update.

[chise/xemacs-chise.git.1] / src / ChangeLog.GTK

2000-10-03  William M. Perry  <wmperry@aventail.com>

        * objects-gtk.c (gtk_font_instance_truename): Make sure we get the
        fully expanded version of the font.

        * device-gtk.c (convert_font): Ditto.

        * gtk-xemacs.c (convert_font): Tell __get_gtk_font_truename to not
        expand wildcards.

        * objects-gtk.c (__get_gtk_font_truename): Use the internal
        name-list in a GdkFont structure to find the truename of the
        font.  This protects us from crashing if we get a FontSet
        instead of a Font.

        (__get_gtk_font_truename): Accept new argument 'expandp' for
        whether to return the FULL font name or the wildcarded version.

2000-09-21  William M. Perry  <wmperry@aventail.com>

        * device-gtk.c (Fgtk_init): Moved calls to gtk_init or gnome_init
        to separate function, and expose it to lisp.  It is now possible
        to create GTK applications from batch mode.
        (gtk_init_device): Use the new function.

2000-09-12  William M. Perry  <wmperry@aventail.com>

        * gtk-glue.c (gdk_event_to_emacs_event): Special case double
        and triple clicks when converting events to lisp objects.
        This allows something like GtkCList to treat double-clicks
        differently in the 'select_row' signal.

2000-09-11  William M. Perry  <wmperry@aventail.com>

        * menubar-gtk.c (menu_create_menubar): Set a special name for
        GtkMenuItems directly in the menubar.

2000-09-10  William M. Perry  <wmperry@aventail.com>

        * gtk-xemacs.c (gtk_xemacs_size_request): Deal with frame
        being NULL.

        * gtk-xemacs.c (gtk_xemacs_size_allocate): Ditto.

2000-09-09  William M. Perry  <wmperry@aventail.com>

        * sound.c (init_native_sound): Enable sound for GTK devices.

        * device-gtk.c (gtk_init_device): Attempt to load a default
        gtkrc file from the data directory.  This way we can enable
        the default face font handling in gtk-faces.el but not screw
        the majority of users with a proportional font by default.

        * device-gtk.c (gtk_init_device): Attempt to load
        ~/.xemacs/gtk-options.el when GTK devices are created.  This
        allows for setting a persistent geometry without requiring GNOME.

        * gtk-xemacs.c (gtk_xemacs_style_set): Deal with NULL frame.

        * device-gtk.c (gtk_init_device): Make app_shell a GtkXEmacs
        widget so that style information is retrieved correctly.

        * menubar-gtk.c (gtk_xemacs_menubar_get_type): New subclass of
        GtkMenuBar that always requests the same width as the text
        widget.  This fixes the spurious frame resizes when there were too
        many menu items to display in the desired width.
        (create_menubar_widget): Use the new subclass.

2000-09-08  William M. Perry  <wmperry@aventail.com>

        * device-gtk.c (Fgtk_keysym_on_keyboard_p): Ported function from
        the X side of things.

        * device-gtk.c (gtk_mark_device): Make sure that we mark the
        keysym hashtable or things go boom.

2000-09-07  William M. Perry  <wmperry@aventail.com>

        * menubar-gtk.c (gtk_update_frame_menubars): Don't actually
        update the menubars if the menu is up.  This fixes the weird
        problem where if you had a menu up and 'message' was called
        (this happens a lot with the 'customize' menu), the menu would
        disappear.  This is because XEmacs is fairly lame about when
        it updates the menus - the message logging code eventually
        does a (save-excursion (set-buffer " *Message Log*") ...).
        The set-buffer caused magic variable current-menubar to be
        evaluated, which calls gtk_update_frame_menubars, which would
        nuke the menus.  Gack.

2000-09-06  William M. Perry  <wmperry@aventail.com>

        * event-gtk.c (gtk_event_to_emacs_event): Reworked how we handle
        removing the shift modifier of normal keys (a & A, etc) to be more
        like MS-windows.  This makes everything work pretty happily with
        query-replace and apropos now.

2000-09-05  William M. Perry  <wmperry@aventail.com>

        * select-gtk.c (emacs_gtk_selection_received): Signal a fake
        event here so that the event loop will wake up.  Should fix
        the strange pause seen when pasting.

        * select-gtk.c (Fgtk_get_clipboard): Signal an error if no
        selections are available.  This is more meaningful than 'insert'
        throwing an error when it gets 'nil'.

        * select-gtk.c (emacs_gtk_selection_received): Don't bother
        checking whether the data returned as the selection is a string.
        If it is not, we convert it to binary data anyway.  This fixes the
        bug where you could not paste between two separate XEmacs
        instances w/mule enabled (it sends selections as COMPOUND_TEXT).

        * device-gtk.c (Fgtk_style_info): Return the default font name as
        part of the style info.

        * menubar-gtk.c (__generic_button_callback): make sure that we
        provide a channel for our menu events.  This fixes things like
        get-dialog-box-response that rely on event-channel !nilp

2000-09-05  William M. Perry  <wmperry@aventail.com>

        * glyphs-gtk.c (__downcase): Actually return the downcased string!
         Thanks to Michael Altenhofen <Michael.Altenhofen@sap.com> for
        spotting this.

2000-09-05  William M. Perry  <wmperry@aventail.com>

        * menubar-gtk.c (gtk_popup_menu): Make sure we call
        __activate_menu correctly if the menu is dynamic.  This fixes
        popup menus with :filter effects.

2000-09-01  William M. Perry  <wmperry@aventail.com>

        * gpmevent.c, gpmevent.h: Updated to the latest XEmacs 21.2
        version of gpmevent.c.  This means that GPM support works
        with GTK now.

        * console-tty.c, console-tty.h: Removed old GPM support.

        * device-tty.c, emacs.c, event-Xt.c: Ditto.

        * event-unixoid.c, frame-tty.c: Ditto.

2000-08-30  William M. Perry  <wmperry@aventail.com>

        * gtk-xemacs.c (smash_scrollbar_specifiers): Don't bother looking
        for nextstep themed scrollbars - didn't work anyway.

        * glade.c (Fglade_xml_textdomain): deal with old versions of Glade
        that use 'textdomain' instead of 'txtdomain' in the structure.

        * menubar-gtk.c (gtk_popup_menu): use gtk_widget_show instead of
        gtk_widget_show_all so that the magic `space-saver' menu item does
        not get shown on popups.

2000-08-27  William M. Perry  <wmperry@aventail.com>

        * scrollbar-gtk.c (gtk_update_scrollbar_instance_status): Fiddle
        with scrollbar_values -> GtkAdjustment conversion and scrolling up
        with the arrows works correctly now.

        * event-gtk.c (gtk_event_to_emacs_event): Fixed the shifted key
        lossage (pc-select works now).

2000-08-11  William M. Perry  <wmperry@aventail.com>

        * scrollbar-gtk.c (scrollbar_cb): Need to make sure we look at the
        appropriate scrollbar instances from the mirror.  Was looking only
        at the vertical scrollbar instance.  Don't know if this has
        anything to do with the weird scrolling behaviour, but it is worth
        a shot.

2000-07-26  William M. Perry  <wmperry@aventail.com>

        * menubar-gtk.c (run_menubar_hook): New function that runs
        activate-menubar-hook if the menu shell it was invoked by is
        currently inactive.
        (create_menubar_widget): Hook up the button-press-event to
        run_menubar_hook to cater to broken packages.

2000-07-22  William M. Perry  <wmperry@aventail.com>

        * frame-gtk.c (gtk_popup_frame): When the window is supposed to be
        initially unmapped, we need to make sure we realize the text
        widget (to avoid lossage in redisplay assuming there is a valid
        window), but NOT show it.

2000-07-12  William M. Perry  <wmperry@aventail.com>

        * menubar-gtk.c (__kill_stupid_gtk_timer): New function to kill
        the timer on a GtkMenuItem widget.  This timer and its callback
        appear to be what was causing heavily filtered menus (like
        customize and other things under 'options') to crash.  The GTK
        code installs a timer when the user is moving the mouse around to
        scan menus.  Submenus are only popped up when this timer expires.
        But if the filters are constantly running and creating/destroying
        submenus, then you can blow up when they unconditionally check
        GTK_WIDGET_FLAGS(menu_item->submenu), when submenu is NULL.

2000-07-11  William M. Perry  <wmperry@aventail.com>

        * device-gtk.c (gtk_init_device): Can now pass the entire argv
        array that is in gtk-initial-argv-list, since gtk-init filters out
        unknown options that may cause GTK to puke for us.  This means
        that GNOME session management works now.

        * frame-gtk.c (gnome_parse_geometry): Ripped this out of the GNOME
        libraries to parse geometry settings, in case the user did not
        compile with GNOME.
        (gtk_initialize_frame_size): If gtk_initial_geometry is !NILP,
        then try to parse it and use that.  If the geometry cannot be
        parsed, fall back to using 80x30.

        * device-gtk.c (Vgtk_initial_geometry): New variable exposed to
        lisp holding the desired geometry of initial frames.

2000-07-09  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (Fgtk_call_function): Outlined possible `enhancement'
        if someone calls gtk-call-function with too few arguments.  After
        I implemented it I realized it was probably a bad idea, so I
        commented it out.  Maybe for the future...

        * menubar-gtk.c (menu_convert): Can now pass in a GtkWidget to
        reuse.  It detaches submenus, cleans up any GCPROs on it, and
        attaches a new submenu.  All done in an effort to stop the menubar
        flickering.
        (menu_create_menubar): No longer willy-nilly deletes all of the
        items in the top-level menubar.  We now check to see if the widget
        already existing at the desired position has the same label as
        this menu.  If it does, we pass it in to menu_convert.  This
        drastically reduces the flickering menu problem that most people
        have been seeing (try speedbar now).

2000-07-04  William M. Perry  <wmperry@aventail.com>

        * event-gtk.c (gtk_event_to_emacs_event): If FRAME is NULL, then
        default to the selected frame on Vdefault_gtk_device.  This will
        only happen when being called from gtk-glue.c

        * ui-gtk.c (gtk_type_to_lisp): Properly convert GTK_TYPE_GDK_EVENT
        objects to lisp events.

        * event-gtk.c (gtk_event_to_emacs_event): Made this non-static so
        that gtk-glue.c can use it.

        * gtk-glue.c (gdk_event_to_emacs_event): New function to convert a
        GDK event into something sensible for lisp.  Just uses
        gtk_event_to_emacs_event() from event-gtk.c to avoid code
        duplication.  Not perfect (the channel is set to the selected
        frame, not the GtkObject it actually happened on).

        * event-gtk.c (gtk_event_to_emacs_event): Finally fixed the weird
        selection problem where if you released the mouse button over the
        root window, XEmacs would get confused and still think the
        selection was going on.

        * ui-gtk.c (Fgtk_describe_type): New function to return the
        signals and magic properties of an object given its name or type
        identifier.

2000-07-03  William M. Perry  <wmperry@aventail.com>

        * ui-byhand.c (Fgtk_ctree_recurse): New function gtk-ctree-recurse
        that encompasses gtk_ctree_post_recursive,
        gtk_ctree_pre_recursive, gtk_ctree_post_recursive_to_depth, and
        gtk_ctree_pre_recursive_to_depth.  All hail lisp bindings with
        variable number of arguments.
        (Fgtk_ctree_recurse): Allow nil for the node.

        * ui-gtk.c (emacs_gtk_boxed_equality): New function for comparing
        two GtkBoxed objects.  This is needed because there is no way to
        store a `user_data' or equivalent on them, so we cannot resurrect
        it like a GtkObject so we always get the same Lisp_Object.  This
        allows callbacks to use `equal' on two GtkBoxed types and get sane
        results.
        (emacs_gtk_boxed_hash): New function for hashing GtkBoxed
        objects.

2000-07-01  William M. Perry  <wmperry@aventail.com>

        * glade.c: New file to implement glue code for libglade. 

2000-06-30  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (lisp_to_gtk_type): Know how to deal with
        GTK_TYPE_GDK_GC.

        * gtk-glue.c: Need to defien a GTK type for GdkGC so that we can
        import all the GDK drawing primitives.
        (face_to_gc): New function to convert a face object to a GC.

2000-06-27  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (Fgtk_import_variable_internal): Renamed to -internal.

2000-06-25  Vladimir Vukicevic <vladimir@helixcode.com>

        * frame-gtk.c (gtk_set_initial_frame_size): Added in a call to
        gtk_window_set_policy so that you can resize the window down below
        the minimum size of the menubar.

2000-06-23  William M. Perry  <wmperry@aventail.com>

        * emacs.c (Fkill_emacs): Make sure we clean up the .saves* files
        on normal exit.

2000-06-13  William M. Perry  <wmperry@aventail.com>

        * menubar-gtk.c (__activate_filtermenu): Put in some of the same
        protections for filter menus that the normal menus have for widget
        deletion and selected menu items, etc.

2000-06-12  William M. Perry  <wmperry@aventail.com>

        * scrollbar-gtk.c (gtk_create_scrollbar_instance): hide a
        scrollbar widget until we are absolutely sure we need to see it.
        Fixes the problem we were seeing with mouse-avoidance-mode causing
        severe scrollbar breakage.
        (update_one_widget_scrollbar_pointer): Move the call to
        gtk_widget_realize() here instead of in the upper code.  Isolates
        the dependency-on-having-a-window logic.

        * gtk-xemacs.c (smash_scrollbar_specifiers): When setting the
        scrollbar specifiers, need to take the x/y thickness of the border
        into account.  Horizontal scrollbar placement is much much nicer
        now.  Some themes would cause them to get positioned partially on
        top of the modeline.

2000-06-08  William M. Perry  <wmperry@aventail.com>

        * console.c (select_console_1): Duh, forgot to put in an 'else'
        statement, so Vwindow_system was getting set to Qnil all the
        time.  *sigh*.

2000-06-02  William M. Perry  <wmperry@aventail.com>

        * glyphs-gtk.c (gtk_xpm_instantiate): Do not allow pixmaps to be
        instantiated as pointers under GTK.  The pixmap and cursor
        routines under GDK do not expose enough information to let use do
        the same magic that glyphs-x.c does.  *sigh*

        * ui-byhand.c (Fgtk_toolbar_insert_item): Hand-rolled function.
        (Fgtk_toolbar_prepend_item): Ditto.
        (generic_toolbar_insert_item): Utility function to take care of
        all the common code for the Fgtk_toolbar_*_item functions.

2000-06-01  William M. Perry  <wmperry@aventail.com>

        * gtk-glue.c (face_to_style): DOH!  You should only load it if
        IMAGE_INSTANCEP, not !IMAGE_INSTANCEP you doofus.

        * ui-byhand.c (Fgtk_toolbar_append_item): Hand-rolled function.

        * ui-gtk.c (Fgtk_import_function_internal): No longer need to use
        dll_function() to get the marshallers.  They are now stored in a
        hashtable.  Lookups should be a bit quicker, and it will work on
        platforms where you cannot do a dll_open (NULL) to look at your
        own symbol table.

2000-05-31  William M. Perry  <wmperry@aventail.com>

        * select-gtk.c (emacs_gtk_selection_handle): Better MULE string
        handling.

        * gtk-xemacs.c (gtk_xemacs_realize): Make sure we set the style on
        the widget from the realize function.  Otherwise for some themes
        the color slots are not allocated yet, and we end up with icky
        looking colors for things like the modeline/toolbar.

        * select-gtk.c (Fgtk_get_clipboard): If we cannot get the
        selection from anyone, return the last selection we received.
        This make us work more like the X selection behaviour with
        cutbuffers enabled.

2000-05-30  William M. Perry  <wmperry@aventail.com>

        * ui-byhand.c: Removed definitions of gtk-clist-prepend,
        gtk-clist-append, and gtk-clist-insert.

        * ui-gtk.c (lisp_to_gtk_type): Use it.

        * gtk-glue.c (face_to_style): Routine to convert from faces to
        GtkStyle

        * menubar-gtk.c (gtk_popup_menu): Honor popup_up_p here.
        (popdown_menu_cb): and here.

2000-05-29  William M. Perry  <wmperry@aventail.com>

        * frame-gtk.c (gtk_popup_frame): Do not show the widgets if we
        were told to be initially unmapped.
        (gtk_init_frame_1): Remember whether we were told to be initially
        unmapped.  Balloon help looks a little better now.

2000-05-28  William M. Perry  <wmperry@aventail.com>

        * redisplay-gtk.c (gtk_output_string): Fixed multi-dimensional
        text run drawing.  gdk_draw_text does too much, by dividing the
        length by 2.  So we fake them out my multiplying the length by the
        dimension of the text.  This will do the right thing for
        single-dimension runs as well of course.

2000-05-26  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (get_enumeration): Utility function that does its best
        to import unknown enumeration types.

        * glyphs-gtk.c (resource_symbol_to_type): Fixed pointer
        instantiation.

        * gtk-xemacs.c (FROB_FACE): Make sure to pass the device to
        Fmake_image_instance or the initial background pixmaps would not
        show up.

2000-05-25  William M. Perry  <wmperry@aventail.com>

        * device-gtk.c (gtk_init_device): Call gnome_init if available.

        * menubar-gtk.c (create_menubar_widget): Use gnome_app_set_menus
        instead of dealing with all the handlebox/menu crap ourselves.

        * frame-gtk.c (gtk_create_widgets): Use GnomeApp if it is
        available.  Looks much sexier. :)

        * glyphs-gtk.c (gtk_resource_instantiate): New image instantiator
        gtk-resource, similar to the mswindows-resource stuff.  This
        allows you to get to the stock cursors defined in GDK.  May
        eventually allow you access to things like stock gnome pixmaps,
        not sure.

        * frame-gtk.c (gtk_set_frame_pointer): Actually handle setting the
        frame cursor.

        * redisplay-gtk.c (gdk_draw_bitmap): New function to output a
        bitmap using XCopyPlane instead of XCopyArea.
        (gtk_output_gdk_pixmap): Use it when PIXMAP_DEPTH == 0.  This
        means bitmaps look correct now.

2000-05-24  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (flags_to_list): New function to convert a GtkFlags
        argument to a list of symbols.
        (gtk_type_to_lisp): Use it for converting from a flag.

2000-05-23  William M. Perry  <wmperry@aventail.com>

        * frame-gtk.c (gtk_set_frame_position): Do not use
        gtk_window_reposition - this apparently does not exist in some
        versions of GTK 1.2

        * gtk-glue.c (xemacs_gtklist_to_list): Don't call g_list_foreach
        on a NULL list, just in case.

        * redisplay-gtk.c (gtk_text_width_single_run): Use gdk_text_width
        instead of gdk_text_measure.  Proportional fonts appear to work
        now.

        * objects-gtk.c (gtk_initialize_font_instance): Use X specific
        crap lifted from objects-x.c to figure out proportionality of a
        font, and better location of the default character.

2000-05-22  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (enum_to_symbol): Routine to convert an enum to a lisp
        symbol.  We now return this instead of an integer for GtkFlags or
        GtkEnum types.
        (lisp_to_flag): Routine to convert from a symbol or list of
        symbols to a GtkEnum or GtkFlags type.
        (lisp_to_gtk_type): Use it exclusively.  No more sending integers
        to functions.
        (import_gtk_enumeration_internal): Do not import the GTK_FOO_BAR
        flags/enums into the lisp namespace.  We now store mappings from
        symbol names (both the 'real' and nickname fields) to the actual
        internal values.  Much more lispy way of doing things.

        * menubar-gtk.c (__maybe_destroy): If we delete the menu item that
        was 'active' when the menu was cancelled, GTK gets upset because
        it tries to remove the focus rectangle from a (now) dead widget.
        This widget will eventually get killed because it will not be
        `precious' the next time the window is shown, because it is
        unselectable.

        * frame-gtk.c (delete_event_cb): Signal a fake event to make the
        event loop wake up and notice we've been destroyed.

2000-05-20  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (object_putprop): Allow `put'ing of arbitrary slots on
        gtk objects.  This will be useful for tacking lisp shit onto
        composite widgets, etc.
        (object_getprop): Ditto for `get'ing.

        * frame-gtk.c (gtk_set_initial_frame_size): Don't delay when
        changing the frame size, or creating initially unmapped frames is
        screwed up.  This showed up when trying to use the cheesy XEmacs
        file selector dialog implemented using frames.

        * ui-byhand.c: Removed a lot of functions that can now be imported
        directly by gtk-import-function using the
        Gtk(Array|List)Of(String|Object) types.

        * ui-gtk.c (type_to_marshaller_type): Deal with new array/list
        types.
        (Fgtk_call_function): Need to free array and list storage after
        calling functions.
        (lisp_to_gtk_type): Deal with the new list and array types when
        converting to GTK types.

        * gtk-glue.c: New file to deal with lists of strings or
        GtkObjects.  Defines new types GtkArrayOf, GtkListOf,
        GtkArrayOfString, GtkListOfString, and GtkListOfObject.
        (xemacs_list_to_array): Convert from a lisp list to a GtkArrayOf
        of items.
        (xemacs_list_to_list): Convert from a lisp list to a GtkListOf of
        items.

        * dialog.c (Fpopup_dialog_box): Don't crap out if the car of
        dbox_desc is not a string... this allows us to follow ben's
        system/native dialog box specification.

2000-05-19  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (lisp_to_gtk_type): Can now convert to GDK colors.  Can
        deal with color specifiers or instances.
        (lisp_to_gtk_type): Can now convert to GDK fonts.  Can deal with
        face names, actual faces, font specifiers, or font instances.

2000-05-18  William M. Perry  <wmperry@aventail.com>

        * gtk-xemacs.c (smash_scrollbar_specifiers): Function that
        attempts to set the scrollbar width/height correctly, but does not
        do a very good job.  Commented out for now.

        * redisplay-gtk.c (gtk_output_vertical_divider): Got rid of lots
        of useless code, since we ended up just calling gtk_output_shadows
        anyway.
        (gtk_output_vertical_divider): Make sure we fill the rectangle
        completely, otherwise the transparent background shows thru.

        * menubar-gtk.c: Don't nuke menus all the time... should speed up
        submenu traversal a bit, and save on the GtkMenuItem creation.

        * device-gtk.c (Fgtk_style_info): Return a list of all the
        pixmaps, not just GTK_STATE_NORMAL.

        * menubar-gtk.c (menu_descriptor_to_widget_1): Better menu labels
        with keybindings.

        * frame-gtk.c (gtk_set_frame_size): This function actually works
        now.
        (gtk_set_initial_frame_size): Better default sizing method.

        * event-gtk.c (init_event_gtk_late): Push an error trap handler so
        that XEmacs won't abort at the drop of a hat on X errors.  We
        could get X specific here and override the default GDK XError and
        XIOError handlers to be more like those in device-x.c.  Not sure
        if that is worth it or not - you would at least get information
        that an error occurred.

        * scrollbar-gtk.c (gtk_update_scrollbar_instance_status): Don't
        always call gtk_widget_set_usize because that causes the widget to
        redraw itself, which can cause some annoying flicker.
        (gtk_update_scrollbar_instance_status): Don't always move it
        either, because that can cause the GtkFixed container to get a
        resize event.
        (update_one_widget_scrollbar_pointer): Try to set the cursor 

2000-05-17  William M. Perry  <wmperry@aventail.com>

        * device-gtk.c (Fgtk_style_info): Back to taking only 1 argument -
        the device.
        (Fgtk_style_info): Now returns ALL of the information about
        colors.  Returns a list of colors instead of just the one
        associated with GTK_STATE_NORMAL.

2000-05-16  William M. Perry  <wmperry@aventail.com>

        * gtk-xemacs.c (smash_face_fallbacks): New function to reset the
        gtk-specific fallbacks on various face specifiers.  This means
        that if the user has not changed the face, when a theme or style
        is changed, the faces will automatically change as well.
        (gtk_xemacs_style_set): Call it.

        * toolbar-gtk.c (get_toolbar_gc): Swap the fg/bg of the toolbar
        face when getting the GC.  It looks better this way.

        * gtk-xemacs.c (gtk_xemacs_style_set): Override the style-set
        method so that we can nuke the pixmaps again.
        (__nuke_background_items): Moved the voodoo out into its own
        function so that it can be called from both style_set and
        realize.

        * console-gtk.h (struct gtk_frame): Removed hardcoded GCs ala X.

        * toolbar-gtk.c (get_toolbar_gc): New function that dynamically
        creates a GC from the `toolbar' face.
        (gtk_draw_blank_toolbar_button): Use it instead of the hardcoded
        GC.
        (gtk_output_toolbar_button): Ditto.
        (gtk_output_toolbar): Ditto.

        * event-gtk.c (gtk_check_for_quit_char): Quit handling actually
        works now.  Feh!

        * device-gtk.c (gtk_device_init_x_specific_cruft): New function
        that gets the socket we are listening to the X server on so that
        the SIGIO lossage works correctly for GTK devices.
        (gtk_init_device): Call it at device creation time.

2000-05-15  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (__internal_callback_marshal): We now correctly handle
        the 'data' argument.  This is an arbitrary lisp object passed to
        the callback routine as its last argument.

2000-05-14  William M. Perry  <wmperry@aventail.com>

        * event-gtk.c (gtk_event_to_emacs_event): Needed to reinstate the
        magic to NOT differentiate betwen ! and shift-!.  *sigh*

        * ui-gtk.c (lisp_to_gtk_type): Allow 'nil' for string types so
        that we can pass NULL to gtk-frame-new.

2000-05-13  William M. Perry  <wmperry@aventail.com>

        * gtk-xemacs.c (gtk_xemacs_size_request): Needed to override the
        size_request method to get frame sizing correct.  Whoo hoo!
        (gtk_xemacs_realize): Don't set the background on the GtkXEmacs
        window - this reduces the flicker even more.

        * device-gtk.c (gtk_init_device): Don't use shared memory under
        FreeBSD - it is apparently flaky as hell and causes lots of themes
        to crash and burn quite prettily.

        * gtk-xemacs.c (gtk_xemacs_realize): Added new widget method that
        makes sure to nuke the background pixmap of the XEmacs text area
        and its parent (the GtkWindow it is contained in).  This fixes the
        flashing you would see with themes/styles that define a background
        pixmap.

2000-05-12  William M. Perry  <wmperry@aventail.com>

        * menubar-gtk.c (gtk_update_frame_menubar_internal): Duh, should
        actually pay attention to menubar_will_be_visible instead of just
        calling gtk_widget_show_all in both branches. :)

2000-05-11  William M. Perry  <wmperry@aventail.com>

        * menubar-gtk.c (vars_of_menubar_gtk): New variable
        menubar-dockable-p that controls whether to use the GtkHandleBox
        or not.

        * select-gtk.c: Implemented all of the selection callbacks for GTK

        * frame-gtk.c (resize_event_cb): Force a redisplay when the frame
        is resized.

        * event-gtk.c (gtk_event_to_emacs_event): When we are doing our
        hackery to make sure we don't see button events inside the
        scrollbar, make sure we correctly forward events that did not
        happen over ANY widget, otherwise selection gets royally screwed
        and thinks you still have the mouse button down.

        * redisplay-gtk.c (gtk_output_string): Don't bother calling
        gdk_draw_text_wc - I misunderstood what XDrawString16 did -
        gdk_draw_text encapsulates this nicely for us.

2000-05-10  William M. Perry  <wmperry@aventail.com>

        * menubar-gtk.c: Changed how the menubar is created and managed.
        We no longer create and destroy it at will.  Only one GtkMenuBar
        is ever created, and the children are just added/removed from it.
        Much less flickering when switching buffers - cleaner in general.
        (create_menubar_widget): Wheee - menubars are now detachable.

        * ui-gtk.c (Fgtk_import_function_internal): Don't drop everything
        down to its fundamental type just yet - we need to know specifics
        about boxed types.
        (object_putprop): Duh, actually SET the property.
        (Fgtk_fundamental_type): New function that does the obvious.
        (Fgtk_object_type): New function that does the obvious.
        (lisp_to_gtk_type): Implement glyph handling!  GtkPixmap works!

        * ui-byhand.c (Fgtk_pixmap_get): Implemented by hand.  *sigh*

        * dialog-gtk.c: Call into lisp to do dialog boxes.

2000-05-08  William M. Perry  <wmperry@aventail.com>

        * make-src-depend (PrintPatternDeps): Make sure we generate the
        xx-gtk.* dependencies.

        * depend: Regenerated 

        * make-src-depend: Handle inclusion of .c files.

        * Makefile.in.in (extra_doc_files): Need to include ui-byhand.c
        here as well.

        * ui-gtk.c (type_to_marshaller_type): Don't abort if we don't know
        about the marshaller.  Just return Qnil and let the caller figure
        it out.
        (Fgtk_import_function_internal): Deal with new return value of
        Qnil for error signalling from type_to_marshaller_type().
        (lisp_to_gtk_type): Half-assed implementation of dealing with
        callbacks.
        (gtk_type_to_lisp): Ditto.

        * ui-byhand.c: New file that hand-codes some annoying functions in
        Gtk.  *sigh*

        * emacs-widget-accessors.c (Fgtk_adjustment_lower): New
        auto-generated file that spits out all the widget slot accessor
        functions.

        * ui-gtk.c (Fgtk_signal_connect_internal): Holy shit, signals work!
        (lisp_to_gtk_type): Allow ints to be passed in as arguments to
        functions that expect floats, otherwise you have to write stupid
        code like 9.0 instead of just passing in '9'.

2000-05-07  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (gtk_type_to_lisp): Return Qnil if we get a NULL
        pointer/object/boxed.
        (lisp_to_gtk_type): Allow Qnil to be passed in to mean a NULL
        pointer/object/boxed.

        * event-gtk.c (gtk_event_to_emacs_event): Make sure a button press
        focuses on the text widget, otherwise if someone packs a widget
        into the container from lisp, we end up not getting keyboard focus
        quite a bit.  Eek.

        * frame-gtk.c (gtk_create_widgets): Set the name of the xemacs
        shell so we can control sylte issues from .gtkrc
        (gtk_set_initial_frame_size): Set the default size of the frame
        correctly.
        (gtk_create_widgets): Expose the frame shell and container widgets
        to lisp.

        * emacs-marshals.c: Added a whole shitload of marshallers - don't
        edit this file by hand if you need to add one.  Check out
        ../lisp/ui/gtk-marshal.el for a script that auto-generates them
        for you.

2000-05-05  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (describe_gtk_arg): Debugging function to spit out
        human-readable version a GtkArg.
        (lisp_to_gtk_type): Made this function much more sane.
        (Fgtk_call_function): New function to allow calling generic
        functions from DLLs in lisp.
        (CONVERT_RETVAL): New macro (ugh) to convert a pointer that the
        return value filled in back into a valid GtkArg.
        (Fgtk_call_function): This actually works now!  Scary stuff.

2000-05-04  William M. Perry  <wmperry@aventail.com>

        * ui-gtk.c (Fgtk_import_type): New function to import a GtkType
        into the XEmacs namespace.  For starters this only fully imports
        enumerations and flags.  Importing a GtkObject or subclass results
        in getting a list back of possible signals and argument lists.

        * scrollbar-gtk.c (gtk_create_scrollbar_instance): Fixed some
        compiler warnings.
        (scrollbar_cb): Aha!  Thanks to the magic of reading the
        gtkrange.c source code, we now have better behaving scrollbars.
        Clicking in the trough/end arrows now works correctly instead of
        the semi-hackish way it did earlier today.

        * ui-gtk.c (Fgtk_define_enumeration): New function that can 

        * scrollbar-gtk.c (scrollbar_cb): Combined the horizontal and
        vertical scrolling callbacks.  Differentiate between them by
        user_data. 
        (scrollbar_cb): Don't bother trying to figure out whether this is
        a line-by-line or page-wide scrolling operation.  Just go by what
        is in the value of the adjustment and issue a scrollbar_*_drag
        event.  This is not 100% correct, but it at least gets it
        working.  Doing it 'right' might not be possible with standard Gtk
        scrollbars.
        (scrollbar_drag_hack_cb): New function to set
        vertical_drag_in_progress and inhibit_slider_change when we might
        be starting a drag in the scrollbar.
        (gtk_update_scrollbar_instance_status): Better setting of the
        adjustment variables.  Scrolling by pages (clicking in the trough)
        works now.  Line-by-line seems to work too.

2000-05-03  William M. Perry  <wmperry@aventail.com>

        * scrollbar-gtk.c (gtk_update_scrollbar_instance_status): Got the
        thumb sizing correct finally.  Thanks to scrollbar-msw.c for
        showing how to do it sanely - scrollbar-x.c is a mess.

        * window.c (specifier_vars_of_window): Added GTK specific fallback
        for Vvertical_divider_line_width.

        * toolbar.c (specifier_vars_of_toolbar): Handle specifier
        fallbacks for GTK toolbars.

        * gui-gtk.c (button_item_to_widget): Compiles under Gtk at least -
        more than likely completely broken.

        * glyphs-gtk.c (write_lisp_string_to_temp_file): Resurrected this
        blasphemous function to deal with XPMs.
        (gtk_xpm_instantiate): Now writes the XPM to a temp file, since
        GTK cannot deal with XPMs from memory.  Lame, but not fixable for
        gtk 1.2 - maybe for 1.4.
        (gtk_xpm_instantiate): Transparency works now.

        * gccache-gtk.c (gc_cache_lookup): Made non-hash based code
        compile.  It is not used, but what the hell.

        * faces.c (complex_vars_of_faces): Do GTK specific mucking about
        with face property fallbacks (fonts and colors)

        * events.c (event_equal): Added special case for GTK.
        (event_hash): Added special case for GTK.
        (format_event_object): Added special case for GTK events.

        * event-gtk.c (event_name): Use gtk_event_name helper function
        here.
        (handle_focus_event_1): Got rid of gtk-redisplay-hack variable and
        usage.

        * device.c (delete_device_internal): Delete 'popup' frames under
        Gtk as well as X.  Should this happen for HAVE_MSWINDOWS as well?

        * console.c (select_console_1): Make sure we set Vwindow_system
        like all the others.

        * frame-gtk.c (gtk_update_frame_external_traits): Added comments
        about why I didn't implement some portions of this function.

        * redisplay-gtk.c (gtk_output_string): Fixed the bizarre redisplay
        problem where all the default text would be drawn twice - once
        with the normal background colour, and once with the text-cursor
        background.  This was caused by a stupid typo and using the wrong
        GdkGC in the second call to gdk_draw_text_image... basically no
        clipping was being done on the image, so the whole thing was
        redrawn.
        (gtk_output_string): Call gdk_draw_text if we have a background
        pixmap, otherwise things look REALLY weird.
        (gtk_clear_region): Had a misplaced brace that caused the non-text
        area of a frame not to get the correct background.

2000-05-02  William M. Perry  <wmperry@aventail.com>

        * glyphs-gtk.c (gtk_xpm_instantiate): Need to write pixmaps to a
        temp file and then read with gdk_pixmap_create_from_xpm () - there
        is no way to read from a memory buffer (yet - I might write one
        for Gtk 1.4)

        * glyphs.c: Don't include xpm.h when building with Gtk.
        (pixmap_to_lisp_data): Alternate implementation for Gtk.

        * device-gtk.c (gtk_init_device): Call gdk_imlib_init if
        available, otherwise the 'Pixmap' theme engine doesn't work.
        Losers.

        * glyphs-gtk.c (gtk_xpm_instantiate): Now at least tries to
        instantiate XPM images.
        (init_image_instance_from_gdk_pixmap): Utility function to create
        a glyph from an existing GdkPixmap.

        * device-gtk.c (Fgtk_style_info): Attempt to expose the background
        pixmaps from a Gtk style.