The check for a possible implicit paint flush before queueing an
antiexposure was wrong. An implicit flush doesn't actually NULL
the implicit paint, se we have add a flag to explicitly track if
it is flushed.
Passing region into _gdk_gc_set_clip_region_internal takes ownership,
so don't use it after that. We can just as well just move the usage
above the call.
Generally you only need to work around bugs in one specific app, so we
don't want to affect the applications that application will start.
Thus we unset GDK_NATIVE_WINDOWS after reading it.
For toplevels, never apply clip as shape, instead apply shape.
This way we don't have to re-set it all the time as the window size
changes. Furthermore, this change fixes unsetting a shape on a
toplevel window which didn't actually unset the shape before.
Additionally we never apply clips as shape if the shape would just
be the same as the regular window size. This means we won't unnecessarily
add a useless shape to most native child windows (and additionally this
helps apps that do weird X stuff that don't expect these shaped windows).
I.e. we use:
impl_iface = GDK_WINDOW_IMPL_GET_IFACE (private->impl);
And then use impl_iface instead of the full macro when calling vfuncs.
Also, in some places we avoid getting the iface multiple times.
For instance, two clients selecting for button events can cause BadAccess.
This fixes bug 592624, where a gdk_window_reparent caused us to re-set
the event mask, breaking the workaround for the mozilla BadAccess bug.
When updating the clip region on cairo context (due to window changes or
paints of different sizes) we reset the old clip region, but this was
erronously done inside a cairo_save/cairo_restore pair, which made the
reset not take effect.
This fixes bug #592263.
Instead of doing some magic in gdk_draw_drawable() to avoid double
offsetting when calling gdk_draw_drawable on the impl we call
the vfunc directly on the impl. Thus removing the weird magic from
gdk_draw_drawable().
I tested this with the testgtk test "text", where if the original magic
code is disabled typing a newline in the middle of a text line causes
the double offset issue to appear.
When the clip mask is completely inside the drawable clip region we
don't change the clip at all. However, we did set region_tag_applied,
so when the drawable clip was removed we removed the original clip mask.
This is no good, so we fix that by returning early in this case.
Fixes issue reported in bug #592752.
The fallback pixbuf rendering case ends up calling gdk_draw_image() on the
destination drawable wrapper, which resets the previously set clip region.
So, we need to manually get the impl and draw on that directly.
This fixes bug 592752 where we don't clip pixbuf rendering on non-render
Xservers.
Some applications make weird assumtions on Gtk+ that do not work anymore
with the new client-side windows support. For instance SWT/Eclipse reorders
the stacking order of the X windows directly without telling gdk this,
which breaks gdk drawing as gdk now relies on knowing the stacking order
for window clipping.
This introduces a GDK_NATIVE_WINDOWS environment variable, which if set
causes Gtk+ to always use native windows. Its more compatible with
pre-csw Gtk+ behaviour if you do weird X-specific hacks, although it does
limit the size of GdkWindows to 65535x65535.
gdk_window_get_geometry calls the native function for all non-native
windows. This returns coords relative to the native parent. We need
to convert this to be relative to the client side parent.
This fixes DnD coordinates in firefox (bug 588437).
If antiexposures are queued and then we do the outstanding moves
we will queue translations that will affect incomming exposes
wrongly wrt the antiexposure. So, make sure we process the outstanding
moves before doing the antiexposure.
Additionally this commit adds a bunch of comments about how the
expose handling works and fixes a small bug that made us always
flush outstanding moves a little to early than necessary.
Set the input_window_destroy and input_window_crossing methods of the
GdkWindowImplIface on Windows. Add implementation of
_gdk_input_crossing_event that is mostly dummy, though.
Apps may change the window hierarchy while recursing over it by
destroying windows from the expose event handler. We need to copy
the children list and ref all the children while recursing.
This fixes some crashers in gedit (bug #589367, bug #591434)
The X11 queue_translation operation uses NextRequest to get the serial
of the XCopyArea operation where the translation should end. However,
if the gc passed to gdk_draw_drawable has a non-flushed clip region
(which it commonly has now for the window clipping) then the next
operation will be the GC flush, not the XCopyArea.
To handle this right we now pass in the GC to be used to
queue_translation and ensure that it is flushed before calling
NextRequest().
When we copy the region we need to also re-expose the areas of
the copied region that was not also in the destination.
And, we need to do this invalidation after the move, as the
move also moves any invalid area.
If there are outstanding moves in an area that intersects
the source of an outstanding move we need to move the invalid
area correspondingly, otherwise we will expose the wrong area
as the outstanding move copy will happen before we expose
the invalid area.
When moving an area we move any invalid region in this area to the
new place, but there really is no need to remove the old invalid
area as it would just be invalidated again (being newly exposed).
This extends the usage of the native clear region call such that its
called also for windows that have parent relative background all the
way up to a native window. That way we get true background relative
background clearing even to a foreign parent, which means that some
transparent notification icons look right again.
GdkDrawable->draw_drawback was replaced with a new vfunc
draw_drawback_with_src that is now called from gdk_draw_drawable.
However, some code seems to call the vfunc directly (see bug #591288),
so make it chain to the new call.
Note that such direct vfunc calls are a bad idea and won't work for all
cases.
This moves the native show/hide calls to the generic code
for calculating viewable rather than in its own separate code
called from gdk_window_show/hide. This simplifies the code a bit,
but most significantly it means things are correctly shown when
they become viewable for other reasons than a show/hide call.
For instance, this fixes bug 590442 (gvim embedding) where the
toplevel GtkPlug is mapped by the embedder and we didn't previously
pick up that the native children became viewable and should be shown.
We used to invalidate the whole window when raised, but this is
unnecessary much, we now just invalidate the visible area minus the
previously visible area.
This also fixes a problem where expose calling raise caused a loop (#588438)
This never worked before csw since the root window is never
set as IS_MAPPED, but with the new viewable check (which is
true for the root window) we could erronously queue exposes
on the root window.
This happened unexpectedly in bug 589369, where metacity
got a GraphicsExpose event on the root window due to some
graphics operation, queueing an expose which would be handled
by clearing that area. That is fixed with this commit.
This seems to more or less fix the build. On Tiger there are still issues
with libresolv missing on the link line, I will figure out what's up with
that soonish.
We get the real pixmap size and use as cairo surface size rather
than doing some magic to try to get clipping on the right hand size
(and the magic looks wrong anyway).
The previous code could result in the width/height being to big for
the cairo 28.4 fix point size and thus not drawing anything.
This fixes bug #588076.
When the fallback for gdk_draw_pixbuf (gdk_drawable_real_draw_pixbuf)
is called with a window destination we have already applied any clip
regions and offsets for the window, but the window we get is a wrapper
and not an impl. We have to ensure we really draw to the impl, as
otherwise the pixbuf drawing will be clipped by client side subwindows.
This fixes bug 588553.
gdk_window_input_shape_combine_mask() can accept NULL for the mask
parameter, but it wasn't checking for NULL before passing the
resulting GdkRegion to gdk_region_destroy(). Fixes#589275
When the region is empty we can return early, because there
is no more area to remove. This happens often for children
of scrolled windows (i.e. things that are clipped out.
Implement _gdk_win32_window_destroy() by just renaming
_gdk_windowing_window_destroy(), removed superfuous stub.
Also cleaned up implementations of gdk_win32_window_set_background()
and gdk_win32_window_set_back_pixmap() - removed stuff now done at the
respective gdk_window_*() function.
Resurrcetion and adaption of find_window_for_mouse_event(). The window
receiving the WM_MOUSEMOVE, WM_?BUTTONDOWN is not necessarily the one
interested in GDK_(ENTER|MOTION|LEAVE)_NOTIFY
http://bugzilla.gnome.org/show_bug.cgi?id=588373
Also added some more more TODO_CSW and disabled print_event(): it can not
cope with the new _gdk_windowing_got_event() eating/morphing events.
The scrolling and region moving code needs to avoid drawing when the
window is mapped, which it did. However, it also needs to avoid
drawing when any of its parents are not mapped, which it didn't so
switch to using gdk_window_is_viewable().
This fixes the index rendering in evolution (#588169)
When we ungrab the pointer we don't get enter events for the window the
pointer is in at the time of the ungrab, so we manually query for the
window the pointer is in. The same thing actually happens on re-grab if
the previous grab was !owner_events (meaning we don't get crossing events
for windows other than the grab) but the new grab is owner_events (and
thus non-grab windows need to get crossing events).
This factors out some common code and enables it also for the re-grab
to owner_events case.
If we get a nonlinear enter/leave notify on the toplevel we need
to set nonlinear in all the events we send, even if the in-toplevel
tree is linear.
This fixes combobox menus popping down immediately when you click
(not hold). (bug #587559)
We really need to wait for the MapNotify from the xserver to ensure
that the window has been mapped, as it may be delayed by the WM, network
or similar things.
This fixes a problem in the /ui-tests/keys-events gtk test
This is more important in the client side windows world, as clip masks
may be used for clipping to non-native subwindows. This fixes a bug
in aisleriot where it uses masked pixbuf drawing and it ends up drawing
over the moving card subwindow.
Turns out pygtk build broke due to the argument addition to draw_drawable.
So, we now add a new vfunc for the new draw_drawable and are thus
backwards compat.
we now use gdk_offscreen_window_set_embedder() instead of a signal
to get the parent. This also replaces set_has_offscreen_changes.
Rename "parent" in all embedding related names to "embedder" to make it
more obviously different than the normal parent.
Rename gdk_window_get_offscreen_pixmap to gdk_offscreen_window_get_pixmap
to match the other offscreen calls.
Rename gdk_window_offscreen_children_changed to gdk_window_geometry_changed
as this is more descriptive.
We emulate visibility by unmapping (in X) all mapped (according to
gdk) windows that are not visible (all parents mapped). This is because
there may be client side windows inbetween the native windows in the
hierarchy, so you can't know a native window is visible just because
all the parent native windows are mapped.
However, we don't want to unmap foreign windows, as that may cause all
sort of unexpected issues. This should be safe, because generally the
parent of a foreign window is a native window (e.g. a socket), so its
gdk visibility state is the same as the Xserver one.
This fixes an issue with the GDM notification area where the unmap of
the plug caused spurious UnmapNotify events that confused GtkSocket and
caused icons to become one pixel wide.
Last commit was bad, as it didn't clip against client side
children. This implements such clipping first and then
only clears the rectangles that need to be cleared.
This is how cairo works with native children.
This fixes an issue where the drag target outline is visible on the
evolution folder treeview when it isn't with native windows.
Without this we can't draw to them, which caused problems for e.g.
gnome-settings-daemon clearing the background when the desktop
background changed.
Note: We don't actually clip away child windows from the root window,
the clip is just based on the size of the root window.
Use g_*gettext functions in gdk-pixbuf instead direct gettext
calls to benefit from the maybe-dont-translate functionality
in GLib. Also, replace a hand-rolled version by g_dpgettext2
in gtkbuilderparser.c. Fixes bug 585791.
In the destroyed window case in do_synthesize_crossing_event we didn't
return a value which can cause infinite "loops". Always return FALSE
to make sure the idle doesn't run again.
Doing this directly had some issues with picking going recursive in
clutter-gtk. Furthermore, doing it in an idle means we can coalesce
multiple calls (which is common due to widget changes) in the same
toplevel to just one call.
You can call this if you have offscreen children and the geometry of
them changed. This will cause re-picking of the active window sending
enter and leave events as needed.
We want to be able to map any window coordinate to a root coordinate, not
just the origin, because you can't rely anymore on a simple translation
from window coordinates to parent with offscreen windows. This lets
us e.g. pop up menus in the right place even if they are popped up from
a no-window widget.
We use the offscreen signals for getting parent, picking
children at a point and mapping coordinates between windows
embedding offscreens and offscreens.
This means we have two hierarchies more or less, one visible to apps via
the standard APIs and for drawing where the offscreens are their own
separate toplevels, and another one for event handling where embedded
offscreens appear as if they were children of the embedding window.
3 signals are for offscreen windows
get-offscreen-parent: Get the parent window an offscreen is embedded in
to-parent: Convert coordinates from offscreen to parent
from-parent: Convert coordinates from parent to offscreen
1 signal is for the window embedding offscreens:
pick-offscreen-child: This picks what (if any) offscreen is at a specific position
The last signal is only used if you call gdk_window_set_has_offscreen_children
to tell gdk that the window has embedded offscreen children.
Add get-pointer signal for offscreen window pointer getting
Apps using offscreen windows can connect to get-pointer on offscreen
windows in order to make gdk_window_get_pointer() return correct
values.
Add get-offscreen-parent signal
Add signals for from-parent and to-parent coordinate mapping
Add pick-offscreen-child signal
We returned the innermost child that has the pointer, which is not right.
Only the direct child that has the pointer in it should be reported (if any).
This also removes the (unimplemented) possibility to change a window to
non-native. This seems generally not very useful, and there are some problems
with it, for instance if two "users" need a window to be native and then
one of the "users" doesn't need it anymore it can't change it back, because
it is unaware of the other reason the window is native.
If a native window or a window with a native subwindow is moved or resized
then the client window moves and implicit paints need to be flushed since
the native window move will copy/overwrite data. This may happen while there
is an outstanding paint if the move is inside an expose event (weird, but
flash embedded in webkit hit this).
Right now we're asserting here, but the right fix is to allow this but
to not flush the parts that are currently in a (non-implicit) paint. This
means we flush all results from previous not-yet-flushed exposes, but not
the ones being drawn.
Calling gdk_window_get_events() had the side-effect of letting
property change notification through to the application, which
was not intended. Now we keep StructureNotify and PropertyNotify
filtered out when they were before. Reported in bug 582003.
