Alexander Larsson [Thu, 30 Oct 2014 09:30:33 +0000 (10:30 +0100)]
GdkWindowX11: Keep the impl window cairo surface around until destroy
We used to have a weak ref to the cairo surface and it was keep
alive by the references in the normal windows, but that reference
was removed by
d48adf9cee7e340acd7f8b9a5f9716695352b848, causing
us to constantly create and destroy the surface.
https://bugzilla.gnome.org/show_bug.cgi?id=738648
Matthias Clasen [Tue, 28 Oct 2014 03:51:17 +0000 (23:51 -0400)]
Make the inspector window translucent instead of lowering
Make the inspector window click-through and see-through
when we can, instead of lowering it. This is useful, since
Wayland has no lowering.
Lapo Calamandrei [Wed, 29 Oct 2014 18:54:43 +0000 (19:54 +0100)]
Adwaita: fix for the previous colorswatch fix...
Make the selected colorswatch inhert on backdrop.
Lapo Calamandrei [Wed, 29 Oct 2014 18:50:58 +0000 (19:50 +0100)]
Adwaita: fix selected colorswatch white border
Lapo Calamandrei [Wed, 29 Oct 2014 17:10:39 +0000 (18:10 +0100)]
Adwaita: first round of OSD style converage
Some code refactoring, entry and spinbuttons and flat buttons
styled, more to come...
Carlos Garnacho [Wed, 29 Oct 2014 16:31:15 +0000 (17:31 +0100)]
wayland: Emit GDK_OWNER_CHANGE events around wl_data_offer changes
This way the upper GtkClipboard code emits correctly ::owner-change when
receiving these events.
Alexander Larsson [Wed, 29 Oct 2014 12:52:04 +0000 (13:52 +0100)]
GdkX11GLContext: Fix leak
Alexander Larsson [Wed, 29 Oct 2014 12:33:08 +0000 (13:33 +0100)]
Cache default gdk visuals in the GDK_VISUALS property on the root window
This means we don't have to try to initialize opengl in every gtk
instance that is stated. It will only happen for the first one.
https://bugzilla.gnome.org/show_bug.cgi?id=738670
Alexander Larsson [Wed, 29 Oct 2014 11:35:07 +0000 (12:35 +0100)]
X11: Pick better system and rgba visuals for GL
We want to create windows with the default visuals such that we then
have the right visual for GLX when we want to create the paint GL
context for the window.
For instance, (in bug 738670) the default rgba visual we picked for the
NVidia driver had an alpha size of 0 which gave us a BadMatch when later
trying to initialize a gl context on it with a alpha FBConfig.
Instead of just picking what the Xserver likes for the default, and just
picking the first rgba visual we now actually call into GLX to pick
an appropriate visual.
Matthias Clasen [Wed, 29 Oct 2014 10:42:08 +0000 (06:42 -0400)]
Remove a redundant call
No need to set shapes twice in a row.
Matthias Clasen [Sun, 26 Oct 2014 23:14:01 +0000 (19:14 -0400)]
GtkPopoverMenu: Add API to open a submenu
This allows alternative menu switching to be implemented, besides
the builtin GtkModelButton support.
Matthias Clasen [Sun, 26 Oct 2014 23:11:33 +0000 (19:11 -0400)]
Add a scale to the popover menu example
Demonsttrate that this is flexible enough for the intended uses.
Matthias Clasen [Sun, 26 Oct 2014 23:04:28 +0000 (19:04 -0400)]
Add documentation
Document GtkModelButton and GtkPopoverMenu.
Matthias Clasen [Sun, 26 Oct 2014 13:08:37 +0000 (09:08 -0400)]
Adapt the testpopover example
Add the required boxes for each submenu. We now need to specify the
submenu name only once per submenu.
Matthias Clasen [Sun, 26 Oct 2014 22:58:55 +0000 (18:58 -0400)]
GtkPopoverMenu: Simplify
Change things so each submenu is a single child, and users are
required to create the box themselves.
Matthias Clasen [Sun, 26 Oct 2014 22:13:29 +0000 (18:13 -0400)]
Adapt the testpopover example
Drop all all explicit role setting from the testpopover example.
Matthias Clasen [Sun, 26 Oct 2014 22:13:02 +0000 (18:13 -0400)]
GtkModelButton: Don't require an explicit role
When we have an action-name, we can deduce the role from the
action that is looked up by the action helper.
Matthias Clasen [Sun, 26 Oct 2014 22:10:23 +0000 (18:10 -0400)]
Adapt the testpopover example
Matthias Clasen [Sun, 26 Oct 2014 22:08:26 +0000 (18:08 -0400)]
GtkModelButton: Rename action-role to role
Rename action-role to role, and make all properties readable.
Matthias Clasen [Sun, 26 Oct 2014 03:30:12 +0000 (23:30 -0400)]
GtkModelButton: Drop unused :accel property
Matthias Clasen [Sun, 26 Oct 2014 20:23:50 +0000 (16:23 -0400)]
Adapt the testpopover example
Drop all uses of has-submenu.
Matthias Clasen [Sun, 26 Oct 2014 20:23:31 +0000 (16:23 -0400)]
Drop GtkModelButton:has-submenu
Replace the has-submenu property with checking menu-name != NULL.
Matthias Clasen [Fri, 24 Oct 2014 03:39:39 +0000 (23:39 -0400)]
Adapt the testpopover example
We now use GtkPopoverMenu, and just add the model buttons
directly to it.
Matthias Clasen [Fri, 24 Oct 2014 03:38:46 +0000 (23:38 -0400)]
GtkModelButton: Add a menu-name property
Add a menu-name property and use it in a default implementation
of ::clicked to switch menus if we are inside a stack. This means
GtkModelButton is no longer entirely generic, but rather expects
to be used inside a GtkPopoverMenu. It still works in other contexts
too, of course.
Matthias Clasen [Fri, 24 Oct 2014 03:38:09 +0000 (23:38 -0400)]
Add GtkPopoverMenu
GtkPopoverMenu is a GtkPopover subclass that has a menu-like stack
built-in.
Matthias Clasen [Fri, 24 Oct 2014 01:55:20 +0000 (21:55 -0400)]
Add a manual popover menu example
Reproduce the testpopover popover manually, by packing model
buttons into a stack in a popover.
Matthias Clasen [Fri, 24 Oct 2014 01:43:28 +0000 (21:43 -0400)]
Make GtkModelButton a better GtkActionable
Rename the "toggled" property to "active", since that is what
GtkActionHelper expects to update for check and radio actions.
Also make the property readable, since GtkActionHelper wants
to read it.
Matthias Clasen [Fri, 24 Oct 2014 01:31:17 +0000 (21:31 -0400)]
Make GtkModelButton public
Matthias Clasen [Wed, 29 Oct 2014 02:36:50 +0000 (22:36 -0400)]
Drop an unused variable
Alexander Larsson [Wed, 29 Oct 2014 10:25:21 +0000 (11:25 +0100)]
GdkGLContextX11: Don't free NULL visual info
Matthias Clasen [Wed, 29 Oct 2014 00:23:47 +0000 (20:23 -0400)]
Fix toplevel input shape for csd windows
We need to maintain the application-set and the the csd
input shape separately, and intersect them to pass them
down.
Matthias Clasen [Tue, 28 Oct 2014 20:07:25 +0000 (16:07 -0400)]
Plug another memory leak
Also found by valgrind.
Matthias Clasen [Tue, 28 Oct 2014 20:06:42 +0000 (16:06 -0400)]
Plug some memory leaks in GL support
valgrind pointed these out.
Alexander Larsson [Tue, 28 Oct 2014 17:29:02 +0000 (18:29 +0100)]
GdkGlContextWayland: Don't require specific bit depths
This just means we may fail to get a gl context, we just require
*some* color depth.
Alexander Larsson [Tue, 28 Oct 2014 17:27:41 +0000 (18:27 +0100)]
Pick the first rgba visual, not the last
The visuals are typically sorted by some sort of "most useful first"
order. And picking the last one is likely to give us the weirdest
matching glx visual.
Alexander Larsson [Tue, 28 Oct 2014 15:43:38 +0000 (16:43 +0100)]
GdkGLContextX11: Ensure we get the fbconfig with the exact same visual
We really want a gl context with exactly the same visual, or we will
get a badmatch later, which hits us on nvidia as per:
https://bugzilla.gnome.org/show_bug.cgi?id=738670
Matthias Clasen [Tue, 28 Oct 2014 14:48:49 +0000 (10:48 -0400)]
Document Wayland opacity handling
Matthias Clasen [Tue, 28 Oct 2014 14:28:23 +0000 (10:28 -0400)]
wayland: Make window opacity work
Under wayland, the compositor doesn't have a 'overall window alpha'
knob, we just need to add the alpha to the buffers we send.
Client-side alpha, if you want to call it that.
Implement this by reusing the existing alpha support for non-toplevel
widgets. As a side-effect of the implementation, windows with RGBA
visual under X will now also use per-pixel alpha, instead of
overall alpha.
Matthias Clasen [Tue, 28 Oct 2014 12:08:49 +0000 (08:08 -0400)]
Bump atk dep
Needed for ATK_STATE_HAS_TOOLTIP.
Matthias Clasen [Tue, 28 Oct 2014 11:04:23 +0000 (07:04 -0400)]
Inspector: Make picking work without a grab
Turns out Weston takes an X grab while delivering button events.
Thus, trying to take a grab in response to a button press is
futile.
Jasper St. Pierre [Tue, 28 Oct 2014 04:44:48 +0000 (21:44 -0700)]
wayland: Don't use the default display
Pass a display in so that multi-display technology works.
Jasper St. Pierre [Tue, 28 Oct 2014 05:29:29 +0000 (22:29 -0700)]
gdkevents: Use the correct display to copy event data
Jasper St. Pierre [Tue, 28 Oct 2014 04:47:58 +0000 (21:47 -0700)]
gdkdisplay: Provide a default event_data_copy / event_data_free
Every single implementation but Quartz is a no-op for this, so just
provide it once rather than in every backend.
Matthias Clasen [Tue, 28 Oct 2014 04:56:46 +0000 (00:56 -0400)]
gtk-demo: Drop uses of gdk_cursor_new()
Use gdk_cursor_new_for_display() instead.
Matthias Clasen [Tue, 28 Oct 2014 04:54:39 +0000 (00:54 -0400)]
Add another multi-device invariant
The device and window passed to gdk_device_grab() must be
on the same display, or things are unlikely to work.
Document and assert this.
Matthias Clasen [Tue, 28 Oct 2014 04:53:59 +0000 (00:53 -0400)]
Drop an unused variable
Matthias Clasen [Tue, 28 Oct 2014 04:51:32 +0000 (00:51 -0400)]
Don't return GDK_GRAB_SUCCESS on failure
That seems ... counterproductive.
Matthias Clasen [Tue, 28 Oct 2014 04:44:33 +0000 (00:44 -0400)]
Mention the 'same display' requirement in the docs too
Matthias Clasen [Tue, 28 Oct 2014 04:38:59 +0000 (00:38 -0400)]
Deprecate gdk_cursor_new()
Too easy to introduce an unintended assumption about the display
being the default this way.
Matthias Clasen [Tue, 28 Oct 2014 04:36:17 +0000 (00:36 -0400)]
GtkAboutDialog: Don't use gdk_cursor_new
gdk_cursor_new returns cursors for the default display, and
thus is not safe to use in code that may run on non-default
displays.
Jasper St. Pierre [Tue, 28 Oct 2014 04:21:00 +0000 (21:21 -0700)]
gtkwindow: Use gdk_cursor_new_for_display
Otherwise, this crashes the inspector, since it now uses multi-display
technology, and on Wayland, resources like the cursor are client-private.
Jasper St. Pierre [Tue, 28 Oct 2014 04:18:15 +0000 (21:18 -0700)]
gdkwindow: Specify some invariants for cursors
Cursors should not be on a different display than their window /
device, as that would break Wayland.
Matthias Clasen [Mon, 27 Oct 2014 23:12:03 +0000 (19:12 -0400)]
Update for 3.15.1
Matthias Clasen [Mon, 27 Oct 2014 23:31:04 +0000 (19:31 -0400)]
Update expected results of a11y tests
The new has-tooltip state is showing up in a number of places,
as expected.
Jasper St. Pierre [Mon, 27 Oct 2014 23:54:33 +0000 (16:54 -0700)]
x11: Scale the frame extents as well
Matthias Clasen [Mon, 27 Oct 2014 21:13:36 +0000 (17:13 -0400)]
gtk3-demo: Add a frame to the sidebar
With overlay scrollbars, there's no visible separation between
the sidebar and the content otherwise.
Alexander Larsson [Mon, 27 Oct 2014 20:15:37 +0000 (21:15 +0100)]
opengl: Use GL_TEXTURE_2D rather than GL_TEXTURE_RECTANGLE_ARB when possible
This is more standard, and most driver support non-power-of-2 TEXTURE_2D
these days. We fall back for ancient drivers.
Alexander Larsson [Mon, 27 Oct 2014 20:13:28 +0000 (21:13 +0100)]
opengl: Use proper blend function for premultiplied alpha
Alexander Larsson [Mon, 27 Oct 2014 20:12:40 +0000 (21:12 +0100)]
gdkglcontext: Track whether to use GL_TEXTURE_2D or GL_TEXTURE_RECTANGL_ARB
Jasper St. Pierre [Mon, 27 Oct 2014 17:18:37 +0000 (10:18 -0700)]
cairoblur: Fix the blur again once more
The differences between the mutter and GTK+ code are subtle, but it
turns out that _gtk_cairo_blur_compute_pixels actually returns the
shadow *spread*. Since we use a triple box blur, the constant was
multiplied by 1.5 to approximate three chained box blurs. Split this
out and use the correct value for the lobe width.
Jasper St. Pierre [Mon, 27 Oct 2014 17:05:35 +0000 (10:05 -0700)]
cairoblur: Fix the blur code
We weren't passing in the right "d" value, which was causing the blur to
behave incorrectly, especially in the case of 1px blurs, which would
cause no blurs at all.
The blur should now match the web.
Alexander Larsson [Mon, 27 Oct 2014 15:33:37 +0000 (16:33 +0100)]
Rename gdk_gl_context_flush_buffer to gdk_gl_context_end_frame
This makes a lot more sense.
Matthias Clasen [Mon, 27 Oct 2014 11:19:50 +0000 (07:19 -0400)]
Indentation fix
Sebastien Lafargue [Sat, 18 Oct 2014 09:19:26 +0000 (11:19 +0200)]
popover menus: Take iconic into account
We don't want separators in both side of an iconic section
and use a 10px margin to separate two iconic sections.
Separators are also updated in case of dynamic insertion
( often used with menu items for plugins )
https://bugzilla.gnome.org/show_bug.cgi?id=738650
Matthias Clasen [Mon, 27 Oct 2014 11:05:22 +0000 (07:05 -0400)]
Ignore gdk-pixbuf deprecations
Cosimo Cecchi [Fri, 24 Oct 2014 22:26:31 +0000 (15:26 -0700)]
testlevelbar: fix CSS to be aligned with documentation
https://bugzilla.gnome.org/show_bug.cgi?id=707695
Cosimo Cecchi [Fri, 24 Oct 2014 22:25:36 +0000 (15:25 -0700)]
levelbar: don't use the current value style class for empty blocks
We want to be able to style the empty blocks independently of all the
offset styles, so remove the current style class when painting an empty
block.
https://bugzilla.gnome.org/show_bug.cgi?id=707695
Cosimo Cecchi [Fri, 24 Oct 2014 21:53:26 +0000 (14:53 -0700)]
testlevelbar: don't segfault when quitting testlevelbar
Quit the mainloop when the window is closed.
https://bugzilla.gnome.org/show_bug.cgi?id=707695
Matthias Clasen [Mon, 27 Oct 2014 10:49:46 +0000 (06:49 -0400)]
quartz: Move the FullscreenSavedGeometry definition earlier
Move the FullscreenSavedGeometry before the first (conditional)
use, to fix the build on OS X 10.6.
https://bugzilla.gnome.org/show_bug.cgi?id=737561
Marcus Karlsson [Sun, 26 Oct 2014 11:26:48 +0000 (12:26 +0100)]
quartz: fix build on 10.6 and below
It is not possible to successfully build GTK+ on OS X 10.6 and below
since NSFullScreenWindowMask is only available starting with 10.7. Add
ifdef guards around setStyleMask: in order to allow it to build on
earlier OS X releases.
https://bugzilla.gnome.org/show_bug.cgi?id=737561
Christophe Fergeau [Thu, 23 Oct 2014 19:13:29 +0000 (21:13 +0200)]
Return correct font from gtk_font_chooser_widget_find_font
Commit
30a1c4ab fixed several memleaks including one in
gtk_font_chooser_widget_find_font.
However, the fix causes one extra call to gtk_tree_model_iter_next()
after finding the font we look for (ie pango_font_description_equal
returns TRUE): the 'increment' part of the for loop
(gtk_tree_model_iter_next) is run before the 'exit condition' of the for
loop is evaluated.
This commit reverts this part of commit
30a1c4ab and adds an extra
call to pango_font_description_free in order to fix the leak.
https://bugzilla.gnome.org/show_bug.cgi?id=739111
Christophe Fergeau [Thu, 23 Oct 2014 21:30:35 +0000 (23:30 +0200)]
fontchooserwidget: Don't invalidate priv->font_iter in load_fonts
When using GtkFontChooserButton, the same GtkFontChooserWidget can be
hidden and shown multiple times. When doing that, the font that was
chosen the previous time should be the selected one in the
GtkFontChooserWidget, however this does not work as expected and a
somehow 'random' font gets selected (or none) instead.
Every time the font chooser widget is shown, its style will be updated,
causing gtk_font_chooser_widget_style_updated and then
gtk_font_chooser_widget_load_fonts to be called.
gtk_font_chooser_widget_load_fonts starts by clearing the GtkListStore
listing the available fonts, repopulates it, and then makes sure the
current font is selected.
However, this does not work as expected, as during the call to
gtk_list_store_clear, the cursor_changed_cb will be invoked multiple
times when the GtkTreeView cursor gets moved when the line where the
cursor currently is gets removed. This will cause the 'current font'
state (priv->font_desc) to be unexpectedly modified, and when
gtk_font_chooser_widget_load_fonts tries to reposition the cursor to the
'current font', we won't get the expect result.
This commit avoids that by making sure cursor_changed_cb does not get
called when we call gtk_list_store_clear in
gtk_font_chooser_widget_load_fonts.
https://bugzilla.gnome.org/show_bug.cgi?id=739111
Matthias Clasen [Mon, 27 Oct 2014 01:48:59 +0000 (21:48 -0400)]
Fix an out-of-bounds access in MyEnhancedXkbTranslateKeyCode
Commits
314b6abbe8d8daae and
eb9223c008ccf1c2faab were ignoring
the fact that the code where found is set to 1 was modifying
col - which was an ok thing to do when that part of the code
was still breaking out of the loop, but it is no longer doing
that (since 2003 !). Fix things up by storing the final col
value in a separate variable and using that after the loop.
https://bugzilla.gnome.org/show_bug.cgi?id=738886
Matthias Clasen [Sun, 26 Oct 2014 19:07:26 +0000 (15:07 -0400)]
Test h/vhomogenenous in teststack
Add separate checkboxes for hhomogeneous and vhomogeneous.
Matthias Clasen [Sun, 26 Oct 2014 19:03:58 +0000 (15:03 -0400)]
Use GtkStack:vhomogeneous for popover menus
This allows us to get rid of a size group.
Matthias Clasen [Sun, 26 Oct 2014 18:59:21 +0000 (14:59 -0400)]
GtkStack: split homogeneous
Split homogeneous into separate hhomogeneous and vhomogeneous
properties. vhomogeneous is what we want for popover menus.
Timm Bäder [Sun, 26 Oct 2014 14:21:06 +0000 (15:21 +0100)]
popover: Use the right area for drawing the background
The border shouldn't be subtracted.
Matthias Clasen [Sun, 26 Oct 2014 00:27:18 +0000 (20:27 -0400)]
GtkBox: Fix self-assignment for center widget
This is happening when glade is loading ui files containing
center widgets.
Marek Černocký [Sat, 25 Oct 2014 22:35:56 +0000 (00:35 +0200)]
Updated Czech translation
Marek Černocký [Sat, 25 Oct 2014 22:35:50 +0000 (00:35 +0200)]
Updated Czech translation
Matthias Clasen [Sat, 25 Oct 2014 19:55:54 +0000 (15:55 -0400)]
GtkPlacesSidebar: Don't crash on dnd
gtk_get_current_event() can return NULL. Don't crash if it does.
https://bugzilla.gnome.org/show_bug.cgi?id=739150
Matthias Clasen [Sat, 25 Oct 2014 17:58:29 +0000 (13:58 -0400)]
Revert "Inspector backend hack"
This reverts commit
f88a5dc8c78e6530baeebb6cba75fe3061926052.
Didn't mean to push this.
Matthias Clasen [Sat, 25 Oct 2014 17:31:57 +0000 (13:31 -0400)]
GtkSpinButton: Set the screen on the panel contexts
Failure to do so results in custom styling leaking through in
the inspector. This is pretty obvious, now that the inspector
is using a separate display connection and is generally isolated
from style changes.
Matthias Clasen [Fri, 24 Oct 2014 22:00:19 +0000 (18:00 -0400)]
Inspector backend hack
Matthias Clasen [Sat, 25 Oct 2014 14:31:31 +0000 (10:31 -0400)]
Broadway: Remove display assumptions
In various places, the broadway backend was just using
the default display and assumed that it is the broadway
display. That may not be the case in a multi-backend world,
so instead iterate over all displays and use the first
broadway display - still not perfect, but enough to survive
for now.
Matthias Clasen [Sat, 25 Oct 2014 14:00:57 +0000 (10:00 -0400)]
inspector: Plug a memory leak
Matthias Clasen [Sat, 25 Oct 2014 14:00:04 +0000 (10:00 -0400)]
inspector: Add style classes to the object list
We have a lot of space here now, so show some
useful information.
Matthias Clasen [Sat, 25 Oct 2014 03:36:35 +0000 (23:36 -0400)]
Fix a mixup in popover realization
When the window is on a non-default screen, popover_realize
ended up passing a visual and a parent_window from different
screens into gdk_window_new, which doesn't work. Fix it by
using the visual of the parent window.
Matthias Clasen [Sat, 25 Oct 2014 01:52:23 +0000 (21:52 -0400)]
inspector: Improve details buttons
Ensure that the details buttons are only sensitive when
we actually have details to show.
Matthias Clasen [Sat, 25 Oct 2014 01:14:16 +0000 (21:14 -0400)]
inspector: Improve tab switching
Switch the object and resource trees away from activate-on-single-click
and add a 'view details' button.
Matthias Clasen [Fri, 24 Oct 2014 22:38:15 +0000 (18:38 -0400)]
X11: Don't assume all displays are X11 displays
When iterating over the list of displays gotten from the
display manager, we have to check if what we got is actually
an X11 display.
Matthias Clasen [Fri, 24 Oct 2014 20:23:07 +0000 (16:23 -0400)]
Properly update the style cascade on screen change
GtkStyleContext was not properly handling the style cascade when
setting a screen, causing the inspector global CSS to affect the
inspector window, even though the inspector is using a different
screen now.
Matthias Clasen [Fri, 24 Oct 2014 20:12:33 +0000 (16:12 -0400)]
Fix the build
Some forgotten include path adjustments.
Matthias Clasen [Fri, 24 Oct 2014 19:55:31 +0000 (15:55 -0400)]
inspector: Use a separate display connection
This helps isolate the inspector from some of the changes that
it can trigger. To specify a different display, set
GTK_INSPECTOR_DISPLAY to the name of the display to use for
the inspector window. If no display is specified, GTK+ will
use a separate connection to the default display.
Matthias Clasen [Fri, 24 Oct 2014 17:30:19 +0000 (13:30 -0400)]
Fix a problem with screen changes and csd
When a new screen is set on a window, we unrealize it, to
recreate all the resources. But we don't reset the client_decorated
flag, so realize() doesn't call create_decoration() - which makes
sense, since the decoration already exists. But the side-effect
of create_decoration() is to select the rgba visual, and visuals
are per-screen.
Fix this by looking for the rgba visual in set_screen(), and
replacing it with the rgba visual for the new screen, if necessary.
Benjamin Otte [Fri, 24 Oct 2014 19:31:15 +0000 (21:31 +0200)]
styleproperties: Move into deprecated directory
Benjamin Otte [Fri, 24 Oct 2014 19:22:43 +0000 (21:22 +0200)]
styleproperties: Deprecate
Benjamin Otte [Fri, 24 Oct 2014 19:12:44 +0000 (21:12 +0200)]
styleproperties: Remove header where it's not used
Benjamin Otte [Fri, 24 Oct 2014 19:03:36 +0000 (21:03 +0200)]
styleproperties: Remove unused function
Benjamin Otte [Fri, 24 Oct 2014 18:17:11 +0000 (20:17 +0200)]
css: Disallow bypassing value computation
This special code was added back in the days when computation wasn't
idemptotent. These days it is.
Also, the bypass code path is only used in fallback code that is pretty
much unused.