Mon Sep 13 01:34:53 1999 Tim Janik <timj@gtk.org>
* gtk/gtksignal.c: renamed two variables, hopefully didn't introduce
short lived bugs, that would allow language bindings to do surgeries
to our guts.
Fri Sep 10 15:22:50 1999 Owen Taylor <otaylor@redhat.com>
* gtk/gtkwindow.c (gtk_window_size_allocate): Fix typo
where comparison was being done against an uninitialized
value causing intermittant results depending on
compiler flags. Also make it clearer that we aren't
ever initializing the child as 0x0 (though this will
be caught in gtk_widget_size_allocate())
Fri Sep 10 10:06:56 1999 Owen Taylor <otaylor@redhat.com>
* gtk/gtkwindow.c (gtk_window_set_transient_for): Move
gtk_window_unset_transient_for() call after we do checks
involving the old transient parent.
[ From Lance Capser <lmc@cyberhighway.net> ]
Mon Sep 6 00:11:56 1999 Lars Hamann <lars@gtk.org>
* gtk/gtkctree.c (gtk_ctree_is_ancestor): return FALSE if node has
no children. (Reported by: Chris Rogers <gandalf@pobox.com>)
* gtk/Makefile.am (gen_sources): Improve VPATH builds.
`$@' is valid only in the build dir, not after we've done
`cd $srcdir'. Also use `test -f' instead of less portable
`test -e'.
* new configure.in option --disable-rebuilds to completely disable
rebuilds of autogenerated sources.
* check for 5.002 now, to avoid failing autogeneration build rules due
to old perl versions.
* fonts (and fontsets) are cached now.
* more autogeneration make rules and dependancy fixups, we should be
save with autogeneration up to make -j12 now ;)
* new window position GTK_WIN_POS_CENTER_ALWAYS, which will recenter the
GtkWindow on every size change.
* major rework of window manager hints handling code, fixed a bunch of
races with the new resizing code.
* the new wm hints and resizing code is absolutely perfect and bug free now,
it only lacks testing ;)
* fixed up various rc style memory prolems.
* gtk_widget_modify_style() now properly changes the style of realized widgets
and references the style passed into it. if people worked around this bug,
this will introduce a slight memory leak in their code.
The code should typically look like:
GtkRcStyle *rc_style = gtk_rc_style_new ();
[...]
gtk_widget_modify_style (widget, rc_style);
gtk_rc_style_unref (rc_style);
* fix problems with positioning menus offscreen.
* GtkText fixes for some crashes and drawing errors.
* Better handling for unexpected window destroys in GDK and GTK+.
This should make it possible to use a GtkPlug and catch the
case where its parent socket is randomly killed.
* FAQ updates.
* FileSelection i18n patches, RadioButton fixups.
* miscellaneous other bugs fixed.
Fri Sep 3 15:59:56 1999 Owen Taylor <otaylor@redhat.com>
* gtk/gtktext.c (correct_cache_insert): Rewrite
for simplicity, and hopefully correctness.
(Fixes bug #1322, which was a segfault when
on some insertions with the properties around
the insertion set up just wrong.)
* gtk/gtktext.c (gtk_text_adjustment): When we receive
a "changed" signal, clamp the new value to the adjustment
bounds to avoid segfaulting if someone tries to change
the adjustment to a bogus value. (Bug #1795)
Thu Sep 2 16:33:59 1999 Owen Taylor <otaylor@redhat.com>
* gdk/gdkwindow.c: When we receive an unexpected
destroy notify on one of our windows, don't just
warn about it, also mark our windows as destroyed.
* gtk/gtkmain.c: Ignore unexpected destroy notifies
for children, for toplevel windows handle them
like delete_event.
* gtk/gtkplug.c: Add an unrealize handler so that
we unref plug->socket_window when we are done
with it.
Fri Sep 3 14:52:54 1999 Owen Taylor <otaylor@redhat.com>
* gtk/gtktext.c (clear_area): Fix stupid signedness
problem that was causing background to sometimes
be misaligned.
Fri Sep 3 12:26:33 1999 Owen Taylor <otaylor@redhat.com>
[ Fixes pointed out by Ettore Perazzoli <ettore@comm2000.it> ]
* gtk/gtkmenu.c (gtk_menu_position): Make
sure we never position menus with negative x, y,
since gtk_widget_set_uposition() can't handle that.
* gtk/gtkmenuitem.c (gtk_menu_item_position_menu):
Modify the positioning code a bit so that we always
put the top-left corner onscreen. (This is for
UI reasons, gtk_menu_position() now takes care of
gtk_widet_set_uposition() brokeness.)
Fri Sep 3 16:04:41 1999 Tim Janik <timj@gtk.org>
* gtk-config.in (--version): don't echo @GTK_VERSION@, but
@GTK_MAJOR_VERSION@.@GTK_MINOR_VERSION@.@GTK_MICRO_VERSION@, so the
AM_PATH_GTK() macros don't get confused by the -pre1.
Fri Sep 3 03:06:30 1999 Tim Janik <timj@gtk.org>
* configure.in: changed version number to 1.2.5-pre1,
interface age 0, binary age 5. we still require Glib-1.2.4 though.
* gtk/Makefile.am: fixed up things for -jx, x > 1.
* NEWS: updates for 1.2.5.
Sun Sep 5 08:48:51 1999 Owen Taylor <otaylor@redhat.com>
* gtk/gtkwidget.c (gtk_widget_modify_style): Ref the
RC style that is passed in. The lack of the ref
before was a bug. If people worked around this
bug, this will introduce a slight memory leak
in their code. The code should typically look like:
rc_style = gtk_rc_style_new ();
[...]
gtk_widget_modify_style (widget, rc_style);
gtk_rc_style_unref (rc_style);
* gtk/gtkwidget.c (gtk_widget_modify_style): Reset
the style if it was already set.
* gtk/gtkwidget.c (gtk_widget_set_name): Only set the
style if it was set before.
Sun Sep 5 08:10:53 1999 Owen Taylor <otaylor@redhat.com>
* gdk/gdkfont.c (gdk_font_hash_insert): Add
name => font and name => fontset hashes. The
name => fontset hash is a _big_ win since we
weren't previously caching fontsets at all and loading
fontsets is expensive. The name => font hash
is less of a win, but it does save us from doing
repeated XQueryFont calls on the same font.
* gdk/gdkprivate.h (struct _GdkFontPrivate): Add a names
list so we can remove font/fontset from hash.
Thu Sep 2 19:02:37 1999 Owen Taylor <otaylor@redhat.com>
* gdk/gdkproperty.c (gdk_atom_intern): Remove useless
and slightly confusing test. [ XInternAtom (,,TRUE)
will never return None ].
* gtk/testgtk.c (main): Add a check to see if we
are being run from the correct directory and
to quit nicely if we are not.
* gtk/gtkrc.c (gtk_rc_slist_remove_all): Make function
static.
* configure.in (REBUILD): Change check for perl5
to check explicitely for v >= 5.002. (5.001
does not work with our scripts.)
Thu Sep 2 23:00:03 1999 Tim Janik <timj@gtk.org>
* gtk/gtkenums.h (GtkWindowPosition): added GTK_WIN_POS_CENTER_ALWAYS.
* gtk/gtkwindow.c:
queue resizes unconditionally (gtk_widget_queue_resize will figure
what to do if the window is not realized).
(gtk_window_move_resize): only recenter the window
for GTK_WIN_POS_CENTER_ALWAYS.
(gtk_window_compute_reposition): handle GTK_WIN_POS_CENTER_ALWAYS in
the same way as GTK_WIN_POS_CENTER.
Thu Sep 2 22:39:27 1999 Tim Janik <timj@gtk.org>
* gtk/gtkwindow.c (gtk_window_reposition): check for the last position
with (!(info->last_flags & GDK_HINT_POS)) instead of
(!info->last_flags & GDK_HINT_POS).
* gtk/gtkwindow.c (gtk_window_move_resize): constrain new_width and
new_height unconditionally, because we use these values even if
!default_size_changed && !hints_changed.
comented the (default_size_changed || hints_changed) case with
respect to resize rejects from the window manager.
* gtk/gtkwindow.c (gtk_window_move_resize): save info->last values
in the zvt condition hack, since this includes the window hints, set
the hints after the handling_resize case.
* gtk/gtkwindow.c (gtk_window_show): constrain the default size that a
window is initially shown with to the geometry.
Thu Sep 2 07:38:56 1999 Tim Janik <timj@gtk.org>
* gtk/gtkwindow.c (gtk_window_move_resize):
s/size_changed/default_size_changed/g so i know what's
really going on (frying brain on smaller flame now).
Thu Sep 2 05:47:47 1999 Tim Janik <timj@gtk.org>
* gtk/gtkwindow.c (gtk_window_move_resize): queue a new resize if
we have size_changed upon handling_resize. this is a gross
workaround for the broken zvt widget and should be removed in
1.3 again (search for FIXME).
Owen provided an accurate comment for this:
/* We could be here for two reasons
* 1) We coincidentally got a resize while handling
* another resize.
* 2) Our computation of size_changed was completely
* screwed up, probably because one of our children
* is broken. It's probably a zvt widget.
*
* For 1), we could just go ahead and ask for the
* new size right now, but doing that for 2)
* might well be fighting the user (and can even
* trigger a loop). Since we really don't want to
* do that, we requeue a resize in hopes that
* by the time it gets handled, the child has seen
* the light and is willing to go along with the
* new size. (this happens for the zvt widget, since
* the size_allocate() above will have stored the
* requisition corresponding to the new size in the
* zvt widget)
*
* This doesn't buy us anything for 1), but it shouldn't
* hurt us too badly, since it is what would have
* happened if we had gotten the configure event before
* the new size had been set.
*/
Sat Sep 4 08:39:26 1999 Owen Taylor <otaylor@redhat.com>
* gdk/gdkwindow.c (gdk_window_set_geometry_hints)
gdk/gdkwindow.c (gdk_window_set_hints):
Don't omit setting the properties if flags == 0 -
there may be an existing set of properties there
already. (Very old bug. Would it be better to
delete the property instead?)
* gdk/gdkselection.c (gdk_selection_property_get): Fix
spelling error in comment.
Wed Sep 1 20:46:11 1999 Owen Taylor <otaylor@redhat.com>
* gtk/gtkrc.c: deal properly with the fact that RC
style lists may include rc styles more than once.
* gtk/gtkrc.c (gtk_rc_append_pixmap_path): Removed
unused static function.
* gtk/gtkrc.c (gtk_rc_find_pixmap_in_path): Keep a
stack of directories of RC files currently being
parsed and implicitely add them to pixmap path.
This fixes a bug where the directory would get
appended then overwritten by pixmap_path declarations.
(bug #1462, from Peter Wainright <prw@wainpr.demon.co.uk>)
* gtk/gtkthemes.c (gtk_theme_engine_unref): Call
theme's exit function. (Patch from Peter Wainwright,
bug #1454)
* gtk/gtkradiomenuitem.c (gtk_radio_menu_item_destroy):
Add a destroy() handler to take care of removing
group for menu item. (Fixes bug #1197)
* gtk/gtkwidget.c (gtk_widget_size_request): Fixed thinko
in warning message.
Wed Sep 1 21:27:42 1999 Tim Janik <timj@gtk.org>
* gtk/gtkwindow.c (gtk_window_move_resize): don't require a server
roundtrip to figure window's width and height, since we know that
anyways from widget->allocation.
Wed Sep 1 14:05:30 1999 Owen Taylor <otaylor@redhat.com>
* gdk/gdkimage.c (gdk_image_new): Use gdk_error_trap_push()
to avoid stomping over gdk_error_warnings.
* gdk/gdkimage.c (gdk_image_new): compute image->bpp
as (bits_per_pixel + 7) / 8. This gives the same
result as before for multiples of 8, but actually
a "reasonable" value for 1bit or 4bit displays.
Wed Sep 1 10:38:37 1999 Owen Taylor <otaylor@redhat.com>
* gtk/gtkwindow.c (gtk_window_compute_hints): Removed
GTK_WIDGET_REALIZED() assertion - we can compute the
hints before we are realized.
* gtk/gtkwindow.c (gtk_window_move_resize): Reorder
hint changing so that we have a value of hints_changed
when we decide whether to constrain the window size.
* gtk/gtkwindow.c (gtk_window_move_resize): Spelling fix.
* gtk/gtkwindow.c (gtk_window_constrain_size): cleanups,
change back to G_MAXINT.
Wed Sep 1 06:54:59 1999 Tim Janik <timj@gtk.org>
* gtk/gtkwindow.c (gtk_window_get_geometry_info): zero initialize
new GtkWindowGeometryInfo, so fields like GdkGeometry geometry
contain uncluttered values.
(gtk_window_compute_hints): simply assert that window is realized
and that geometry_info is valid, since we rely on this anyways.
(gtk_window_constrain_size): major cleanups to the code.
if (flags & GDK_HINT_BASE_SIZE) use geometry's base width and height
for the base size, instead of the minimums. use 32767 as max width
and height (like in gtkwindow.c) instead of G_MAXINT.
Wed Sep 1 04:41:25 1999 Tim Janik <timj@gtk.org>
* cleaned up the GtkContainer.need_resize flag handling mess, we
only need to force resize requests when we were prematurely
realized, or our widget tree was modified when we were temporarily
hidden. handling these cases directly upon showing the window (i.e.
while the GdkWindow is still unmapped) avoids the need to wait for
a configure event response and therefore makes the GUI more snappier
and avoids blank windows during the roundtrip.
* gtk/gtkwidget.c:
(gtk_widget_hide):
(gtk_widget_show): don't queue resizes on toplevels, they know how
to deal with matters.
* gtk/gtkcontainer.c (gtk_container_queue_resize): set the ->need_resize
flag directly for not visible resize containers and spare us unecessary
signal emissions.
* gtk/gtkwindow.c:
(gtk_window_realize): if we need to enforce premature size allocation,
queue a container resize so we are correctly resized later on.
(gtk_window_init):
(gtk_window_size_request):
don't freak around with the ->need_resize flag,
gtk_container_queue_resize() will care about that.
(gtk_window_show):
handle initial resizing issues here, we can handle matters better in
this place, especially since we know that our GdkWindow is still
unmapped.
(gtk_window_move_resize):
don't care about ->need_resize at all.
handle size changes properly that occoured while we waited for a
configure event.
Tue Aug 31 15:58:46 1999 Owen Taylor <otaylor@redhat.com>
* gtk/gtkwidget.c (gtk_widget_set_uposition):
* gtk/gtkwindow.[hc] (gtk_window_reposition):
Move the hint setting code from gtk_widget_set_uposition
to here; set the hints so that we respect any previously
set geometry hints.
* gtk/gtkwindow.c (gtk_window_compute_reposition): Don't
change the window hints here or move the window here,
let that happen in gtk_window_move_resize().
Tue Aug 31 06:58:52 1999 Tim Janik <timj@gtk.org>
* gtk/gtkwidget.c (gtk_widget_event): when discarding exposes due
to already queued resizes (and therefore redraws) on a widget, check
its anchestry as well.
* gtk/gtkcontainer.c:
(gtk_container_queue_resize): clear resize widgets for resize
containers before aborting prematurely. this is especially important
for toplevels which may need imemdiate processing or their resize
handler to be queued.
(gtk_container_dequeue_resize_handler): added new internal function for
gtkwindow.c.
* gtk/gtkwindow.c (gtk_window_move_resize): if we are resizing due to a
configure event, take possible changes in window position into account
as well.
if we request a new window size, queue up a resize handler that will
last until the configure event response arrives.
combined the ->need_resize case (initial show) with the general size
(hints) changed case and added even more comments.
if !auto_shrink, only revert to the old allocation if the new size
is smaller than the current allocation.
Tue Aug 31 11:55:20 1999 Owen Taylor <otaylor@redhat.com>
* gtk/gtkwindow.c (gtk_window_set_arg, gtk_window_set_policy,
gtk_window_set_geometry_hints
gtk_window_set_default_size):
When hints are set, queue a resize so that the hints will
be eventually reset on the toplevel.
* gtk/gtkwindow.c (gtk_window_show): Use
gtk_window_compute_default_size(). Clear the need_resize flag
on the initail map so that we don't unnecessarily trigger the
resize code.
* gtk/gtkwindow.c (gtk_window_move_resize): Split apart
into separate functions. Compare the hints we are setting
with what we set last time so that we can accurately
tell when we need to reset the hints.
* gtk/gtkwindow.c (gtk_window_compute_default_size): New
function to figure out the size from requisition
and default_size.
* gtk/gtkwindow.c (gtk_window_constrain_size): Function
from fvwm to constrain a size to the geometry hints.
* gtk/gtkwindow.c (gtk_window_compare_hints): New functio
to compare two sets of geometry hints.
* gtk/gtkwindow.c (gtk_window_compute_hints): Renamed
from gtk_window_set_hints(), just compute the hints,
don't set them.
* gtk/gtkwindow.c (gtk_window_compute_reposition): Move
code from gtk_window_move_resize() to separate function,
rationalize a bit.
Tue Aug 31 13:05:03 1999 Owen Taylor <otaylor@redhat.com>
* gtkrc.h: Move the ref_count member out of the GtkRcStyle
structure into a new private structure.
* gtkrc.c: Split GtkRcStyle into public/private.
In the private part, add a list of pointers to the
RcStyle lists this RcStyle participates in.
* gtkrc.c: When a RcStyle is free, remove all
lists referencing it from the
realized_style_ht hash, and free those lists.
* gtk/gtkrc.c (gtk_rc_clear_styles): Don't call
gtk_rc_init(), since that adds the default styles
to the list of parsed RC files again.
* gtk/gtkrc.c: Use gtk_rc_style_find() consistently.
Thu Aug 26 14:14:42 1999 Tim Janik <timj@gtk.org>
* gtk/gtkwidget.c (gtk_widget_size_allocate): reactivated the sanity
checks that ensure that a widget's allocation is at least 1 in width
and height. (GNOME note: this doesn't affect old panel code anymore,
because GtkSocket will request width and height of at least 1 since
Fri Jul 23).
* gtk/gtkwindow.c (gtk_window_realize): if the widget hasn't been
allocated yet (happens if the user realizes the window prematurely),
size request and allocate it.
(gtk_window_size_allocate): guard against guint underflows.
Wed Aug 25 19:01:36 1999 Lars Hamann <lars@gtk.org>
* gtk/Makefile.am (gtk_built_sources): reordered stamp-gtk.defs
in gtk_built_sources, so gtk.defs gets built prior to all other
sources.
