc]: removed G_TYPE_GTK_POINTER cludge.

2000-10-25  Robert Brady  <robert@suse.co.uk>

	* gdk/x11/gdkproperty-x11.c (gdk_property_get, gdk_property_change,
	gdk_property_delete): If window == NULL, assume the root window.
	(restoring the behaviour that was in 1.2).
--------------------------------------------------------------------

2000-10-25  Tor Lillqvist  <tml@iki.fi>

	* gdk/gdkwindow.c: Include x11/gdkx.h only when building for X11.

	* gdk/win32/gdkdrawable-win32.c
	* gdk/win32/gdkimage-win32.c
	* gdk/win32/gdkprivate-win32.h: Corresponding changes as in the
	X11 backend.

	* gdk/win32/gdkcc-win32.c
	* gdk/win32/gdkcolor-win32.c
	* gdk/win32/gdkfont-win32.c
	* gdk/win32/gdkinput-win32.c
	* gdk/win32/gdkmain-win32.c
	* gdk/win32/gdkproperty-win32.c: Silence gcc -Wall.

2000-10-24  Havoc Pennington  <hp@redhat.com>

	* gtk/gtktextview.c (gtk_text_view_scroll_to_mark_adjusted): make
	it a static function

	* gtk/gtktextbtree.c (gtk_text_btree_tag): Gee, maybe we should
	redraw text when a tag is applied to it.

	* gtk/gtktexttag.c (gtk_text_tag_affects_size)
	(gtk_text_tag_affects_nonsize_appearance): private functions to
	see if a tag requires various kinds of redraw/layout to be queued
	up.

	* gtk/gtktexttag.h (struct _GtkTextTag): Remove relief crackrock

	* gtk/testtext.c (fill_example_buffer): Put the cursor
	at the start of the buffer, so search works by default

	* gtk/gtktextiter.c (lines_match): init match_start always

	* gtk/gtktextbuffer.c (gtk_text_buffer_get_iter_at_line_index): New
	function, get iter at a line + a byte index

	* gtk/gtktextiter.c (gtk_text_iter_set_line_index): New function,
	to set byte position within a line
	(gtk_text_iter_check): remove leftover G_BREAKPOINT thing

2000-10-23  Havoc Pennington  <hp@redhat.com>

        * gtk/testtext.c: Re-enable the "find" dialog

	* gtk/testgtk.c: Add test for gdk_drawable_get_image

	* gdk/gdkwindow.c (gdk_window_begin_paint_region): Fix bug where
	the arguments to gdk_draw_drawable were in the wrong order
	(gdk_window_paint_init_bg): This function was ignoring the
	init_region, instead of clipping to it, so the entire backing
	pixmap was cleared on every begin_paint()
	(gdk_window_begin_paint_region): Hmm, the same list-walking bug
	was in here again, the loop kept using the same GtkWindowPaint
	over and over.
	(gdk_window_begin_paint_region): Fix a bug where we had two
	x_offset instead of x_offset and y_offset

	* gdk/gdkdraw.c (gdk_drawable_get_image): get composite drawable
	before we get the image.
	(gdk_draw_drawable): get the composite before we draw the drawable.
	(gdk_drawable_real_get_composite_drawable): default
	get_composite_drawable implementation that returns the drawable
	itself

	* gdk/gdkdrawable.h (struct _GdkDrawableClass ): Add
	get_composite_drawable virtual function

	* gdk/gdkwindow.c (gdk_window_begin_paint_region): Fix a cheesy
	list-walking bug

	* gdk/x11/gdkdrawable-x11.c (gdk_x11_draw_drawable): Add a hack to
	make this work if the source drawable is a GdkDrawableImplX11
	instead of a public drawable type. This is really broken; the
	problem is that GdkDrawable needs a virtual method get_xid(), but
	of course that doesn't work in practice. Enter RTTI.

	Also, improve mismatched depth message.

	* gdk/gdkpixmap.c (gdk_pixmap_get_image): Implement get_image for
	GdkPixmap

	* gdk/x11/gdkdrawable-x11.c (gdk_drawable_impl_x11_class_init):
	install _gdk_x11_get_image as our implementation of get_image

	* gdk/x11/gdkimage-x11.c (gdk_image_get): Rename to
	_gdk_x11_get_image and export for use in gdkdrawable-x11.c

	* gdk/gdkimage.c (gdk_image_get): Make this just a wrapper around
	gdk_drawable_get_image

	* gdk/gdkdraw.c (gdk_drawable_get_image): call virtual get_image

	* gdk/gdkdrawable.h (struct _GdkDrawableClass ): Virtualize
	get_image

	* gtk/gtktreestore.c (gtk_tree_store_get_node): remove weird
	trailing semicolon after for loop

2000-10-22  Tor Lillqvist  <tml@iki.fi>

	* gdk/win32/gdkgc-win32.c
	* gdk/win32/gdkimage-win32.c: Remove gcc -Wall warnings.

	Patches by Hans Breuer:

	* gdk/win32/gdkdrawable-win32.c (gdk_win32_draw_glyphs): Call
	SetBkMode() and SetTextAlign() before calling
	pango_win32_render().

	* gdk/win32/gdkgc-win32.c (gdk_win32_gc_values_to_win32values):
	Set hcliprgn to NULL after destroying it.
	(gdk_gc_set_clip_rectangle): Delete hcliprgn and set to NULL.

	* gdk/win32/gdkgeometry-win32.c
	(_gdk_windowing_window_queue_antiexpose): More effective
	implementation.

	* gdk/win32/gdkpixmap-win32.c (gdk_pixmap_new): Use drawable's
	depth as default, not the visual's.

	* gdk/win32/gdkwindow-win32.c (gdk_window_set_title): Guard
	against empty title.

	* gdk/win32/gdkprivate-win32.h: Define a helper macro,
	IMAGE_PRIVATE_DATA.

	* gdk/win32/gdkimage-win32.c (gdk_image_get_type): Use it.