The new css tree may change the order of selectors (keeping the
same semantics). This affects how the selectors are printed later,
which causes some css parsing tests to not match the references.
Fortunately the order is consistent between runs given the same
css, so we just have to switch around the order in some of the
.ref.css files.
Now we use the selector tree everywhere, so there is no need to
keep around the linear selectors unless we're using them to
verify the tree correctness, so free them.
We add some "artificial" ordering to the otherwise unordered
tree nodes. This means the tree will be the same every time for the
same input. This is good because e.g. tree order affects the
reordering of the simple selectors, which may affect how
css providers are printed, which need to be consistent for
the css tests to work.
When building the tree we generally reorder the selectors inside
the same simple selector in order to pick a good first selector
to balance the tree better. However, some kinds of selectors
can't really be reordered, even thought they are simple.
This is since the matching code for some types handle
the existance of a directly preceeding selector differently:
REGION and ANY selectors look for a DESCENDANT previous
POSITION selector look for a REGION previous
From a set of GtkCssSelectors and the rulesets they match to
we create a large decision tree that lets us efficitently match
against all the rules and return the set of matched rulesets.
The tree is created such that at each level we pick the initial rule[1]
in all the considered selectors for that level and use put the
one that is in most selectors in the node. All selectors matching that
are put in the previous part of the tree.
This returns true if the matcher matches *anything*. We need
to check this later, because such matchers are dangerous in loops
that iterate over all parents/siblings since such loops would not
terminate.
gtk+ was trying to display already freed strings, leaking memory,
...I noticed this because I was getting weird blinking characters
as the status of my cups printers, and valgrind confirmed something
was wrong.
https://bugzilla.gnome.org/show_bug.cgi?id=683072
... so we don't bump a refcount whenever we get the initial singleton.
We want to use this function instead of
_gtk_css_style_property_get_initial_value() everywhere where we compute
values, because some initial values may depend on settings soon.
Resizes are queued via
gtk_widget_propagate_state()
=> gtk_style_context_set_state()
=> gtk_style_context_queue_invalidate()
=> gtk_style_context_validate()
=> _gtk_widget_style_context_invalidated()
so there's no need to queue an extra one.
Symbolic colors are an implementation detail of the CSS engine and have
been superceded by GtkCssColorValue. We don't want them clobbering the
public API. In particular because the only use I could find in the
public API is people using it to shade colors.
Make _gtk_style_provider_private_get_color() return a GtkCssValue (a
GtkCssColorValue to be exact) instead of GtkSymbolicColor.
With this, the symbolic color usage inside GTK is minimized.
The documentation for gtk_file_chooser_get_filenames() states that the
returned filenames are absolute paths, and uses g_file_get_path() to
construct the filename. The same function is used to construct the
filename in gtk_file_chooser_get_filename(), so it should also return
absolute paths.
https://bugzilla.gnome.org/show_bug.cgi?id=371034
When event capturing is enabled, stop propagating scroll events
at insensitive widgets, but don't handle them (don't return TRUE),
so they can bubble up again and reach their handling widgets.
Activate the "hides on deactivate" behavior for splashscreens,
torn-off menus, utility windows, tooltips and notifications: when
another application is brought to the front, these windows are hidden
so as not to obscure it. This is the expected behavior for
application-specific floating windows on OS X.
(cherry picked from commit 0596f5591f)
Render a background with gtk_render_background() in draw() instead.
Note that we still use gtk_style_context_set_background() for the header
window.
https://bugzilla.gnome.org/show_bug.cgi?id=688744
Make it so that the repeating vs normal test only uses sharp color
cutoffs instead of real gradients. That removes rounding errors and
makes the test pass.
so GtkMenu works properly. This is not right, but not more
wrong than always sending GDK_NOTIFY_ANCESTOR either.
(cherry picked from commit 35a9322e45)
Based on a patch from Paul Davis, inject synthetic enter events directly
into the Quartz event stream, instead of trying to synthesize them in GDK.
This seems to magically fix most combo box popup weirdness, I guess
some code is relying on a specfic order of events, or any other state
imposed by the "proper" code path of events coming in the usual way.
The patch also removes _gdk_quartz_events_send_enter_notify_event()
which is now obsolete.
(sortof cherry-pixked from 979e5061a0
but needed manual editing because GdkQuartzWindow.c was renamed
and apparently earlier patches not picked correctly/completely)
so they can appear on top of popup menus. Also, reorder the switch()
statement in window_type_hint_to_level() so it resembles the stacking
order, to avoid confision like this in the future. Fixes bug 688512.
(cherry picked from commit 1a2509a6ab)
Don't try to handle button press events on the window frame, they
have out-of-window coordinates. Also, break grabs on such events
so popup menus go away.
Patch from Kristian Rietveld, fixes bug 684419.
(cherry picked from commit 43e1354b71)
This reverts the size_allocate removal from commit
8449e05865. That code was using
_gtk_window_set_allocation() instead of gtk_widget_set_allocation(). And
that broke glade.
We can set for_size to -1 earlier than we did. Doing so makes sure we
only cache one value (as we should in the first place). In GTK 3.6, this
worked properly, but with Previously, this check was moved further up to
avoid interacting with size groups. But after recent refactorings, size
groups are handled way earlier anyway.
... instead of GtkSizeGroupMode. Orientation is what we're interested in
after all. When we need a GtkSizeGroupMode, we can do the translation
where we need it.
which does not really have a different effect than the previously
used NSPopUpMenuWindowLevel, but is what all code examples I found
are using, and it does make more sense.
(cherry picked from commit 47f0e3f1e1)
Application code can set shortcut folders that are already bookmarks.
This code causes the bookmarks to be refreshed after the shortcut is
added removing any possible bookmark duplicates
https://bugzilla.gnome.org/show_bug.cgi?id=577806
Expose GtkEntry icons as child accessibles of a GtkEntry, and provide
actions to simulate clicking them. Also, refactor the a11y children test
slightly to add a test.
https://bugzilla.gnome.org/show_bug.cgi?id=686347
The bitmasks with the 31st, 32nd and 63rd bit set are added. The make up
the largest bitmasks on 32bit/64bit that can be represented without
allocating and the smallest bitmask on 32bit that must be allocated.
With the fix in 77912a65e2, another bug
got visible: booleans are 32 bits, so if the intersection between the 2
bitmasks happened in higher bits, the return value would be truncated to
FALSE.
This actually made slider handles disappear, so it was pretty visible.
BIG NOTE: We fail on some of these to give the colors defined in the CSS
specs. This is not good, but I'm not sure how to best fix it.
For those cases, I've kept the correct color in the CSS file but added
the correct one next to it.
If lookup->missing is empty we don't need to continue looking.
We short circuit in several places as this can happen
after iteratively makeign lookup->missing smaller.
We need to use the allocated codepath if *any* argument is
allocated, not if one arg is not allocated.
This bug caused unnecessary calls to _gtk_bitmask_is_allocated,
as well as return completely wrong result if both bitmask are
allocated.
What is this bin doing with all these crazy deltas? Company does:
<Company> that can safely be removed
<Company> in general, code that isn't obvious can either be understood
<Company> with a bit of thinking or it can be removed
<Company> if in doubt, go for the 2nd of those :)
Most GtkBin subclasses override this strange garbage anyway, so it's
not like this code is ever *run*, per se. Just make it proxy directly
to the child, and hope nothing goes wrong.
Implement get_preferred_width, get_preferred_height, and size_allocate.
This allows GtkBin subclasses to be quick and easy, without the
author doing the subclassing to have to do much work.
If the "wider" label is the smaller one, use the wider size for both
cases. This can happen when ellipsizing a single character, which is
often smaller than the ellipsizing glpyph(s).
Functions should not have a space before the opening parenthesis. So
change output like
alpha (@color, 0.5)
to
alpha(@color, 0.5)
and do the same for "shade" and "mix".
Tests have been updated accordingly.
This is so newer versions of those libraries don't cause more warnings
with a stable GTK version.
We don't ever want to turn off deprecation warnings for master however,
because that's where we get rid of deprecated API we use.
Note that only glib allows use to easily do this, so nothing is done for
Pango, gdk-pixbuf or Cairo here.
With ellipsizing, the ellipsized text can have a smaller height than the
non-ellipsized text. So the wider text is also higher. Example:
.<big>TEXT</big>
will ellipsize to the small text.
Reported-By: Rico Tzschichholz <ricotz@t-online.de>
We must make sure to remove the weak pointer when disposing the widget
or when resetting the align widget otherwise glib will try to nullify
invalid memory.
This way we don't need a marker on GtkWidgetParivate that needs to be
unset later, so we have all our data in the same place and can avoid
problems with reentrancy and shenanigans like that.
But the main reason I wrote that is cleaner code.
With this function now available, we can do size computation in 2
ways:
(1) Compute size with size groups
(2) Compute size without size groups
And have (1) use (2) instead of setting flags on widgets. This patch
does exactly that.
With size groups now doing hfw, doing the optimization for CONSTANT_SIZE
was done too early. Size groups need to know that it's a hfw request, so
the other widgets in the size group get the correct behavior.
The label code assumed that Pango treats this as "wrap to as much space
as possible and then ellipsize all the lines", but for Pango, ellipsize
takes precedence over wrap. So do the same thing in GtkLabel.
Also updated is the reftest that checked this behavior.
Get rid of all the event boxes in this test. Event boxes need GDK
windows which cost a lot of performance when running the test and they
clip the label output.
Getting rid of the clipping also shows 2 bugs in this test that weren't
visible before. Those will be fixed in a followup patch.
We compute on-demand for size groups anyway, so we can (in theory, this
patch doesn't do that yet) get around costly cache blowing when
invalidating single widgets of a size group this way.
The current approach of using gtk_widget_get_mapped() is broken:
The usual steps taken when showing a window are:
(1) request the sizes
(2) allocate the sizes
(3) show the window in the allocated size
Showing the window with a random size between steps (1) and (2) would of
course
result in extra work and potential flickering when the widgets get
resized to
their proper sizes.
However, as GtkSizeGroup::ignore-hidden uses gtk_widget_get_mapped() to
determine visibility for a widget, the following will happen:
(1) the widget will request a 0 size
(2) the widget will be allocated a 0 size
(3) the widget will be too small when it is shown
gtk_widget_get_visible() however is set in advance. Note that toggling
visibility also causes a gtk-widget_queue_resize() call already so we
take care of changes in here automatically.
Instead of only checking the ignore_hidden flag when getting the
preferred sizes, respect it already when constructing the list of
widgets. This way, widgets don't queue resizes for groups they're
ignored in anyway.
For loops to loop over lists look nicer and actually do the right thing
with "break" and "continue" statements. So they are vastly preferred to
while loops.
This simplifies code and because sizes are cached by the widgets
themselves, it's not a large performance problem (unless people use huge
amounts of widgets in a single size group, but who does that?
The main problem is that we were emitting the row-deleted signal for the model in the middle
of the process that actually deletes the row from the model (remove the row from the array,
update the model->file_lookup hash table, etc.). In the model's caller, one of the row-deleted
callbacks was requesting an iter, which caused the model to revalidate itself - but it did
this while it was in an inconsistent state. This led to an assertion failure later when the
model resorted itself.
The fix in remove_file() is like this:
* The filteredness/visibility of the deleted node is not updated. The
node will simply be gone; we don't need to update those values at
all.
* We invalidate just the node that is being deleted.
* The model->file_lookup hash table is not completely nuked; instead,
we carefully adjust its indices.
* The row-deleted signal is only emitted at the very end, when
deletion is complete and the model is consistent.
Many thanks to William Hua for doing the detective work on this bug!
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
They were in the semi-public API of GtkFileSystemModel, but never actually used outside of it.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This is a function internal to the file system model; let's not pollute the gtk_tree_path namespace.
Also, make the 'i' variable into 'r' as it refers to a row index, not a file-array index (for
consistency with the docs and the rest of the code).
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
Right now we support loading and recoloring symbolic GFileIcons, but
only if the underlying GFile has a local path. This breaks when the
GFileIcon is loaded from a GResource, which is a reasonable option for an
application that wants to ship a custom symbolic icon.
This patch changes GtkIconInfo to store a GFile together with the file
path, and changes the symbolic icon lookup code to use the GFile URI,
which transparently makes the code work also for GResources.
https://bugzilla.gnome.org/show_bug.cgi?id=687059
Move variable initialization outside the first code with side effects.
This allows adding some more early returns, including one for code that
used to trigger g_return_if_fail() in certain corner cases.
Old code tried to use the "background-image" proeprty for setting the
default image background. While this used to work in the early days of
GTK3, today it is grossly misleading as the backgronud image may be
resized, repositioned and semi-translucent which causes very weird
artifacts when rendering.
So we use the background-color only instead.
This way we create one provider per settings object instead of stuffing
it into a global unchanging never-deleting hash table.
Also, we now reload the theme when instructed instead of keeping the old
loaded (and possibly stale) data forever.
https://bugzilla.gnome.org/show_bug.cgi?id=683896
This makes sure the full theme loading logic resides in one function and
isn't scattered around.
As a side-effect, the hash table kept by gtk_css_provider_get_named()
will now be populated with fallback themes. This will not be a problem
after the next commit though.
Split maintaining the global themes hash table and the theme loading
code into two functions.
This also fixes leaking the provider when loading a theme from a builtin
resource.
libxkbcommon has had some changes to its API. However, it now has a
stable release (0.2.0), so this makes the necessary changes, and
replaces all uses of the deprecated API.
Signed-off-by: Ran Benita <ran234@gmail.com>
Themes may want to render handles differently depending on whether
the widget is in selection mode (2 handles enclosing a selection) or
cursor mode (one handle pointing out the insertion cursor).
This improves both interaction and theming, as it allows
arbitrary handle shapes while just being draggable from
the visible areas.
This way themes can set up handles with the hotspot visually
displaced from the horizontal center, as long as the hotspot
lies centered in the image/svg asset.
The check on the handle to be drawn on the mask was based on the yet to
be set priv->windows pointers, pass explicitly the handle position to
have the shape correctly initialized on non-composited environments
The GtkNotebook drag-motion event handler may install a timeout when
hovering over a tab, in order to switch to it.
On the other hand it's desirable for applications to use the empty tab
area as a drop target, so the drag-motion handler returns FALSE
(also in case it installs the switch tab timeout), as explained in
https://bugzilla.gnome.org/show_bug.cgi?id=350665.
Unfortunately, applications can use the tab label widget (or a child
of it) as a different drop target area, and install their own
drag-motion handler there.
In this scenario, the timeout will still be installed by GtkNotebook's
handler, but since it returns FALSE, it will never get the matching
drag-leave event, causing it to trigger also when the mouse pointer
moved elsewhere before it expired.
Fix this by returning TRUE from drag-motion when the event is over a
tab. Note that this makes automatic tab switching not work anymore when
drag and drop is handled in the tab label widget; applications are
expected to also handle tab switching if desired in such a case.
https://bugzilla.gnome.org/show_bug.cgi?id=684415
GtkScrollbar used to rely on style-updated being emitted every time
after the widget was created in order to set the right values from its
style properties on GtkRange.
Nowadays we try to be smarter and avoid emitting style-updated at
creation time, so we need to manually initialize the GtkRange values.
This fixes a regression from 35e36b9fe5.
https://bugzilla.gnome.org/show_bug.cgi?id=686280
Currently we use gtk_style_context_set_background() when the state flags
change in order to propagate the background color to the overshoot
window, but this is actually only needed because the window doesn't get
expose events, since we always draw a full background in draw().
This also fixes some problems when the GdkWindow of the scrolled
window's child is composited, as seen in oxygen-gtk3.
https://bugzilla.gnome.org/show_bug.cgi?id=686265
The implementation of transition for GtkCssShadowValue can return NULL
at least when the two values have a different inset; all other parts of
the GTK/CSS machinery (e.g. GtkCssArrayValue) handle this by returning
NULL too. Instead, GtkCssShadowsValue was returning an invalid value,
where "len" was set, but some values in the array were NULL, which would
lead to a segfault when this value is later evaluated by the compute
function.
Fix this by making GtkCssShadowsValue return NULL if a shadow transition
fails, like GtkCssArrayValue does.
https://bugzilla.gnome.org/show_bug.cgi?id=686013
Don't use a repeating linear gradient, since it can't be easily
reftested against a non-repeating one for the reasons described in the
test header.
Instead, add a separate test for repeating gradients (against another
repeating gradient).
This makes the test pass, so it can be added to the Makefile now.
Parsing a shorthand background property was running into unexpected
errors when trying position values where there were none. To fix this,
introduce a try_parse variant of the position parse function that
silently returns NULL.
Move instance fields to a private struct, in preparation
for installing a11y headers.
This also required removing access to GtkWidgetAccessible innards
from several accessible implementations.
Move instance fields to a private struct, in preparation
for installing a11y headers.
This also required removing access to GtkToplevelAccessible innards
from the GtkWindowAccessible implementation.
Move instance fields to a private struct, in preparation
for installing a11y headers.
This also required removing access to GtkRendererCellAccessible innards
from various cell accessible implementations.
Move instance fields to a private struct, in preparation
for installing a11y headers.
This also required removing access to GtkContainerCellAccessible
innards from the GtkCellAccessible implementation.
Move instance fields to a private struct, in preparation
for installing a11y headers.
This also required removing access to GtkContainerAccessible innards
from the GtkMenuItemAccessible implementation.
Before we used a window's background color, which resulted in corrupted
display in some cases, presumably because we didn't reset the active
pattern. This patch seems to eliminate the observed corruption.
(cherry picked from commit 0e42cf81f1)
This avoids a case where the display has been opened, but calling
gdk_display_get_default() in the callback doesn't work.
Reviewed-by: Benjamin Otte <otte@redhat.com>
... instead of from the intrinsic value. This way, we respect running
animations.
Note that the concept of "reversing" transitions is not implemented yet.
Otherwise, that value will never get reset and remain frozen in time.
This is problematic for example when the value is inherited and the
parent changes the value.
When positioning the scrollbar we were doing several miscalculations
when accounting for CSS paddings and borders. This also fixes a number
of problems with RTL and when scrollbars-within-bevel is FALSE.
https://bugzilla.gnome.org/show_bug.cgi?id=685449
I'm adding a bunch of fixes for gcc complaining about
-Wmissing-declarations after finding a bunch of cases today where I
had forgotten to make functions static in the CSS code.
This patch fixes the tests in gtk/tests.
After this last patch, the gtk/ subdir should now compile without
warnings when this flag is enabled.
This is part of a bunch of fixes for gcc complaining about
-Wmissing-declarations.
It puts functions into headers and includes those headers both where the
functions are defined and where they function are used.
Also remove the starting underscore from function names where
appropriate, as those functions are static now and not exported anymore.
This is part of a bunch of fixes for gcc complaining about
-Wmissing-declarations.
I'm adding a bunch of fixes for gcc complaining about
-Wmissing-declarations.
This set of patches makes private classes in gtk/*.c that use
G_DEFINE_TYPE() safe by adding definitions for the get_type() function
that can't be made static.
I'll add a bunch of fixes for gcc complaining about
-Wmissing-declarations after finding a bunch of cases today where I had
forgotten to make functions static in the CSS code.
A thorn in those patches is G_DEFINE_TYPE() which doesn't allow making
the get_type() function static, so I added definitions for that function
above the G_DEFINE_TYPE().
After those patches, GTK should compile without warnings when this flag
is enabled.
It seems we missed updating this since GTK+3, widgets cannot be
allocated less than the size they requested in thier request
phase, and explicit sizes are used only to grow the size request.
This is intended mainly to speed up the current situation with spinners
on debug kernels. Because we now don't use a cross-fade to draw the
transition but instead have a real gradient that we draw, we don't need
to use the slow cross-fade code.
https://bugzilla.gnome.org/show_bug.cgi?id=684639
We need to store the border widths independant of them being set to 0 by
border styles, because otherwise we'd need to track that dependency and
recompute on changes, and I don't want to add more entries to
GtkCssDependencies just for this special case.
By moving the code that does the setting to 0 from the compute stage to
the query stage, we can achieve this.
Now we need to just be aware that the actual value stored is not set to
0 when we use gtk_css_computed_values_get_value().
By calling XSync in _gdk_x11_display_after_process_all_updates we
effectively make gdk rendering sync, which avoids problems with the
client animations running faster than the Xserver rendering, thus
filling up the X rendering pipes and essentially "locking up" the
Xserver (i.e. you can't even close the offending window because the
WM is starved too).
I verified this worked by making GtkSpinner paint multiple times on my
intel driver (which has some issue making this rendering slow atm),
and without this patch i get severe lag where even window dragging
stops for 5 seconds when i drag the mouse around. However, with the
patch everything is smooth.
https://bugzilla.gnome.org/show_bug.cgi?id=684639
GDK_WINDOW_XID() has the side-effect of turning a window native;
this in turn can have unexpected effects such as black backgrounds.
Avoid this by using the XID of the toplevel.
https://bugzilla.gnome.org/show_bug.cgi?id=682395
Otherwise the evil widgets that don't chain up their map and unmap
vfuncs will not get updated style contexts. This is in particular true
for GtkWindow and the CSS Theming / animated backgrounds demo in
gtk-demo.
Here's the shortest description of the bug I can come up with:
When computing values, we have 3 kinds of dependencies:
(1) other properties ("currentColor" or em values)
(2) inherited properties ("inherit")
(3) generic things from the theme (@keyframes or @define-color)
Previously, we passed the GtkStyleContext as an argument, because it
provided these 3 things using:
(1) _gtk_style_context_peek_property()
(2) _gtk_style_context_peek_property(gtk_style_context_get_parent())
(3) context->priv->cascade
However, this makes it impossible to lookup values other than the ones
accessible via _gtk_style_context_peek_property(). And this is exactly
what we are doing in gtk_style_context_update_cache(). So when the cache
updates encountered case (1), they were looking up the values from the
wrong style data.
So this large patch essentially does nothing but replace the
context argument in all compute functions with new arguments for the 3
cases above:
(1) values
(2) parent_values
(3) provider
We apparently have a lot of computing code.
The garbage would be visible if any widget enabled the toplevel
NSView's CALayer in order to do custom native rendering.
(cherry picked from commit 92ea94af5f)
This is needed for the SELECTION_NONE mode where nothing is ever
selected, but its also needed for CTRL-<key> keynav that moves the
focus without changing the selection.
https://bugzilla.gnome.org/show_bug.cgi?id=684984
Currently the GdkWindow used for dragging is created once when
the first drag starts, and the reused identical each time.
Instead, just recreate it for each drag, with the correct size.
This reverts commit f2cb8f1270.
The patch actually didn't work for at least text. I currently have no
clue why, but I suspect it requires investigating Cairo code and
recording surfaces, and I'll not do that right now.
Split out the blurred shadow rendering in three steps:
- creation of a surface of the appropriate size - we use the clip
rectangle as a good measurement for the size, since we won't render
out of it anyway
- painting the unblurred shape on the surface - this is responsibility
of the single shadow implementations
- blur the surface and compose the result back on the original cairo_t
This means we can share code between the implementations for the first
and third steps; it also makes the code independent of the rendered
size, so we can avoid passing down a cairo_rectangle_t with e.g. the
icon coordinates.
The code accesses pixels in a chunks of 4 bytes, so we must only support
formats where the size of a single pixel is 4 bytes.
Fix RGB24 to be 4 bytes (the alpha channel is ignored) and disallow A8.
We were adding one child too much to the style context path when
generating it for the internal buttons, which in turn caused sibling
selectors from the theme such as :first-child to apply to both buttons
under certain circumstances. Spotted by Lapo Calamandrei.
As long as we don't have an API for explicitly inverting the bar, it
makes more sense for the progress in vertical orientation to fill from
the bottom.
In the event that a GtkAccelKey was present for the closure but it
contained a keyval of 0 the previous code would show "". After the
recent adjustments, "-/-" would be shown in this case.
It turns out to be a pretty common case, so fix the logic to stop using
'0' as a magic value to mean "don't have an accel" and add a separate
boolean for that purpose.
This reverts commit 1f5dea9eba,
since it was causeing noticable behaviour changes.
Previously, GTK_DATA_PREFIX=/ ./gtk3-demo would start
gtk3-demo with the Raleigh theme. With that change, it
was starting with no theme at all (i.e. all black).
While regular animations should always be created, transitions should
not. This patch allows to express this by passing NULL as the values to
transition from.
It also adds a gtk_style_context_should_create_transitions() function
that returns TRUE when transitions should be created.
... that actually was both wrong, a performance failure and has been
there since the original checkin.
Updating the cached style data absolutely does not mean clearing all
cached style data first. There's nothing to update then.
This will be useful to not trigger updates all the time when nothing is
happening (ie due to animations being paused or due to them having
reached their final value).
This adds the GtkCssAnimation class and the code needed to hook it into
GtkStyleContext. It takes the values out of the CSS "animation"
properties and does animations. See
http://dev.w3.org/csswg/css3-animations/
for details.
Note that the code for starting and stopping animations with widget
visibility doesn't work yet.
This change is necessary because the old code did not accound for corner
cases (like translucent child windows), which could stop
gtk_widget_queue_resize() to not trigger redraws.
Make the main (and only) entry-point to gtkmodelmenu.c the now-public
gtk_menu_shell_bind_model().
Move the convenience constructors (gtk_menu_new_from_model() and
gtk_menu_bar_new_from_model()) to their proper files.
Remove the private header file.
Simplify the code a bit by making the initial populate part of the
bind() call.
https://bugzilla.gnome.org/show_bug.cgi?id=682831
Add an API to GtkAccelLabel for hardcoding the accel key to be displayed
(ie: allowing us to bypass the GtkAccelGroup lookup).
Use that from the GMenuModel-based GtkMenu construction code instead of
passing around the accel group.
This makes accel labels work in bloatpad again.
This patch effectively removes any hope of automatic runtime accel
changes in GMenuModel-based menus without additional application
support but it leaves the door open for this to be supported again in
the future (if we decide that it's important).
https://bugzilla.gnome.org/show_bug.cgi?id=683738
Add support for a stateful action associated with a submenu. The action
state is set to TRUE when the menu is shown and FALSE when it is
unshown.
This is useful to avoid unnecessary processing for menus that have
frequently-changing content.
A possible future feature is to add support for asynchronously filling
the initial state of the menu by waiting until the action actually emits
its state-change signal to TRUE before showing the menu.
A silly example has been added to Bloatpad to demonstrate the new
feature.
https://bugzilla.gnome.org/show_bug.cgi?id=682630
In gtk_menu_bar_draw, the check for shadow type != none
disables rendering of the background instead of the frame.
The check should be moved down to gtk_render_frame.
Patch by Peter de Ridder,
http://bugzilla.gnome.org/show_bug.cgi?id=670390
A button is highlighted if the private variable in_button is TRUE.
This variable is set when the pointer is over the button and cleared when
it left the button. When a button is hidden while there is the pointer over
it, GTK generates a leave notification event, in_button is set to FALSE.
But when a button is removed from a container but not destroyed, it is
unrealized and loose its window. It cannot receive the leave notification
event and in_button stay TRUE. So when the button get a new parent it is still
highlighted.
https://bugzilla.gnome.org/show_bug.cgi?id=676890
Scroll to the selection when setting it so the selected font is
visible on screen. This is especially useful if an initial font is
set for the user to see it.
https://bugzilla.gnome.org/show_bug.cgi?id=684156
Previously, we would avoid setting the prelight state flag when
button_down was TRUE and draw_indicator = FALSE, which is the normal
case of a GtkToggleButton during a mouse press.
It looks like this behavior was introduced a long time ago with commit
b94e6c0a80. I believe the reason was that
a widget in GTK2 couldn't have more than a single state (e.g.
hover+active) at a given moment.
https://bugzilla.gnome.org/show_bug.cgi?id=684038
Apply patch from Kristian Rietveld which addresses two issues
in gdkeventloop-quartz.c:
This patch moves the autorelease pool drain and introduces protection against
the invalidated ufds. Basically, when we suspect ufds has been invalidated by a
recursive main loop instance, we refrain from calling the collect function.
(cherry picked from commit 79b3326eaa)
After my recent fix for this, nautilus was still having problems
telling keeping F10 and Shift-F10 apart. With this change, we are
treating levels with the same symbol like inactive levels, ignoring
them entirely.
The widget-factory was pretty much overflowing, so I've
made it page, and started to fill the second page with
vertical spin buttons. New examples and widgets should
be added to page 2 now.
Be a bit more careful in get_pango_attr_list() and
get_utf8_preedit_string() to ensure that the client_window is properly
created before proceeding, to avoid access violation/segfault crashes on
Windows with IME installed, especially when running the pickers demo.
https://bugzilla.gnome.org/show_bug.cgi?id=682919
Don't hook on the widget style context and set up instead
a widget path for itself. Also use a common style class
for both handles, with an extra top/bottom class for each
handle.
This is to allow animating arrays properly. I'm not really thrilled
about this solution (we leak propertys into the values again...), but
it's the best I can come up with - I prefer it to having N different
array types...
I want to get away from the ability to have 0-length arrays, all css
arrays are single element.
Even if the element is "none", it is still a "none" element.
GtkTextHandle is used to indicate both the cursor position
and the selection bound, dragging the handles will modify
the selection and scroll if necessary.
Backwards text selection is also blocked for touch devices,
so the handles don't get inverted positions and possibly
obscure portions of the selected text.
GtkTextHandle is used to indicate both the cursor position
and the selection bound, dragging the handles will modify
the selection and scroll if necessary.
Backwards text selection is also blocked for touch devices,
so the handles don't get inverted positions (This is more
important though on GtkTextView, as inverted handles may
obscure portions of the selected text, good for consistence
though)
This is a helper object to allow text widgets to implement
text selection on touch devices. It allows for both cursor
placement and text selection, displaying draggable handles
on/around the cursor and selection bound positions.
Currently, this is private to GTK+, and only available to
GtkEntry and GtkTextView.
GtkTextHandle creates temporary override redirect windows, but still
hook to the text widget for events, so those are effectively captured
by GtkScrolledWindow if a text widget is within it
A change in xkeyboard-config 2.4.1 made it so that function keys
now have a shift level which has the same symbol, but 'eats' the
shift modifier. This would ordinarily make it impossible for us
to discriminate between these key combinations.
This commit tries harder to discriminate in 2 ways:
- XKB has a mechanism to tell us when a modifier should not be
consumed even though it was used in determining the level.
We now respect such 'preserved' modifiers. This does not fix
the Shift-F10 vs F10 problem yet, since xkeyboard-config does
not currently mark Shift as preserved for function keys.
- Don't consume modifiers that do not change the symbol. For
the function keys, the symbol on the shift level is the same
as the base level, so we don't consider Shift consumed.
For more background on the xkeyboard-config change, see
https://bugs.freedesktop.org/show_bug.cgi?id=45008https://bugzilla.gnome.org/show_bug.cgi?id=661973
Showing mnemonics immediately on modifier press can be annoying and
distracting when the user is just trying to Alt+Tab into another
application/window since the mnemonic will show up and quickly vanish
again when we receive the focus out event.
https://bugzilla.gnome.org/show_bug.cgi?id=672431
The file type radio group has a mnemonic on its label, but
activating it did not work, since GtkPrinterOptionWidget did
not know to forward the activation to one of the buttons.
* This patch gets rid of the separated
fields for selecting a print to file
target by removing the folder selection
button and the entry. It is replaced by
a browse button, which opens a file
selection dialog, that can select both
the path AND the filename.
* If the filename is relativ to the home
folder it will substitute ~/ instead of
the home folder. Additionally if the
resulting filename is longer than 30
characters, it cut of the first part
and replace it by '...' so that
the button text never gets too long.
https://bugzilla.gnome.org/show_bug.cgi?id=682129
The file chooser is asynchronous, so doing 'select_file (old_file)' and subsequently querying
the file for updating the labels is not going to work. However, the underlying file chooser
will emit 'selection-changed' as appropriate when it finishes restoring the old file. So,
we only need to update the labels when the file chooser dialog is confirmed, not cancelled.
This commit moves all the entry completion implementation
into gtkentrycompletion.c. It also gets rid of an unnecessary
completion_device member in GtkEntryPrivate.
When compiling gtk on Win32 then the file gtkdbusgenerated.c also needs to be
compiled and linked into the gtk library as it's needed for GtkMountOperation
https://bugzilla.gnome.org/show_bug.cgi?id=682825
- don't poke at the children's background pattern at draw time, but just
call gtk_render_background()
- we should propagate rendering of the background to the overshoot
window when the state flags or the style changes, or it won't respond
to e.g. focused/backdrop changes correctly
https://bugzilla.gnome.org/show_bug.cgi?id=682854
GTK_CSS_DEPENDS_ON_EVERYTHING was used as a placeholder when
implementing dependencies. Now that dependencies are completely
implemented, it's no longer necessary.
... in the case where no change of the DOM tree actually happened.
We don't do anything yet with that information, this patch just
correctly computes it.
When values are computed, they might depend on various other values and
we need to track this so we can update the values when those other
values change. This is the first step in making that happen.
This patch does not do any dependency tracking at all, instead it uses
GTK_CSS_DEPENDS_ON_EVERYTHING as a sort of FIXME.
Both _gtk_css_style_property_print_value() and
_gtk_css_style_property_compute_value() aren't necessary anymore and are
replaced by _gtk_css_value_print() and _gtk_css_value_comptue()
respectively.
This gets rid of the public function
_gtk_css_rgba_value_compute_from_symbolic().
The fallback is now handled using a switch statement instead of letting
the caller pass the function.
This is a reorganization of how value computing should be done.
Previously the GtkCssStyleProperty.compute vfunc was supposed to take
care of special cases when it needed those for computation. However,
this proved to be very complicated in cases where values were nested and
only the last value (of a common type) needed to be special cased.
A common example for this was the fallback handling for unresolvable
colors.
Now, we pass the property's ID along with all compute functions so we
can do the special casing where it's necessary.
Note that no actual changes happen in this commit. This will happen in
follow-ups.
This commit is essentially a large reorganization. Instead of all value
subtypes having their own compute function, there is the general
_gtk_css_value_compute() function that then calls a vfunc on the
subtype.
This hint may be used in text widgets to inhibit their
input methods. The most compelling usecase is calculator
applications, which already have a builtin and better
suited onscreen keyboard.
https://bugzilla.gnome.org/show_bug.cgi?id=651244
This improves the layout of the mount operations dialog to:
* Make primary labels bold in all cases
* Lay out the widgets in a grid
* Put space between the two radio groups to distinguish them
https://bugzilla.gnome.org/show_bug.cgi?id=682552
We no longer support modifying GdkWindow hierarchies during
expose events. This is not working anymore anyway as the
flush operation now does not push already rendered pixels
in the flushed window from the double buffer to the window.
https://bugzilla.gnome.org/show_bug.cgi?id=679144
Avoid copying back partially drawn double-buffer data
when flushing to avoid flicker. This means non double
buffered widgets must draw opaque pixels in its expose
handlers, and that you are not allowed to use direct
rendering (or modify GdkWindow pos/size/order) from
inside the expose handler of a double buffered widget.
See https://bugzilla.gnome.org/show_bug.cgi?id=679144 for more
details
gtk_widget_insert_action_group (widget, "foo", NULL) is valid, but
g_action_muxer_insert (muxer, "foo", NULL) is not. Use
g_action_muxer_remove() for that case.
The code was calling _gdk_window_ref_cairo_surface in a few places
where the intent was not to read/write to the surface, but just look
at its type (to e.g. create a similar surface). This is bad, as that
operation causes a flush which may cause unnecessary work and/or
flashing. Instead we just get the impl surface in these cases.
get_time_from_ns_event(): apply patch from Michael Hutchinson which
makes sure the returned guint32 wraps correctly on 32 bit machines
when the uptime exceeds 2^32 ms.
First, ensure we always consume floating values, as documented.
Second (and more serious), don't try to query the action if the
action name is not set yet. This will cause crashes...
It's too close to the release so we'll keep _set_menu() around, but
deprecate it, pointing to the new API. It will be removed (and the name
reused) before the 3.6.0 release.
https://bugzilla.gnome.org/show_bug.cgi?id=682235
If a section or submenu item has a "action-namespace" attribute, the
action names of the created GtkModelMenuItems will be prefixed with that
namespace. Namespaces can be cascaded.
There are no remaining users of the GActionMuxer in GtkApplicationWindow
because they've all been ported over to using the one on GtkWidget (via
GtkActionHelper, for the most part).
There are no remaining users of the GtkApplicationWindow API to create
GSimpleActionObserver or to get the GActionObservable (ie: muxer) for
the appwindow. Drop those APIs.
Drop the explicit passing of GActionGroup into the GtkMenu(Bar)
constructors and operate from the action context instead.
With GtkMenuItem implementing GtkActionable, this turns out to be pretty
easy (and most of the code can be removed from GtkModelMenuItem,
including the GActionObserver implementation).
Each GtkWindow with an associated GtkApplication should add this as
"app" to its action context. Each GtkApplicationWindow is its own
GActionGroup, and it should add itself to itself with the prefix "win".
There is now some duplication here because we have the new GActionMuxer
hierarchy managed by GtkWidget, but GtkApplicationWindow still carries
its own muxer. The redundancy will be removed in a future patch.
The current process of implementing GActionObserver is annoying and the
GSimpleActionObserver interface leaves a lot to be desired. Introduce a
new class, GtkActionHelper that gives you pretty much everything you'd
want to do as an implementor of GtkActionable.
The GtkActionHelper also features an "application" mode that is not
associated with a particular GtkWidget but rather with whatever widget
happens to be the active window of the given GtkApplication at a
particular point in time. This will be useful for the Mac OS menubar.
This allows adding a GActionGroup with a given name at an arbitrary
point in the widget tree.
This patch also adds an internal _get_action_muxer() API. Calling this
will create a GActionMuxer associated with the widget. The parent of
the muxer will be the muxer of the widget's conceptual parent. For
non-menus, that is the normal parent. For menus, it is the attach
widget.
In this way, we end up with a hierarchy of GActionMuxer that largely
reflects the hierarchy of GtkWidget, but only in places that the action
context has been requested. These muxers are the ones on which the
inserted actions groups are installed.
A following patch will add a user of this API.
If a muxer does not contain an action group with the given prefix, chain
up to the "parent" muxer to look for it.
This initial implementation is rather inefficient. It will lead to
changes on action groups associated with parent muxers being broadcast
to all children (regardless of if anybody there is interested or not).
An optimised version will follow soon.
-Include fallback-c89.c for the usage of round(), where an implementation
of round() is provided for compilers that don't have it
-Use g_ascii_strtod() instead of strtof as strtof() may not be universally
available.
-Turn on Whole Program Optimization for all Release builds.
-Disable Incremental Linking for all Release builds.
-Use MultiByte character set for all configurations for consistency.
-Enhance optimization by turning on WholeProgramOptimization for all
Release builds
-Disable IncrementalLinking for all Release builds
-Make sure we are using MultiByte character set, to be consistent across
the board for all configurations
Call _gtk_entry_get_display_text()from gtkentryaccessible.c to make
sure we always consider the actual text being displayed when
implementing the functions from the AtkTet interface.
https://bugzilla.gnome.org/show_bug.cgi?id=681591
Translating "ON" as "EK" (verb prefix for beginning of an action) and
"OFF" as "FOR" (away) is suboptimal. Just use the suggested 1/0 glpyhs
instead.
The previous translator was clearly insane.
These widgets have ancestors other than GObject which could eventually
implement the notify vfunc for their properties. For correctness, they
should chain up the notify vfunc.
https://bugzilla.gnome.org/show_bug.cgi?id=673478
While shadow-type *properties* can make sense, to opt-out of the
padding/border machinery programmatically, having it as a style
property doesn't make any sense, since we have a better way to change
the bevel style from the theme already.
This commit deprecates the shadow-type style property in GtkToolbar.
This is a regression from commit
d0d21a4f00.
We are requesting the CSS padding twice: once unconditionally and
another time if SHADOW_TYPE != NONE, which is usually the case.
gtk_widget_path_copy() currently calls g_array_append_val() in a loop,
which is inefficient due to reallocating the array's memory. Calling
g_array_set_size() before entering the loop reduces the number of CPU
cycles used by roughly 30%.
Patch by John Lindgren,
https://bugzilla.gnome.org/show_bug.cgi?id=679978
Not defining these macros at all causes harsh build breakages.
Better to leave them defined (but documented as deprecated) for now.
Everybody will still get the deprecation warnings for the underlying
gdk_threads_enter/leave.
We can hide the macros again later on when the world has had some
time to port off GDK threads.
The widget is already calling gtk_render_frame, but is not measuring css
border and padding when negotiating its size. This patch replaces the
already existing get_internal_padding static helper with a function that
sums the old internal-padding value with the values specified via css.
Use $(AM_V_GEN) for generating man pages, and set some parameters
for the XSL stylesheets. Among other things, don't generate AUTHORS
and COPYRIGHT sections.
The story is slightly different for applications vs libraries;
make it clear that libraries should continue using the lock so
we don't break applications that haven't been ported to the
'single thread' model yet.
When the tab label gets removed from the notebook on widget
desctruction, we should still unconditionally unparent it from the
notebook, since failing to do so will keep a stray reference alive.
In case applications rely on the tab label being destroyed to release
other references (e.g. because the tab label is a custom object, or
another object's lifecycle is tied to it using g_object_set_data_full()),
this will also possibly cause other references to get leaked.
In Nautilus, the result was we were failing to release the reference to
a NautilusWindowSlot, and other parts of the application relied
on it being destroyed at a specific time instead, causing the
application to crash when closing a window.
This is a regression from commit
325cf071d1.
This commit restores the previous unparenting behavior in case we're not
in a DnD operation.
https://bugzilla.gnome.org/show_bug.cgi?id=680349
We need to do this here so that an explicit gtk_widget_destroy
on either the widget or a container that holds it will kill the
reference to the menu or model.
https://bugzilla.gnome.org/show_bug.cgi?id=680803
This commit deprecates gdk_threads_init, gdk_threads_enter,
gdk_threads_leave and gdk_threads_set_lock_functions. Using GTK+
from multiple threads does not work at all on Windows, and is
problematic on other platforms as well. We want to move to a world
where all GTK+ calls are made from the main thread.
Use g_main_context_invoke, g_idle_add and related functions if you
need to schedule GTK+ calls from other threads.
http://bugzilla.gnome.org/show_bug.cgi?id=680754
These are just wrappers for the functions, and we want to
deprecate them. Stopping to use them internally is a good
first step. Also define GTK_COMPILATION so we can keep using
gdk_threads_enter/leave without causing deprecation warnings.
Remove the "-win32-" from the output file names for the GDK and GTK+ DLLs,
like what is now done for quite a while on other platforms
(and MinGW builds), for consistency reasons. This is due to GDK/GTK+
are buildable with multiple backends.
Note: For references, the Windows build only builds the Win32 backend
for the time being.
If we don't have a wl_seat - because a grab hasn't been initialised by GTK+
then fallback to making the shell surface transient to the parent rather than
a popup surface.
We now support independent selection and primary clipboards, and avoid
wiping clipboard on modifying its contents from the same owner. This fixes
most of the interaction issues with clipboard and selection.
Signed-off-by: Rob Bradford <rob@linux.intel.com>
GSequence iterators point at the position between two elements so an
iterator pointing at the N tree model node is actually between the N-1
and N sequence elements. This means that asking for the previous
sequence iterator first and then checking if it is the begin iterator
would yeld true for an iterator pointing at the 2nd tree model node
and make us return FALSE mistakenly.
https://bugzilla.gnome.org/show_bug.cgi?id=679910
Review comment: I think the implementation of the vfuncs in gdkkeys-wayland.c
depend on that we're using the keysysm as the hardware keycode. I think that
needs to be evaluated for the future. But for now this patch gives reasonably
complete keyboard input.
Signed-off-by: Rob Bradford <rob@linux.intel.com>
This is then logically associated with the input device since each (keyboard)
input device has its own keymap.
Signed-off-by: Rob Bradford <rob@linux.intel.com>
Although GDK expects the keymap to be associated with the display under
Wayland this is really associated with the input device so expose this by
finding the first keyboard device.
Signed-off-by: Rob Bradford <rob@linux.intel.com>
This program launches an application specified by its desktop name
optinally taking list of URIs which are passed as arguments.
Uses GdkAppLaunchContext to get proper startup notification and
display handling for graphical apps.
https://bugzilla.gnome.org/show_bug.cgi?id=679342
The code that was trying to limit preview size changes was
running into some width-for-height pitfalls. It turns out
that the dialog behaves quite ok without this code, so just
remove it.
Now that filters may affect sensitivity of rows, we need
to clear the sensitivity column from the cache when the
filter changes. This fixes the problem where selecting a
different filter does not update the sensitivity of folders
until you change directories.
This way we remove paired function calls (compute/set pairs), and also make
it possible to avoid computing a filter twice, as setting the visibility
depends on filteredness.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
It bothers me that we call gtk_file_filter_filter(), then negate the result,
and the return *that* from node_should_be_filtered(). So, rename 'filtered'
throughout GtkFileSystemModel to 'filtered_out' to mean things that didn't
pass the filter.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This was a copy-paste leftover from node_set_visible(). Filters are not
concerned with model freezes, so node_set_filtered() does not
need to handle freezes, either.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
Specially in the case of comboboxes, those menus could enable scrolling
even if the contents could fit in the work area, and could show blank
space in order to line up the selected item with the combobox.
When such thing happens, take into account scroll_offset when relocating
the menu contents so contents don't jump directly onscreen, and apply
it so scrolling is allowed in the direction that brings the menu onscreen
and blocked in the opposite direction.
Also, wait for cancelling the scroll operation until the touch is released
even if the scrolling arrows disappeared, so the menu item underneath isn't
selected right away.
https://bugzilla.gnome.org/show_bug.cgi?id=678113
Entries don't expand vertically if they are given pixbufs larger
than the calculated height for the current font, resulting in
cropped icons, so force the pixbuf to be rescaled so it fits
on the entry allocated size.
https://bugzilla.gnome.org/show_bug.cgi?id=678087
If the symbolic icon has other size than 16x16, the embedder
SVG that overrides colors would still force that size, resulting
in clipping instead of resizing. So fetch the original pixbuf
size the first time a symbolic icon is requested for a GtkIconInfo,
and use that size for the embedder SVG so it can be scaled properly
afterwards.
https://bugzilla.gnome.org/show_bug.cgi?id=677567
This replaces the wl_input_device with wl_pointer, wl_keyboard, wl_touch all
tied together under a wl_seat.
This is quite a radical change in protocol and for now keyboard handling is
disabled.
This requires the SHM object be initialised - therefore this is the most
logical (if slightly ugly place.)
We also need to make sure that we do something clever to load the correct
cursor theme.
Both GtkListStore and GtkTreeStore had a few methods that allowed
-1 to mean 'append' when specified as a position, but others that
demanded positive position arguments. Make this consistent by
always allowing -1.
https://bugzilla.gnome.org/show_bug.cgi?id=667808
The recently-used.xbel storage for recently used files is located inside
the $XDG_DATA_HOME directory; there's no actual guarantee that the
directory has been created already, even though it's very highly
probable on any modern distribution. We should create it, along with its
intermediate parents, before constructing the file monitor that we use
to get change notifications.
https://bugzilla.gnome.org/show_bug.cgi?id=671817https://bugzilla.gnome.org/show_bug.cgi?id=667808
The builder XML description has an action for "win.parse" but the
application is looking for "win.paste". Rename the label to
"_Paste" and the action to "win.paste" in the window action XML.
https://bugzilla.gnome.org/show_bug.cgi?id=678829
nitems is never guaranteed to be defined in sys/params.h as it is meant
to be defined within a protected ifdef __KERNEL conditional.
Use G_N_ELEMENTS from GLib which is the proper way of doing this.
In gtk_entry_get_icon_pixbuf() we unreference the pixbuf that
_gtk_icon_helper_ensure_pixbuf() gives us back, since the function
doesn't return a reference, and by doing so we're able to return the
reference owned by the cached icon helper.
Since the icon helper method can return NULL though, if no icon
properties have been set on it, guard for != NULL before unreffing the
pixbuf, as that would cause a critical warning.
https://bugzilla.gnome.org/show_bug.cgi?id=679537
Right now, when we create a GtkModelMenu for a GMenuModel, we listen to
changes to the menu's attach-widget to detect when a toplevel
GtkApplicationWindow becomes available to fetch actions from it.
This unfortunately breaks this simple code:
GtkWidget *application_window = gtk_application_window_new();
GtkWidget *menu_button = gtk_menu_button_new();
GMenuModel *menu_model = get_menu_model();
gtk_menu_button_set_menu_model(menu_button, menu_model);
gtk_container_add(GTK_CONTAINER(application_window), menu_button);
Since GtkMenuButton creates a GtkModelMenu and sets itself as its attach
widget before it's added to a hierarchy containing a
GtkApplicationWindow.
Fix the bug by simply listening for changes in the window hierarchy, and
creating the menu model when the attach widget is added to an
application window.
https://bugzilla.gnome.org/show_bug.cgi?id=679454
When the menu is detached, the attach-widget property changes value to
NULL, so we should notify a property change, like
gtk_menu_attach_to_widget() does.
https://bugzilla.gnome.org/show_bug.cgi?id=679454
This is fallout from commit 257b42e2f9 -
those fields were already getting freed in
gtk_application_shutdown_x11() and my commit caused crashes on quit
instead.
Thanks to Rico Tzschichholz for reporting the bug and testing this fix.
As an addition to 1042372670, themes might
want to avoid using the hardcoded GTK defaults for legacy GtkStyle
values. Add a gtkstyle-fallback style class that can be used by themes
to tweak the legacy GtkStyle defaults.
https://bugzilla.gnome.org/show_bug.cgi?id=679092
Since we explictly call popup_menu with NULL when it's
keyboard-activated, we need to be careful and not access event->device
without checking for event != NULL before.
The signals for the action group were being disconnected when the action
group was explicitly removed from the GActionMuxer but the same was not
being done when it was finalized.
This means that a change in the state of an action group that used to be
associated with a finalized GActionMuxer would result in a crash. This
would happen for stateful application actions after closing a window.
We can't expose colors with alpha in the GtkStyle colors, since
GdkColor has no alpha. Currently we throw away alpha completely,
which fails very badly for completely transparent backgrounds,
which now is the default for most widgets, as it typically
end up with black-on-black.
We handle this by falling back on the default/previous colors for
transparent colors. This is is simple and avoids complete failure.
https://bugzilla.gnome.org/show_bug.cgi?id=671437
Don't use ASCII control characters to denote the input of Esc, Tab,
Return/Enter, Backspace and Delete, as it seems that it is not how
Windows handle them, and they cause weird characters to appear in the
input field on GTK+3 programs in non-English Windows. Instead, let
these keys be handled as-is on Windows, like what is done in GTK+-2.x.
Checked with mclasen on IRC, and thanks to the people who verified the
patch to not break anything on English Windows.
When using Shift-Tab to move the focus out of page content onto
the tab label, we end up in a situation where both Tab and Shift-Tab
move focus back into the page, which is not really what is expected
when the notebook is part of a dialog.
Instead, arrange things so that using Shift-Tab with focus on a
tab label moves the focus out of the notebook.
http://bugzilla.gnome.org/show_bug.cgi?id=669986
It turns out that we can end up removing a notebook child while
the tab is still 'detached'. Child removal causes
gtk_notebook_remove_tab_label() to be called on the tab label,
but that function did not deal with the eventuality that the tab
label may be a child of the dnd window.
http://bugzilla.gnome.org/show_bug.cgi?id=677943
Some builders using gtk3 outside of the GNOME cycle want an option to
avoid linking to atk-bridge-2.0. Provide that, and at the same time
ensure we're only looking for it on X11 platforms.
https://bugzilla.gnome.org/show_bug.cgi?id=677491
Make GMountOperation look for an owner of org.Gtk.MountOperationHandler
if possible, and use it instead of the GTK-based dialogs.
This allows applications to use the implementation offered by the
desktop shell, if available, through a DBus private interface:
org.Gtk.MountOperationHandler.
https://bugzilla.gnome.org/show_bug.cgi?id=674963
This gets the current cell area of a particular item. Its similar
to gtk_tree_view_get_cell_area().
The code is extracted from gtk_icon_view_set_tooltip_cell which now
just calls the old code.
https://bugzilla.gnome.org/show_bug.cgi?id=678418
gdk_device_list_slave_devices only makes sense to call on master
devices, yet its g_return_if_fail check made it reject such devices.
Pointed out by monty.
When a widget is app_paintable, its background should not be drawn by
the theme, thus we should not try to override its background again when
style-updated is fired.
This is a bit of a hack, but it fixes gray surfaces observed for DnD
windows with recent GTK+.
As used in Totem and gnome-contacts. The widget
takes either a GtkMenu or a GMenuModel to construct
its menu, and can be given a parent widget to use to
position the drop-down (as used in GtkMenuToolButton).
https://bugzilla.gnome.org/show_bug.cgi?id=668013
If the icon view is empty, we cannot get a reasonable size request from
the cell renderers. So all values we would compute are pretty much
useless.
So we special case it.
This also gets rid of a bunch of crashers from div-by-0 in corner cases.
https://bugzilla.gnome.org/show_bug.cgi?id=677809
This is useful for 2 reasons:
(1) Items actually exist and are clickable
(2) Size computations don't divide by 0
I've not seen problems with this in the wild (mostly because
item-padding defaults to non-0), but noticed this while fixing other
bugs.
Minimum size is necessary so you can see the item. If we can't get that
we need to scroll.
Natural as the maximum is used so that the spacing between items doesn't
increase when resizing the iconview, but empty space is added to the
right/bottom instead.
https://bugzilla.gnome.org/show_bug.cgi?id=677809
While working on the "iconview: Don't shrink items" patch I noticed that
gtk_icon_view_compute_n_items_for_size modifies the natural and minimum
item sizes it got from gtk_icon_view_get_preferred_item_size when
calculating the max number of items which will fit, but later on it
checks against these sizes when calculating the item_size, and these
checks expect these values to be unmodified.
This patch fixes this by modifying the natural and minimum values in
advance and doing all computations with modified values.
https://bugzilla.gnome.org/show_bug.cgi?id=677809
The previous code assumed that the width was always enough for more than
one column, which is obviously not correct when a number of columns is
hardcoded.
With this patch, it will now always check that the width is enough and
otherwise cause scrolling.
https://bugzilla.gnome.org/show_bug.cgi?id=677809
GtkToolbar doesn't have its own GdkWindow to draw on (it calls
gtk_widget_set_has_window(FALSE) in _init), but only an event window
(input-only).
Since gtk_widget_get_window() in that case will return the GdkWindow of
the parent container, by calling gtk_style_context_set_background() here
we're overriding the base background of the container instead of our.
While in most cases this doesn't have any noticeable effect, since
the toplevel GtkWindow will paint its background on top of it at the
beginning of the draw cycle, when the classic window hierarchy is
broken, such as when widgets are rendered through a clutter-gtk
offscreen embedding, the background will become visible, which is
undesirable.
Fix this by having GtkToolbar not call gtk_style_context_set_background
in its style_updated handler.
When I added the versioned annotation, I accidentally backdated
it, so the Deprecated: tag in the docs said 3.4, but the annotation
said 3.0. Fix it so we say 3.4 in both places.
When inline-selection is set, and the completion popup is showing,
pressing left abruptly jumps to the beginning of the entry text.
This is not expected, since the cursor is at the end of the text before
the left key is pressed, and this behavior is completely inconsistent
with how an entry would normally behave.
The behavior can be observed in Epiphany by selecting a completion match
and pressing left.
This patch changes the code so that it just runs the default entry key
press keybindings in such a case.
https://bugzilla.gnome.org/show_bug.cgi?id=677915
Mark the new_order argument as zero-terminated array, even though it does not
need to be zero terminated (it has an implicit length not given by a constant
or another method argument). It does not hurt if bindings append an extra zero
to the array as long as it has enough elements, and this makes the method
introspectable.
https://bugzilla.gnome.org/show_bug.cgi?id=677941
Remove the 'you shall not connect' message from this signal.
While it is a keybinding signal, using it from applications is
fine and, in fact, expected.
There are three bugs here:
- we should check if the value type is transformable instead of being
compatible, since that's all we care about in order to call
g_value_transform()
- the check is only meaningful in the direction
passed-in-type->column-type and not viceversa
- we should init the destination GValue to the column type before
calling g_value_transform on it, or the destination type information
will be missing and the method will fail
Thanks to Jasper St. Pierre and Colin Walters for all the help in
tracking this down.
https://bugzilla.gnome.org/show_bug.cgi?id=677649
There are three bugs here:
- we should check if the value type is transformable instead of being
compatible, since that's all we care about in order to call
g_value_transform()
- the check is only meaningful in the direction
passed-in-type->column-type and not viceversa
- we should init the destination GValue to the column type before
calling g_value_transform on it, or the destination type information
will be missing and the method will fail
Thanks to Jasper St. Pierre and Colin Walters for all the help in
tracking this down.
https://bugzilla.gnome.org/show_bug.cgi?id=677649
gtk_icon_info_load_symbolic checks for the existance of a filename parameter
so it can include it along with the stylesheet. We don't set the filename
parameter when creating the info for a GFileIcon, for some reason.
https://bugzilla.gnome.org/show_bug.cgi?id=676356
This is probably going to be replaced by something different soon, but
in the meantime, fix this annoying bug, visible e.g. in widget-factory
for left/right tab examples.
As used in Totem and gnome-contacts. The widget
takes either a GtkMenu or a GMenuModel to construct
its menu, and can be given a parent widget to use to
position the drop-down (as used in GtkMenuToolButton).
https://bugzilla.gnome.org/show_bug.cgi?id=668013
When we determine the fill rectangle for an image layer, and the image
is not repeating, we should also position the rectangle according to the
values of background-position, or we will always fill a tiny slice at
the top/left of the clip box.
https://bugzilla.gnome.org/show_bug.cgi?id=677109
Many themes want to render the trough background/stroke thinner than the
full height/width (which is constructed around the value of the
'slider-width' style property).
Read and apply the CSS margin from the theme on the trough component, so
that themes can make it smaller at their will without the need to
override the render_background, render_frame and render_activity methods
of GtkThemingEngine.
https://bugzilla.gnome.org/show_bug.cgi?id=676196
This was showing up when using a combo box in list mode. After popping
up the list, the keyboard grab appeared stuck. What was stuck here is
only the client-side grab, since we forgot to clean up our grabs
when receiving an UnmapNotify.
This bug was introduced in 1c97003664.
https://bugzilla.redhat.com//show_bug.cgi?id=812035 has
a stacktrace that shows recursion via
free_node -> unref -> dispose -> ui manager api
which ends in a crash, since we run free_node over the entire
tree and it leaves lots of dangling pointers behind.
So, better be careful by setting all pointers to NULL after
freeing them.
And update/fix it for the new shorthand parsing, since the old test did
a couple of things wrong:
- it erroneously flipped background-clip and background-origin when
testing
- it was missing default initial values for background-size and
background-position from the generated test output
When the toplevel is a GdkOffscreenWindow which doesn't
implement the set_device_cursor() vfunc, we would have
crashed. Implement a dummy ->set_device_cursor vfunc.
https://bugzilla.gnome.org/show_bug.cgi?id=675809
To make setting output directory and filename simpler in the PrintToFile
dialog two gtkprintsettings have been added GTK_PRINT_SETTINGS_OUTPUT_DIR
and GTK_PRINT_SETTINGS_OUTPUT_BASENAME.
This will reduce the code needed to implement a better name than "output.pdf"
and actually makes more sense than the existing setting
GTK_PRINT_SETTINGS_OUTPUT_URI which doesn't work seamlessly with
GTK_PRINT_SETTINGS_OUTPUT_FILE_FORMAT like the new settings do.
https://bugzilla.gnome.org/show_bug.cgi?id=657322
Before, right click events were still let through into GDK. In this
case, also middle/right button events with x-coordinates in the range
[-3, 0] are processed, resulting in failures/crashes in the window
finding code because no GdkWindows are present in this range.
Not setting a URI but catching the activate-link signal is a
valid use of GtkLinkButton, but we shouldn't allow showing a
popup menu which offers to copy the URI if there's none.
Turn dead_doubleacute plus space into '"' and not into a double
acute because that's the way to enter double quotes on the
US-International keyboard layout.
(cherry picked from commit 71164e57b9)
Code was using different places to compute the icon size when layouting
and when doing size requests. This resulted in non-matching behavior in
obscure cases. And that lead to iconviews that were too small.
This makes sure that if the gtk-im-module setting changes we update
our internal state immediately on the next event whichever it is.
In particular this fixes the case of the gtk-im-module setting
changing while the user is typing and the slave context remaining the
same, effectively ignoring the setting change.
https://bugzilla.gnome.org/show_bug.cgi?id=675365
Commit cdf473ec10 deprecated the style
properties associated with the container border width and box spacing of
the action and content areas, in favour of using
gtk_container_set_border_width() and gtk_box_set_spacing() on the
widgets themselves, but failed to initialize those values to the
defaults.
Makes name consistent with other quartz-only modules and makes it clear that this works with the GMenuModel system rather than the older GtkMenu system.
The window's role is 'GtkFileChooserDialog', so that window managers can match it
for positioning.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
Instead, always use PKG_CHECK_MODULES(). That macro actually gets it
right. In particular the erroring out part when you miss xkbcommon or
wayland-client.
This ensures that items stay left-aligned instead of slowly expanding into
empty space when widening the iconview. It's also what the iconview did
pre-refactoring.
Note that for cases where natural width != minimum width, the cells
might still expand and shrink back.
Always assume max-columns and min-rows. The old approach was kinda
insane.
As an example, try to write an algorithm that optimizes the minimum size
for infinite (take a reasonably large number like 2520) word-wrapped
Monospace text cells containing the text "XXXXX XXX XXX XXXXX" (keep in
mind that this is the easy problem, because it's assuming equal cell
renderers). There's 4 ways to reasonably lay out this text:
19 glyphs (19x1):
XXXXX XXX XXX XXXXX
18 glyphs (9x2):
XXXXX XXX
XXX XXXXX
21 glyphs (7x3):
XXXXX
XXX XXX
XXXXX
20 glyphs (5x4):
XXXXX
XXX
XXX
XXXXX
The best thing to do usually is using the 9x2 approach, but that's
neither the one using the natural nor the one using the minimum size.
As a side note, this does not include spacing and padding, which might
also influence the decision. Nor does it include height-for-width
considerations. Look at this table (numbers given in glyphs, not pixels,
as for pixel-sizes it gets even more interesting):
given best solution
width columns sizing glyphs per cell
6 1 6x4 20
7 1 7x3 21
8 1 7x3 24
9 1 9x2 18
10 1/2 9x2/5x4 20
11 1/2 9x2/5x4 22
12 1/2 9x2/5x4 24
13 1/2 9x2/5x4 26
14 2 7x3 21
15 3 5x4 20
16 3 5x4 21.3
17 3 5x4 22.7
18 2 9x2 18
19 1/2 19x1/8x2 19
20 1/2/4 19x1/8x2/5x4 20
21 1-4 any 21
22 1-4 any 22
23 1-4 any 23
24 1-4 any 24
25 5 5x4 20
26 5 5x4 20.8
27 3 9x2 18
28 3 9x2 18.7
29 3 9x2 19.3
30 3/6 9x2/5x4 20
Now of course, nobody wants the number of columns to randomly change in
inexplicable ways while they enlarge or shrink an iconview, so we not
only have to optimize for smallest or other size measurements, but we
also have to optimize for "most pleasing to the eye".
And last but not least, I'd like to once again remind you - if you kept
up until now - that this discussion was for identically-sized cells
only.
Instead of just returning the last allocated numbers, we now compute the
proper sizes from scratch. This is a bit less trivial, but it results in
proper height-for-width handling.
This is a huge quest to remove all caching from GtkIconview to simplify
the code. As it turns out, iconview performance is a joke, so the caches
are kinda unnecessary.
If we need caching, we can add it in a useful way later.
If we use
&item->cell_area
instead of
(GdkRectangle *) item
there is no need anymore to keep the cell_area as the first member of
the the item. And we cget compile-time checks for changes to the item
struct.
It's wrong to hardcode the slider lines here; those should be rendered
as an additional background layer using background-image if the theme
specifies so.
https://bugzilla.gnome.org/show_bug.cgi?id=652304
Instead of setting it on the child scrolled window. This is needed
because the whole window's allocation must be equal to the one of the
entry (in case the popup-set-width property is TRUE); if we set the size
request on a children of the window, there might be other children with
borders/paddings in between the toplevel and the child we set the size
request too, which will break alignment.
https://bugzilla.gnome.org/show_bug.cgi?id=672947
GDK_EVENT_2BUTTON_PRESS and GDK_EVENT_3BUTTON_PRESS can't be used from
some bindings because they'd translate to something syntactically
invalid. Add GDK_EVENT_DOUBLE_BUTTON_PRESS and
GDK_EVENT_TRIPLE_BUTTON_PRESS aliases to work around that.
https://bugzilla.gnome.org/show_bug.cgi?id=671025
This call was forcing needless work since gtk_window_map() already
does a gdk_window_show() which initially sets GDK_WINDOW_STATE_FOCUSED
that we then handle regularly on the widget's window state event
handler.
https://bugzilla.gnome.org/show_bug.cgi?id=673237
Since themes might want different paddings around the color sample
according to where it's being used, don't hardcode a 16px one here.
The theme can specify a padding to get the same effect.
Reported by Christian Persch <chpe@gnome.org>
Add a fallback-c89.c for the gdk/ subdirectory as there is code that uses
functions that are introduced by C99. This currently adds fallback
implementations for MSVC for isnan() and isinf()
Dist this "new" source file as well
nearbyint(), isinf() and isnan() are C99 functions, so check for them.
Also clean up configure.ac a little bit as the checks for rint() and
round() can be a bit simpler, according to Matthias' suggestions.
On Windows, gtkwin32themeprivate.h is needed as
_gtk_win32_theme_get_default() is called on that platform to avoid C4013
warnings/errors (aka implicit declaration of ... for GCC folks).
See inline comments for what it does. Its main use is figuring out if
something has been caused by GTK's caching of CSS properties or if it's
a different problem.
.. instead of the previous stable version.
This ensures that if we use GDK_DEPRECATED_IN_3_6, it will actually emit
a warning in GTK 3.5, and not wait with that until GTK 3.7. This is
particularly useful for stuff that gets deprecated right now. This code
should emit warnings right now, so we know what we're doing while
deprecating.
GApplication now makes the session bus and object path available as a
public API on the application instance. Use that instead of trying to
guess values for ourselves.
This causes this version of Gtk+ to depend on GLib 2.32.2, so bumping
version dependency accordingly.
https://bugzilla.gnome.org/show_bug.cgi?id=671249
We currently have a couple of cases where GtkApplication assumes that
the session bus will be non-NULL causing critical error output or (in
the case of trying to publish menus) an infinite loop.
Three fixes:
- if the session bus is NULL due to not having registered the
GtkApplication yet then give a g_critical on the entry point to the
menu setters instead of going into an infinite loop. Document this.
- check for NULL session bus even when calling the menu setters at the
right time in order to prevent the infinite loop for
non-programer-error cases (ie: because we had trouble connecting to
the session bus)
- check for NULL session bus when publishing the X11 properties on the
GtkApplicationWindow and skip publishing them if we're not on the bus
https://bugzilla.gnome.org/show_bug.cgi?id=671249
The way we use these style properties to set regular properties on
containers accessible from the public API is really just broken, and
could lead to undefined values for the spacing and border-width
container properties (since they could be set from public API and then
changed from under in a style_update handler from GTK).
Take this as an occasion to deprecate these style properties, which do
not make a lot of sense anyway, now that GtkInfoBar supports regular CSS
padding and border.
GtkPlug directly handles X KeyPress/Release events, instead of using
translation in GDK (which expects XI2 events for XI2). When this
was done, the handling of the group was stubbed out and never replaced.
Export gdk_keymap_x11_group_for_state() and gdk_keymap_x11_is_modifier()
so we can fill out the fields correctly.
https://bugzilla.gnome.org/show_bug.cgi?id=675167
InfoBar must take into account the border and padding when requesting
its allocation, since it is then drawing them. Besides, the border and
background should always be drawn, even when the message type is OTHER.
These need to be made independent of the xkb configuration somehow.
As things are now, they will either fail when run on a naked X
server in make check, or fail when run in my session.
If the PPD is not available when the printer is loaded then the PPD options are
not available and we can't get the correct qualifier to use with colord.
When the PPD becomes available, refresh the profile title to reflect reality.
https://bugzilla.gnome.org/show_bug.cgi?id=674890
Shift-click in the slider now starts a drag in 'fine adjustment'
mode, where we move the slider 10-times slower than the mouse.
This can be very helpful when scrolling through a very long document
or webpage, and moving the scrollbar even a single pixel already
jumps too far in the content.
https://bugzilla.gnome.org/show_bug.cgi?id=563688
It seems to be general consensus that button 1 should do the jumping,
so we now jump to the clicked position on primary button clicks and
page on secondary button clicks. Touch behaves like primary.
https://bugzilla.gnome.org/show_bug.cgi?id=563688
Symbolic icons use a "-symbolic" suffix to distinguish themselves from
highcolor variants. Note that the dash character here has a different
meaning than the specificity level defined in the icon-naming-spec [1],
as it identifies a property of the icon itself.
Since they might be provided by a parent theme (e.g. the HighContrast theme
relies on the gnome icon theme for them), when we are looking up one we
should first escape the generic icon inheritance mechanism defined in the
icon-naming-spec [1], and privilege a symbolic icon, if it exists in a
parent theme, before applying the inheritance evaluation.
This fixes symbolic icons not working properly when used in the
HighContrast theme with the GTK_ICON_LOOKUP_GENERIC_FALLBACK flag set.
[1]
http://standards.freedesktop.org/icon-naming-spec/latest/ar01s03.htmlhttps://bugzilla.gnome.org/show_bug.cgi?id=674806
Avoid doing useless translations/rotations, since themes will most
likely set different CSS gradients using left/right/top/bottom style
classes, or use a plain color.
https://bugzilla.gnome.org/show_bug.cgi?id=674809
This allows combo box popup windows to appear in the correct place. This is a
workaround emulating root coordinates by calculating the offset for a chain
of windows that that are relative to each other.
This was broken since before GTK+ 3.0, when we replaced
a use of requisition by allocation. Fix this by using the
requisition height, that is already cached by the menu code.
The math is not quite right here; if you page all the way
down a long menu, you end up on the second-to-last menuitem.
But at least, page up/down let you move up and down the menu
again.
https://bugzilla.gnome.org/show_bug.cgi?id=668931
Having refactored cups_request_printer_list_cb so that the cups
version-dependent block size is small enough to be handled in a single ifdef,
make the ifdef HAVE_CUPS_API_1_6 block.
So that it can be passed as a single parameter to functions as we extract-function to make cups_request_printer_list_cb more manageable.
Note that not all of the affected variables are changed in this changeset. Those are in extracted functions and will be addressed in the next two changes.
Since the event will be ignored anyway after it's translated (slave
devices are disabled), don't let it run in the smooth scroll code path,
as it will burn our caches for the actual event we're interested in.
https://bugzilla.gnome.org/show_bug.cgi?id=673644
Of course, we must pass coordinates in the NSWindow coordinate system
when creating an NSEvent. This fixes drag icon positioning and makes
the icon slide back to the correct position when the drag is
canceled.
Due to the way the tests are structured, a missing libXext will give a
warning about a missing libX11 (even if libX11 is installed). This is
confusing to people who are trying to build Gtk.
https://bugzilla.gnome.org/show_bug.cgi?id=674200
Instead of overriding the font theme settings, just set the Pango
attributes we want on the label. This fixes message dialogs growing on
style_update after recent GTK+ changes.
Instead of using 1 global queue for both resizes and style validation,
use 2 queues. This makes the code a lot simpler and fixes a bug where we
could accidentally stop restylying for very delayed restyles.
We now animate the core style information (see comment in
gtk_style_context_save()). A lot of widgets save + set custom style
classes/states during drawing and so can't be animated. It does work for
labels, menus and buttons though.
This is a GtkCssComputedValues subclass. So it's essentially a store for
computed CSS values. But it can be animated by advancing it to a certain
timestamp.
A StyleAnimation is an immutable object used to track the state of CSS
values. I'd have liked to make it fully immutable - ie not have the
timestamp in there - but couldn't find a place to sanely store the
timestamp.
This is an abstract base class. Implementations for this will be added
later (for both CSS3 transitions and animations, potentially for
animated images).
Actually aplying the information in this object will be done by a
different object commtted later.
This has two goals:
1) Move invalidation code out of a nested if branch. Invalidation is
actually the most important thing this function does.
2) Have the changes bitmask available. It will needed for invalidate
calls to children later.
The design principles were:
- synchronized
If multiple style contexts are animating, they should all do an
animation step at the same time.
- degrades well
Even when there's thousands of style contexts all animating at the same
time, the animation steps don't starve the CPU. This is achieved by
making sure the timeout is really fast. It just sets a bunch of flags.
- no hidden bottlenecks
Turning animatability on or off on a style context is O(1).
So far it is unused.
This is only a small performance boost by itself, but it's necessary
for animations, so we need it.
Benchmark numbers for my Glade benchmark for interested people:
GTK 3.4.0 last commit this commit
Raleigh
real 0m41.879s 0m10.176s 0m9.900s
user 0m41.394s 0m9.895s 0m9.628s
sys 0m0.111s 0m0.096s 0m0.102s
Adwaita (*)
real 0m51.049s 0m13.432s 0m14.848s 0m12.253s
user 0m50.487s 0m13.034s 0m13.218s 0m11.927s
sys 0m0.117s 0m0.151s 0m0.147s 0m0.107s
Ambiance (patched to not use private GTK APIs)
real 0m52.167s 0m13.115s 0m13.117s 0m12.944s
user 0m51.576s 0m12.739s 0m12.768s 0m12.651s
sys 0m0.119s 0m0.137s 0m0.136s 0m0.118s
(*) Adwaita and unico currently use custom properties, and
_gtk_css_value_compare() for custom properties always returns FALSE,
which makes this optimization never trigger. So I modified
_gtk_css_value_compare() to return TRUE for these properties instead and
reran the benchmark. Those are the numbers.
Add an internal API that allows GtkStyleContext to create a widget path
for the widget and with that bypassing gtk_widget_get_path() and that
function caching the path.
Instead, look up the variable upon use. This is more correct (for when
the engine changes due to save/restore() shenanigans.
And it removes code that doesn't use the standard code paths.
Equality tests are done with _gtk_css_value_equal(). There is no need to
do it per-property, equal values will still be equal.
This essentially reverts 24f5d54329e028347bd76af42e86ed190c1229a2 and
92c7a7171e1240b6d961ee5b6f9ab6b596e98904.
... and Make this new value be a real GValue, as we don't need to save
performance for these anymore (it's just used for custom properties).
And I'd rather have code work for all values then be optimized for no
reason.
Deprecate public API where appropriate and make it no-ops.
Remove all calls to it.
Get rid of the 'transition' css property.
For now, this means spinners don't animate anymore.
Instead of keeping a custom GPtrArray, keep it as a GtkCssArrayValue of
GtkCssStringValue. This way, we gain equality comparisons and print
functions for free.
All the properties now are a GtkCssArrayValue of GtkCssSadowValue.
GtkCssArrayValue already does everything we want, so no need to
duplicate its funtionality.
When a parent style context exists, there's no need to queue_resize() on
the widget, because the parent widget will call
gtk_style_context_validate() on us and _then_ we can call queue_resize()
if we have to.
Only the ones that can be animated are marked of course. So more work is
needed.
This is important for
transition-property: all;
because it'll just animate all the properties that can be.
So instead of using
_gtk_css_style_property_get_id (GTK_CSS_STYLE_PROPERTY
(_gtk_style_property_lookup ("name")))
one can now use
GTK_CSS_PROPERTY_NAME
Esaier, eh?
Returns a value that transitions between start and end or %NULL if the
values cannot be transitioned.
So far, all implementations but numbers and rgba return NULL.
Note: custom CSS properties still use the default GtkCssValue and always
will.
So there is a difference in css values used between those, even though
they both carry a GdkRGBA payload.
The compute_value fallback path is only needed for custom properties,
the real style properties have custom compute functions if they need
them already.
This is in preparation for removing the specified type and computed type
properties from GtkCssStyleProperty, which is in preparation for really
using GtkCssValue classes and not GTypes.
For now, we return FALSE for all default css values, so this is not very
useful.
I also think of this as an optimization equal, not a guaranteed equal,
because we don't even have a notion of what "equal" means.
For example, for background-repeat, "repeat, repeat" and "repeat"
are functionally equivalent. But the cssvalue has no idea that it's used
for background-repeat.
As a more complicated example, "repeat, no-repeat" and "repeat" are
equal to what one sees as long as there's only one image listed
background-image-source. But once you start transition'ing to an image
with 2 sources, it's different...
This change isn't strictly correct, but I can't be bothered until we get
a tokenizer that can really differentiate between allowed values and
invalid syntax.
... and actually set the widget on the style context. Note that this
function does not take a reference on the widget, which is a very good
reason to keep it private.
This matcher always matches only on some relvant things and ignores the
rest. This allows you to match only on name and class, but ignore state
and parents/siblings for example.
We now track the position as a (type,a,b) tuple where the numbers make
up the an + b formula from CSS3 nth-child.
Also, the get_sibling() and get_sibling_index() vfuncs were replaced by
a has_position() vfunc. This is mostly so that the matcher can always
return TRUE. And I need that for the everything matcher.
This way, we don't have to do magic inside GtkStyleContext, but have a
real API.
As a cute bonus, this object implements GtkStyleProvider itself. So we
can just pretend there's only one provider.
This change adds --enable-wayland-cairo-gl which turns on the define used in
the Wayland backend to determine whether to use EGL surfaces with Cairo GL or
whether to use the Cairo image backend with an SHM surface (the default).
Part of the fix for: https://bugzilla.gnome.org/show_bug.cgi?id=672361
This moves the freeing of the icon_helper from the destory to the finalize
function to avoid segfaults when trying to access a destroyed object before it
is disposed. This often happens in signal handlers which get called
asynchronously after destroy.
https://bugzilla.gnome.org/show_bug.cgi?id=674050
Try to fetch the name from the application desktop file for the
fallback menu if possible, instead of forcing applications to use
g_set_application_name or hardcoding "Application".
https://bugzilla.gnome.org/show_bug.cgi?id=673882
In the Quartz backend, there are two methods by which windows are
resized. The first method is fully handled by Quartz and does not appear
in the event stream the application resizes. The second method is when
we resize windows by ourselves. In OS X this happens when a GTK+ resize
grip is used. This resize grip is larger than the Quartz resize grip.
When the resize is started outside the "Quartz area", we have to handle
it by ourselves.
This patch fixes this manual window resizing by ignoring events while we
are in the process of resizing (such that the events actually arrive at
the sendEvent handler of GdkQuartzWindow where this resize is handled).
When the resize has finished we break all grabs such that GDK is not
stuck thinking the cursor is still in the resize window.
gtk_entry_completion_set_property() was setting many properties by
directly modifying priv values, bypassing notification invocation and
possibly another actions done by gtk_completion_entry_set_xxx ()
functions. Fix by invoking set_xxx() instead of setting the property
value directly.
The real bug observerd was that setting text-column property using
g_object_set() caused SIGFPE later when entry completion was about to
appear. gtk_entry_completion_set_text_column () apparently does way
more important things than just setting priv->text_column member.
https://bugzilla.gnome.org/show_bug.cgi?id=673693
Instead, just draw the children. The cairo code will keep track of
things, so there's no need to track things.
Also, the old code was doing it wrong.
https://bugzilla.gnome.org/show_bug.cgi?id=672544
Style properties should not be cached, they should be queried live.
Also, this fixes the case where the expander size wasn't set when
constructing the widget which caused expanders to go missing.
This can cause lagging when scrolling as it causes us to repaint
on every scroll event. This wasn't historically a great problem,
but with smooth scrolling we get a lot more events, so this
now creates visible lagging on slower machines.
_gdk_x11_moveresize_configure_done() isn't called for wmspec
moves/resizes so we don't have a way to notice when a wmspec
move/resize ends and consequently untrigger the sending of
_NET_WM_MOVERESIZE_CANCEL which results in this message always being
sent on the next button release event. In that case we are marking
that event as handled so it isn't processed further which breaks
button press/release event handling in several widgets.
To fix this we simply allow the normal event handling machinery to run
after sending the _NET_WM_MOVERESIZE_CANCEL message.
https://bugzilla.gnome.org/show_bug.cgi?id=673328
Since the order in which _NET_WM_STATE and _NET_WM_DESKTOP are set, or
even *if* they are set, isn't defined, we could end up unsetting
GDK_WINDOW_STATE_FOCUSED given that both handlers for these two X
properties end up doing window state changes for all states. As we
want GDK_WINDOW_STATE_FOCUSED to be set by default we need to set its
master flag by default as well.
https://bugzilla.gnome.org/show_bug.cgi?id=673125
Don't handle mouse button events greater than 5 so
they can bubble up to be used by the application.
This was causing nautilus list view to not go forward
and backwards when pressing the extra mouse buttons
designated for that.
Fixes bug 673441
Signed-off-by: Nelson Benitez Leon <nbenitezl@gmail.com>
This is not ideal, we should have a real classic windows theme,
but at least its better than everything being pink, which is what
happens otherwise when theming is not enables.
We now have a proper MASTER/SLAVE input device split, where
the masters are virtual core input devices and we add fake hw
slave devices for the system pointer and real slave devices for
wintab devices.
We also set the proper source_device on the events so you can
tell which device sent it and properly decode the axis info.
gtk_application_set_app_menu(), gtk_application_set_menubar():
Mention that you probably want to call this in the startup signal
handler. If you do it earlier you will likely get a warning about
a missing D-Bus connection, because doing it earlier does not
make sense anyway.
Since we check for !list->next (and !list->prev for RTL) to set the
GTK_REGION_LAST flag, we have to filter out invisible columns before
looping; if we don't do that we might end up assigning GTK_REGION_LAST
to an invisible column.
https://bugzilla.gnome.org/show_bug.cgi?id=672937
When a homogeneous grid has no visible children, we were
accidentally doing a division by zero. Instead, just bail
out early in this case, there is nothing to allocate anyway.
https://bugzilla.gnome.org/show_bug.cgi?id=672763
The code for calculating the per-monitor workarea was ignoring
the fact that the EWMH workarea property can only handle rectangular
workareas, and thus can't really do justice to general monitor
arrangements. As a workaround, we ignore it for anything but
the primary monitor. And we ignore it for the primary monitor
as well if it does not even cover it.
https://bugzilla.gnome.org/show_bug.cgi?id=672163
We want to use Alt+Print for screenshot keybindings, so we have to avoid
translating it into SysRq, since that's what the keymap would dictate.
After talking with Owen, it sounds like doing this generically would be
a lot of work and quite hairy code, so hardcoding this should be OK.
https://bugzilla.gnome.org/show_bug.cgi?id=671006
When we are re-setting the same text for internal reasons
(e.g. when applying the mnemonics-visible change upon Alt press),
we should not needlessly loos the selection.
https://bugzilla.gnome.org/show_bug.cgi?id=671588
Commit de62a1096 broke win32 as it removed the HAVE_X11R6 checks for
building xim, but did not replace them with USE_X11. This made
it try to build xim on non-X targets.
We only support the preview state for checkboxes for GtkButton
because it gave problems when used in treeview where the prelight
state affects all checkboxes in the row.
When GtkPrinterFunc always returns FALSE, for example when looking for
a non existent printer, if print list is done for all backends or print
backend status is UNAVAILABLE, gtk_enumerate_printers() finishes with an
empty backend list and destroy function is never called. We need to
check the backend list again after calling list_printers_init for all
backends and finish the enumeration if it's empty.
https://bugzilla.gnome.org/show_bug.cgi?id=672125
As we don't give out information about the group or level when
giving back keycodes, we should prioritise group-0, level-0, followed by
level-0 only, and then any keycodes.
This fixes "q" being pressed when the Wacom tablet code in
gnome-settings-daemon is supposed to generate an "a" ("uk" keymap
in group-0, "fr" in group-1).
https://bugzilla.gnome.org/show_bug.cgi?id=671065
When the scrolled window has a frame (and the scrollbar is within the
bevel), we should take into account the CSS border/padding of the frame
and offset the scrollbars junction rendering with it.
It seems XP doesn't handle drawing non-alpha theme parts
on alpha destinations. We fix this by using alpha bitmaps only when
needed.
However this means any non-drawn area by the theme part is now draw
black, so we must take more care to only draw where the theme part draws,
so we find the theme part size when available.
We now store the symbolic colors as a GtkCssValue which means that
we can reuse the color when resolving and storing the color in
the computed values in the style context.
Additionally we keep a last_resolved GtkCssValue cache in the
GtkSymbolicColor, and if resolving the color returns the same as
last time we reuse the old value. This further increases sharing
of Css Values.
va_copy() is not universally available, and we already have a G_VA_COPY
macro that emulates the behaviour of va_copy() when it's not available, or
simply calls va_copy() if it's there
Also, in places where we're computing a new CssValue based on an
old one, make sure that if nothing changes we're returning a reference
to the old one, rather than creating a new identical instance.
Some people destroy their widgets in ::response, so trying
to access dialog internals from the class handler that is
running afterwards has the potential to cause problems.
Instead, we can reset the ::show-editor property every
time we map the dialog.
We want to avoid handling focus events for the private focus window,
otherwise the keyboard grab taken by for example buttons will cause a
spurious FOCUS_OUT/FOCUS_IN on the toplevel.
The code that did this seems to have been lost in the XI2 transition for
GTK3.
This patch reapplies db4a6040af which was
backed out in 18406b7b04 to give
developers a chance to get their X servers fixed. As we want to get this
bugfix in for 3.4, we need to commit it now.
https://bugzilla.gnome.org/show_bug.cgi?id=657578
Check for the XIScrollClassInfo struct in addition to the existing
check for XIAllowTouchEvents() because Ubuntu Oneiric seems to
have an incomplete backport which has one but not the other.
Based on a patch by Murray Cumming,
https://bugzilla.gnome.org/show_bug.cgi?id=671453
Scroll events report normalized deltas in terms of an abstract
'scroll unit' now, so our job is to determine a suitable scroll
unit here. Since we are changing the value of the adjustment,
the allocation of the widget does not factor into this at all.
XI2 provides us with an increment for each scroll valuator,
and by dividing the delta by the increment, we obtain normalized
values in some abstract 'scroll unit'.
For mouse wheels, the evdev driver reports an increment of -1,
so doing this division fixes the inverted scrolling with wheels
that we've seen recently.
When doing homogeneous allocation in the presence of
overlapping spanning children, we need to avoid uneven
line allocations, otherwise, the final homogenization
will blow up the size request of the grid.
https://bugzilla.gnome.org/show_bug.cgi?id=671170
In particular gtksettings.h and gtkstylecontext.h needed to be included
in lots of places now.
Also, I order the includes alphabetically in a bunch of headers.
Support long press for customizing, and short press for
selecting/activating. This is simpler than the generic
press-and-hold support in the multitouch branch; we don't
display any feedback, and the timeout is currently hardcoded
to 1 second.
GtkRange was using GDK_POINTER_MOTION_MASK, and it was not
getting any emulated motion events, because we only translate
from GDK_BUTTON_MOTION_MASK to GDK_POINTER_MOTION_MASK, but not
the other way around, and emulated_mask only had
GDK_BUTTON_MOTION_MASK in it. Now we put GDK_POINTER_MOTION_MASK
in emulated_mask and successfully match for windows that
have GDK_POINTER_MOTION_MASK or any of the button motion masks
selected.
This fixes range sliders not following the finger and jumping
to the last position upon release.
If the Window Manager supports the _NET_WM_STATE_HIDDEN, we use it to use
the _NET_WM_STATE protocol when de-iconifying windows (iconification is
unchanged, via XIconifyWindow). Additionally, we no longer interpret all
UnmapNotify events for our window as the result of iconification.
(Based on patch by Tomas Frydrych <tf@linux.intel.com>)
Previously we kept a Selector object for every "simple selector" (term
from CSS spec). Now we keep one for every match operation. So given the
selector
".a b:focus"
we will have 4 elements:
- pseudoclass ":focus"
- element "b"
- match any desendant (the space)
- class ".a"
Each of those is represented by a "selector class" which is basically
the collection of vfuncs for this selector.
NSEvent -scrollingDeltaX and -scrollingDeltaY aren't defined before
10.7, so objc assumes that they return a pointer. Trying to cast to a
float generates a compiler error.
If delta_x/y information is provided in scroll events, use it
to modify the underlying adjustment in steps proportional to
the deltas provided.
If the child widget of a scrolledwindow doesn't set
GDK_SMOOTH_SCROLL_MASK, regular scroll events will be dispatched,
and still handled by these 2 widgets.
XInput >= 2.1 allows for implementing smooth scrolling,
reporting the different scrolling axes as valuators.
Any change in those will be reported as GdkEventScroll
events with delta_x/y information.
the older kind of scroll events is still handled, and
emulated in devices able to provide smooth scrolling,
setting _gdk_event_set_pointer_emulated() in that case.
nsevent scrollingDeltaX/Y (available on OSX >= Lion) is used to
provide the smooth scrolling values. In any case, old fashioned
events are still sent, setting _gdk_event_set_pointer_emulated()
if the event contains both smooth and non-smooth values.
Events of type GDK_SCROLL will be received if the client side window
event mask has either GDK_SCROLL_MASK or GDK_SMOOTH_SCROLL_MASK.
GDK_BUTTON_PRESS_MASK has been removed from type_masks[GDK_SCROLL]
as that bit is often set for other-than-scrolling purposes, and
yet have the window receive scroll events. In GTK+, this forces
non-smooth events bubbling, even if the widgets above want smooth
events, and legitimately set GDK_[SMOOTH_]SCROLL_MASK.
If a device provides both smooth and non-smooth events, the latter will be
flagged with _gdk_event_set_pointer_emulated() so the client side window
receives one or the other. If a device is only able to deliver non-smooth
events, those will be sent, so both direction/deltas may need to be handled.
get_event_window() just checked on GDK_TOUCH_MASK, including for emulated
pointer events, so at the very least those should also match evmasks with
no touch events whatsoever
If an active grab kicks in on a different window, _gdk_display_has_device_grab()
would still find the former implicit grab for the window below the pointer, thus
sending events to an unrelated place.
If a grab with GDK_TOUCH_MASK kicks in due to a touch sequence emulating pointer
events, don't mutate the sequence into emitting touch events right away.
Create the backing GdkTouchGrabInfo for touches even if the pointer
emulating touch sequence is already holding an implicit grab on a
window that didn't select for touch events.
the backing GdkTouchGrabInfo will be needed if the overriding device
grab finishes before the touch does in order to send events back to
the implicit grab window. Instead, wait until the touch is physically
finished before removing the matching GdkTouchGrabInfo
GDK will only receive touch events when dealing with a multitouch
device, so these must be transformed to pointer events if the
client-side window receiving the event doesn't listen to touch
events, and the touch sequence the event is from does emulate
the pointer.
If a sequence emulates pointer events, it will result in a
button-press, N motions with GDK_BUTTON1_MASK set and a
button-release event, and it will deliver crossing events
as specified by the current device grab.
These are equivalent to an implicit grab (with !owner_events), so
if the touch leaves or enters the grab window, the other window
won't receive the corresponding counter-event.
If the touch sequence happens on a window with GDK_TOUCH_MASK set,
a GdkTouchGrabInfo is created to back it up. Else a device grab is
only created if the sequence emulates the pointer.
If both a device and a touch grab are present on a window, the later
of them both is obeyed, Any grab on the device happening after a
touch grab generates grab-broken on all the windows an implicit
touch grab was going on.
Touch events don't generate crossing events themselves, so
do not rely on these to determine whether the button release
happened within the event window.
This widget is too narrow to make touch interaction tricky enough, so
don't add the penalty of having the slider run farther from the touch
coordinates if it happens to miss the slider.
This is so submenus stay open as the parent menu item is
pressed/released, since the user would typically lift the
finger in order to select a submenu item.
This makes kinetic scrolling work with viewports where the
content does not otherwise select for button or touch events,
such as testscrolledwindow's label.
Kinetic scrolling is only done on touch devices, since it is
sort of meaningless on pointer devices, besides it implies
a different input event handling on child widgets that is
unnecessary there.
If the scrolling doesn't start after a long press, the scrolling is
cancelled and events are handled by child widgets normally.
When clicked again close to the previous button press location
(assuming it had ~0 movement), the scrolled window will allow
the child to handle the events immediately.
This is so the user doesn't have to wait to the press-and-hold
timeout in order to operate on the scrolledwindow child.
The innermost scrolled window always gets to capture the events, all
scrolled windows above it just let the event go through. Ideally
reaching a limit on the innermost scrolled window would propagate
the dragging up the hierarchy in order to keep following the touch
coords, although that'd involve rather evil hacks just to cater
for broken UIs.
Anytime a touch device interacts, the crossing events generation
will change to a touch mode where only events with mode
GDK_CROSSING_TOUCH_BEGIN/END are handled, and those are sent
around touch begin/end. Those are virtual as the master
device may still stay on the window.
Whenever there is a switch of slave device (the user starts
using another non-touch device), a crossing event with mode
GDK_CROSSING_DEVICE_SWITCH may generated if needed, and the normal
crossing event handling is resumed.
This patch adds a capture phase to GTK+'s event propagation
model. Events are first propagated from the toplevel (or the
grab widget, if a grab is in place) down to the target widget
and then back up. The second phase is using the existing
::event signal, the new capture phase is using a private
API instead of a public signal for now.
This mechanism can be used in many places where we currently
have to prevent child widgets from getting events by putting
an input-only window over them. It will also be used to implement
kinetic scrolling in subsequent patches.
http://bugzilla.gnome.org/show_bug.cgi?id=641836
We automatically request more motion events in behalf of
the original widget if it listens to motion hints. So
the capturing widget doesn't need to handle such
implementation details.
We are not making event capture part of the public API for 3.4,
which is why there is no ::captured-event signal.
We don't want to fallback for 'random' touch sequences, since
that could lead to all kinds of pairedness and other violations.
Since the X server already tells us what touch events it would
have used for emulating pointer events, we just use that information
here.
Translate XI_TouchBegin/Update/End to GDK_TOUCH_BEGIN/UPDATE/END
events.
At the same time,
set pointer-emulated flags on button events with XIPointerEmulated
and on touch events emulating the pointer.
This commit introduces GDK_TOUCH_BEGIN/UPDATE/END/CANCEL
and a separate GdkEventTouch struct that they use. This
is closer to the touch event API of other platforms and
matches the xi2 events closely, too.
We introduce GDK_SOURCE_TOUCHSCREEN and GDK_SOURCE_TOUCHPAD
for direct and indirect touch devices, respecively. These
correspond to XIDirectTouch and XIDependentTouch in XI2.
GtkButton currently draws itself as active (pressed down) in case we're
pressing and holding the mouse pointer outside its bounds; this is
misleading though, since we won't activate the button unless the mouse
is released inside the button itself.
Fix this by only setting the ACTIVE state flag when the button is
actually pressed down.
https://bugzilla.gnome.org/show_bug.cgi?id=668141
This does nothing but turn all GtkBitmask functions into static inline
functions that call the gtk_allocated_bitmask_*() equivalent.
The implementation of the static functions has also been put into a
private header, to not scare people who want to see how things are
implemented.
My previous fix for this broke the progress bar in epiphany. This fix
makes it work again, and keeps the gimp bug fixed.
Basically, whenever we do a non-double-buffered rendering we have to
flush the entire window as it might be drawn outside the double
buffering machinery.
When we're allocating children of GtkOverlay, compare their allocation
with the overlay one, and set left/right/top/bottom style classes if the
overlaid widget touches one or more of the overlay edges.
https://bugzilla.gnome.org/show_bug.cgi?id=669342
gtk_widget_translate_coordinates() can fail in case the widget is not
realized or there's no common ancestor. Don't use the x/y values
returned by that method in that case, since their value is undefined.
If there's a junction between the two scrollbars (i.e. they're both
visible), draw a background with a style class there, so the theme can
style it.
https://bugzilla.gnome.org/show_bug.cgi?id=669335
If the display server or GDK hides the window - fire the "deactivate" signal
to ensure that the internal state is consistent.
This patch also ensures that the "deactivate" signal will not be fired for a
menu that is not active.
Bug: https://bugzilla.gnome.org/show_bug.cgi?id=670881
Since GtkCellRendererText moved to WFH requests, our get_size
implementation is ignored. We should override get_preferred_width
instead. This fixes the accel renderer being clipped to a wrong size
when trying to edit its shortcut.
Once we've made them popup windows we must also implement the popup_done event
handler on the shell surface listener. The best we can currently do is to hide
the window. This will then signal up to GTK which could then deactivate the
appropriate menu (see https://bugzilla.gnome.org/show_bug.cgi?id=670881)
This allows us to get the device if we need to make the window a popup. This
relies on the side effect that GTK calls into GDK to take a grab before the
popup window is shown.
Since it's generated, we install the header by putting it into
gdkinclude_HEADERS, so it's wrong to put it again into
gdk_public_h_sources.
This fixes the build.
This patch changes all uses of GDK_DEPRECATED(_FOR) in gtk headers
by the versioned variants, GDK_DEPRECATED_IN_3_x(_FOR). At the same
time, we add GDK_AVAILABLE_IN_3_x annotations for all API additions
in 3.2 and 3.4.
This patch changes all uses of GDK_DEPRECATED(_FOR) in gdk headers
by the versioned variants, GDK_DEPRECATED_IN_3_x(_FOR). At the same
time, we add GDK_AVAILABLE_IN_3_x annotations for all API additions
in 3.2 and 3.4.
These macros follow the recent changes in GLibs deprecation
setup. We now annotate deprecated functions with the version
they were deprecated in, and you can define the macro
GDK_VERSION_MIN_REQUIRED to cut off deprecation warnings for
'recent' deprecations.
At the same time, we introduce version annotations for new API
and allow you to avoid 'recent' API additions by defining
GDK_VERSION_MAX_ALLOWED.
The message-type css classes must be in the widget context all the time,
not only when drawing, otherwise they are not propagated to the
children, for instance a label in the InfoBar must inherit the
color. Add a corresponding reftest.
https://bugzilla.gnome.org/show_bug.cgi?id=670555
The widget window is usually covered by the bin_window.
Its background color will become relevant when we introduce
kinetic scrolling with overshooting.
The widget window is usually covered by the bin_window.
Its background color will become relevant when we introduce
kinetic scrolling with overshooting.
The widget window is usually covered by the bin_window.
Its background color will become relevant when we introduce
kinetic scrolling with overshooting.
_gtk_widget_set_device_window() is suppose to make accounting of
the topmost widget under the device at each time, so avoid setting
it on virtual crossing events as the device is already in another
window.
The implicit grab on priv->event_window already warrants that this
widget is the only one getting events while the button is pressed,
so avoid the extra GTK+ grab here.
This last slave device (stored per master) is used to fill
in the missing slave device in synthesized crossing events
that are not directly caused by a device event (ie due to
configure events or grabs).
Store the device, and unset private fields whenever the device
is shadowed by another GTK+ grab, so popping up menus while
selecting (i.e. press-and-hold) doesn't leave the entry in a
confused state.
* Restores the old padding
* Prelight on spin buttons
* Don't have a generic prelight background selector, as that got
picked up by things like images that should have a transparent bg.
The F keys have no unicode mapping, and UCKeyTranslate() returns
a bogus 0x10 as mapping to unicode. Instead of checking for this
random and undocumented return value, simply assign all function
keys explicitly. This patch also splits the ill-named "known_keys"
array into "modifier_keys" and "function_keys" which is much
more obvious.
(cherry picked from commit 55f9e5cbaf)
Move g_return_if_fail() stuff from the backends to the public
functions in gdkscreen.c itself, and some fixes for ugly formatting in
the various gdkscreen-backend.c files.
No need to subtract focus line width again, since the progressbar is
rendered starting at (0, 0).
This also fixes the entry-progressbar-coloring reftest.
The file "installation" part needed a long-overdue update, especially
as some headers were simply moved into gtk/deprecated and some new headers
were introduced, and a new .gschema.xml file needed to be processed.
Instead of firing a 'quit' signal and expecting the application to do
something that will cause it to quit, just call the new
g_application_quit() API for ourselves.
https://bugzilla.gnome.org/show_bug.cgi?id=670485
This seems a bit "too powerful" and unlikely to be used by most
applications. Remove it from now, until someone comes up with a strong
desire for it.
https://bugzilla.gnome.org/show_bug.cgi?id=670485
Instead of having an input/output GdkWindow, make the widget no-window,
and use a separate input-only window for events, and paint on the parent
window directly.
When a subclass of GtkEntry (e.g. GtkSpinButton) resizes the available
text area (by overriding the get_text_area_size vfunc), we need to
ensure we don't draw a possible progressbar over the part that got
removed from the text area.
This fixes drawing a progressbar in GtkSpinButton and in its subclasses,
such as GimpSpinScale, and makes Mitch happy too!
_gdk_x11_keymap_key_is_modifier() never tries to set min/max_keycode
if they haven't been set before, meaning that until another function
sets those, all the keys will be seen as non-modifiers.
This causes GdkKeyEvents to be wrongly tagged with "->is_modifier = 0"
when in actual fact the key is a modifier. This fixes keyboard
shortcuts captured with GtkCellRendererAccel in "raw" mode thinking
a modifier without any actual keys is a valid shortcut.
https://bugzilla.gnome.org/show_bug.cgi?id=670400
We used to set a flushed boolean whenever we flushing double buffered
areas to the window due to a non-db draw. We then read back from the
window if this was set. This broke when we were doing multiple paints
of the same area after a flush as we were re-reading the window each
time, overdrawing what was previously draw.
Subclasses of GtkEntry could set a larger height request, so we need to
apply the same calculations to the insertion cursors than we do on the
PangoLayout to render it centered under all circumstances.
Instead of GtkDrawingArea, since that calls in realize
gtk_style_context_set_background(). We don't want that to happen, given
that we do all the painting ourselves in _draw().
Instead of special-casing Adwaita, apply the half-width logic for themes
that have a scale slider with vertical proportions.
Also, simplify the rendering code a bit by factoring out the trough
sizing logic.
Instead of going GtkAlignment->GtkFrame->GtkAlignment, just pack a
GtkDrawingArea inside the button, and use halign/margin properties to
get the desired layout.
Since we only allow a single row there, it's better to just use a
GtkBox, and use :first-child, :last-child and :only-child to style
swatches in there.
Instead of calling gtk_render_check() there, just render a symbolic
icon, falling back to a built-in one if the icon is not available.
Also, add a style class for the active badge on the swatch:
"color-active-badge".
Themes might want to set different colors on the badge if the displayed
color is light or dark. Use a style class for this when we set a color
on the swatch.
Rename get/set_color to get/set_rgba and show_alpha to use_alpha,
to match existing GtkColorButton API and let GtkColorButton implement
GtkColorChooser.
Implement popups in the editor, fix window sizing, fix RTL flipping.
GtkColorPlane is now using adjustments, and GtkColorEditor is using
adjustments as its model as well.
The custom colors are now pushed over to the right as new ones
are added, and everything beyond the 9th row gets dropped.
Customized colors are added to the custom colors array.
There's no reason this should be a focus ring rather than an actual
frame. In the past this was probably used to get a dashed effect, but
now we even support that natively for borders.
This fixes issues where the new default bg of transparent
didn't work, making offscreen windows black.
I don't think this is a practical performance problem.
Offscreen windows are rarely used and generally used for
graphics tricks like alpha anyway.
use direct access to the RBTree instead of fiddling with the path.
Removes the lat caller of get_index() and with it count_rows(), so those
are gone now.
Refcounting behavior kinda requires that we track of accessibles until
the row goes away. This is necessary for reference handling and for some
assumptions that atk-bridge does.
When moving a page around, all children changing their position
need to be notified.
There are still other places where proper notification is missing
(drag 'n drop, etc.)
https://bugzilla.gnome.org/show_bug.cgi?id=669116
Unsetting the widget causes events to get emitted and we want to
(a) really emit that event to our signal handlers
(b) still be recognized as a real GObject while doing that
https://bugzilla.gnome.org/show_bug.cgi?id=669794
Otherwise we fail to link, seeing this:
/usr/bin/ld: gtk3_widget_factory-widget-factory.o: undefined reference
to symbol 'gdk_pixbuf_new_from_resource'
This implements the following part of the EWMH spec:
"The special value _NET_WM_MOVERESIZE_CANCEL also allows clients to cancel the
operation by sending such message if they detect the release themselves
(clients should send it if they get the button release after sending the move
resize message, indicating that the WM did not get a grab in time to get the
release)."
In particular, it fixes the case of clicking widgets that use
gdk_window_begin_[resize|move]_drag*() and the click "sticking", i.e. the
mouse button getting released but the resize or move operation remaining in
effect.
https://bugzilla.gnome.org/show_bug.cgi?id=669208
It was problematic to maintain Raleigh going forward, as any
changes in it affected all themes. Also, its more robust if
each theme is a full standalone css rather than relying on
an inherited css base.
So, this changes Raleigh to a standalone theme that we can tweak
without accidentally breaking other themes, and makes the
default theme empty. In fact, we don't even add the default
provider anymore as its always empty.
Sometimes we need to read back the window content into our double
buffer due to rendering a window with alpha when there is
no implicit paint or it has been flushed due to non-db drawing
before.
However, in this case we can't use gdk_cairo_set_source_window as
it might trigger an implicit paint flush as we detect what we
think is a direct non-double buffered window draw operation, which
will flush the implicit paint operation that we're just setting up.
To fix this we use the raw gdk_window_ref_impl_surface operation
to get the source surface.
There was a sign issue in a coordinate transform that made us
flush the wrong region when flushing an implicit paint.
The non-double buffered drawing would then be drawn over the
right area, but then at the end of the implicit paint this
would be overdrawn with the area we didn't properly remove
from the implicit paint.
Also, the translation from window coords to impl window
coords is now done before removing any active double
buffered paints, as these are also in impl window coords.
With the changes in default CSS to make the default background transparent
we ran into issues where intermediate GdkWindow (for instance the
view_window in GtkViewport) where we didn't set an explicit background
(because before they were always covered). So instead of showing throught
the transparent windows were showing the default backgroind of the intermediate
window (i.e. black).
With this change we also needed to fix GtkViewport, as it was previously
relying on the bin and view windows to cover widget->window so that the
border was not visible if shadow_type was NONE.
NULL is an actual useful value for
gtk_action_group_set_translation_domain, but this information is buried
in g_dgettext() documentation.
Be extra redundant, to make it obvious that NULL is ok.
https://bugzilla.gnome.org/show_bug.cgi?id=669636
When using GtkTreeStore, don't compute row positions assuming
everything's a list store, which is how get_row_start_for_index() works.
So instead, do a laxer check and don't insist on proper alignment.
The freeing the GResource in gtk_css_provider_reset caused problems
because it was called from gtk_css_provider_load_from_path() inside
gtk_css_provider_get_named(). We fix this by delaying the
setting of priv->resource until we successfully loaded the file.
This lets themes install a .gresource for efficient loading
of resources. This resource file will (if it exists) be automatically
loaded (mmaped) when the theme is used and unloaded when not used.
"border-style: none;" sets border widths to 0. So no border was drawn.
Use "border-style: groove" instead. (If the border was drawn that way,
it'd look noticably different.)
Also include reset-to-defaults.css so we really get default CSS.
Checking if the value is NULL is the wrong thing to do - the bitmask is
usd to keep track of that.
The reason for that will become apparent in the next patch.
Fixes reftests that were broken after images and labels applied padding.
Things look much more cramped now. On the plus side, treeview headers
and comboboxes have better spacing. Menubars look like crap now though.
Needs someone with artistical talent to figure out.
Just set oit for things that need it.
(FIXME: Buttons should probably not need it, but spinbuttons and
treeview headers don't have a proper background without it.)
Hold a ref to the GtkDialog while doing async operations with the
GtkAppChooserOnline object.
This is needed, since somebody could call gtk_widget_destroy() on us
while an async operation is in progress. We don't want to be finalized
in that case, but mark the fact that we were dismissed and just return
from the callback in that case.
This avoids crashing if the dialog is destroyed in the middle of a PK
operation.
https://bugzilla.gnome.org/show_bug.cgi?id=649121
Now that ATK no longer uses a key snooper but is invoked directly,
checking in advance for existing snoopers is wrong and stops ATK from
working.
Also: code reduction without performance loss == good thing.
https://bugzilla.gnome.org/show_bug.cgi?id=669176
When we're size allocating the children widgets, always trim the
allocation for padding and border of GtkComboBox, as that's all the area
they have available.
It was using the wrong padding values for RTL, and generally, using the
same variables in a distant macro makes this method even more
complicated than it already is...
If we're in menu mode with no children, instead of tweaking the
allocation of the arrow and the separator separately, tweak the
allocation of the box that contain both.
GtkComboBox always handles focus through its children, so there's no
need to request and allocate them again.
In fact, Adwaita was forcing these style properties to zero for
GtkComboBox.
The progressbar inside GtkEntry has a progress-border style property,
which is actually the margin of the progressbar inside the GtkEntry
allocation.
Use a CSS margin instead of reading the progress-border property.
Don't assume the padding of the icons is the same padding of the rest of
the entry.
This also allows to set different paddings for left and right icons.
This commit fixes crash which occurs in Firefox, Thunderbird and Inkscape
during printing. This crash was caused because of wrong handling of Custom
CUPS options. (#543520)
Replace the (invalid) DTD in the GtkBuilder docs by a
RELAX NG schema. Also install the schema in /usr/share/gtk-3.0,
so it can be used to validate GtkBuilder ui files.
The timeline frame function emits several signals, assuming that the
timeline will exist after these signals return. This assumption can be
invalid if signal handlers unref the timeline.
https://bugzilla.gnome.org/show_bug.cgi?id=668675
Change the format of GtkBuilder <menu> to be more in-line with the style
of the rest of GtkBuilder so that we can do translation in a consistent
way.
The format is now substantially more difficult to hand-write, but tools
should be along soon.
There is an xslt program attached to the bug to help you convert your
existing .ui files from the old format to the new one.
https://bugzilla.gnome.org/show_bug.cgi?id=668696
This commit adds API that allows to add new named sections
to the Credits part of GtkAboutDialog, in addition to the
hardcoded sections for authors, documenters, artists and
translators.
https://bugzilla.gnome.org/show_bug.cgi?id=484693
When calling gtk_window_present(), gdk_win32_window_raise did not
actually raise the window anymore. Replacing BringWindowToTop() with
SetForegroundWindow() fixes this.
During testing, we also discovered that sometimes SetForeGroundWindow()
will (correctly) refuse to raise the window and fail(for example: sometimes
when dragging a different application at the time of a gtk_window_present()
call). To prevent a GdkWarning from being produced, usage of the API_CALL
macro has been removed for this case.
Additional goodies of SetForeGroundWindow:
- it brings the window to the front when the process owning the
window to raise is the foreground process (for example when
gtk_window_present is called from a GtkStatusIcon's activate
signal handler)
- it limits itself to flashing the task bar button associated
with the window if the process owning the window to raise
is *not* the foreground process (for example when gtk_window_present
is called from a g_timeout_add callback function)
https://bugzilla.gnome.org/show_bug.cgi?id=665760
Otherwise, we could sometimes fail to update the cursor node when the
right row was deleted.
Also, I'd like to file a formal complaint that this node/tree
differentiation makes writing comparisons too complicated.
https://bugzilla.gnome.org/show_bug.cgi?id=668169
GtkApplication adds to the global accel map using the prefix <Actions>
which is also used by GtkAction. This causes GtkApplicationWindow to
try to parse GtkAction-added accels as if they were its own (which
fails).
Switch to a different namespace -- <GAction>/.
https://bugzilla.gnome.org/show_bug.cgi?id=668367
The new semi-private function will allow to implement support for css
padding and border in widgets inheriting from GtkMisc.
Use the new function for GtkLabel, GtkArrow and GtkImage.
This reverts commit 0c8ecba7dc.
The change broke Orca completely, and we need a proper fix first.
So we have to live with the bugs intorduced by this until then.
Due to changes in the behaviour of the virtual modifiers around MOD1 the
implementation of the map_virtual_modifiers vfunc was mangling the modifiers
and making keybindings not work correctly.
This change updates the implementation to match the X11 implementation's
behaviour.
Callers of this function were passing in -1, -1 for the width and height if
they just wanted the window moving - rather than the size changing. We need to
respect that behaviour and don't try and set the width/height to those
dimensions.
Now pop-up windows (ala combo-boxes) work!
We don't want to include the whitespace after an import statement.
Because people tend to add newlines after @import, we report the wrong
lines for failed imports. And that's bad.
"The child window MUST also have the _NET_SUPPORTING_WM_CHECK property set to
the ID of the child window. […] If the _NET_SUPPORTING_WM_CHECK window on the
client window is missing or not properly set, clients SHOULD assume that no
conforming Window Manager is present."
This commit implements that, which allows us to not have to do a
XGetWindowProperty() every N seconds when running under a compliant WM.
This is also a more correct fix for the bug handled in commit
daf29bffed.
https://bugzilla.gnome.org/show_bug.cgi?id=666921
There are new calls to load ui files from resources, and
you can now read pixbufs by using relative paths in a ui
file read from a resource, or by using absolute resource:/// uris.
Was getting this error:
main.c:785:13: error: ‘gtk_demos’ undeclared (first use in this function)
main.c:785:13: note: each undeclared identifier is reported only once for each function it appears in
main.c: In function ‘main’:
Problem pointed out by Stefan Sauer in bug 667598. The solution
here is different from his patch. We always draw marks in increasing
direction, and flip the marks and stop positions to match.
When finding the width for a given height, we can pass the full
height to both the menubar and the content. Instead, give the
menubar its minimum height, and give the rest to the content.
The code was not properly removing the menubar height from the
height given to the content, causing the statusbar to drop off
the bottom of the window in some examples.
This follows the approach used by the Quartz port - that of a separate
implementation matching GtkClipboard.
The simple clipboard tests in gtk3-demo function correctly but there are
almost certainly leaks and bugs.
This is useful to sketch out in GtkBuilder widgets in different states
all at once, so that we can check theming is right for them.
Add some initial UI files for primary-toolbar and inline-toolbar widgets.
gtk_window_get/set_attached_to() is a new API that allows for windows to
be attached to a GtkWidget.
The attachment is a logical binding between the toplevel window and the
widget that generated it; this kind of information is currently used to
propagate style information from the widget to the window, but is also
useful e.g. for accessibility.
https://bugzilla.gnome.org/show_bug.cgi?id=666103
Instead, focus nothing and wait until we get focus before doing so. This
restores previous behaviour but still emits proper cursor-changed
events.
Fixes a bunch of bugs in the filechooser which populates the treeview
asynchronously.
https://bugzilla.gnome.org/show_bug.cgi?id=613728
It was always looking at the 'color' keyword for the parent context
instead of only doing that when used with the 'color' property.
Tested by border-color-default.ui reftest.
Add a switch. This demonstrates:
- that switches can be placed in toolbars
- that GtkSwitch is actionable
- that actions can be shared between multiple actionables
We don't expose ::quit-requested as API anymore. Instead, we expect
users to register inhibitors when needed. Without quit-requested,
there is no need for ::quit-cancelled and gtk_application_quit_response
anymore.
We still emit ::quit when the application is about to quit.
The part in gdkwindow-x11.c which uses XIEvent and friends is "protected"
by HAVE_XGENERICEVENTS, i.e., XGetEventData() has been found. (Xlib.h)
XIEvent and friends are defined in <X11/extension/XInput2.h> which is
included by gdkdisplay-x11.h if XINPUT_2 is defined.
The patch makes sure XIEvent is only used if XINPUT_2 is defined.
https://bugzilla.gnome.org/show_bug.cgi?id=667534
Replace all references to g_application_set_app_menu and
g_application_set_menubar by their gtk variants, which
actually exist. Pointed out in bug 667546
'window-unfocused' is too long and mentions "focus" which is historically
loaded with the meaning "input focus".
'backdrop' isn't generally used in GUI speak and still conveys the state the
widgets in an unfocused or background toplevel window are in.
This reverts commit c276f53796 and
implements the same feature using the compute function.
A nice side effect is that gtk_style_property_get_border() and
gtk_style_property_get("border") to the same thing now.
When using G_TYPE_NONE as the shorthand's type, the property is not
gettable/settable using regular APIs, so it essentially doesn't exist.
As it should be.
We want to make sure to fall back to the initial value of the property
if resolving fails.
This is kinda tricky, because the initial value itself might need
resolving, too.
Instead of on-demand resolvage, we now resolve during lookup. The step
is done via
_gtk_css_style_property_compute_value()
which currently calls into
_gtk_css_style_compute_value()
That function has all the old resolving machinery.
The only part missing for now is the handling of win32 code. It will be
added back later.
This would make the internal properties custom properties. And they are
not.
Also get rid of default initial values, we now specify the correct one
at all times.
We can't print shorthands, so don't try.
In particular, I want to get away from shorthands being representable
using GValue, and this function kinda requires that.
... in gtk_style_properties_lookup_property(). Those properties will
soon stop having pspecs or even being gettable and settable, so better
not tell anyone about it.
One test has all properties set to 'inherit', the other to 'initial'.
This should result in the same result, as inherit will ensure every
widget inherits the same value, and for the toplevel 'inherit' is
defined as being identical to 'initial'.
Passing the length of the widget path that is relevant is not necessary
anymore, it was only useful for inheritance. Instead, we now pass the
state flags and let the selector do the state matching for us.
See inline code comments taken from
http://dev.w3.org/csswg/css3-cascade/#cascade
This now respects the special values "inherit" and "initial" properly.
Note that those values cannot be parsed yet. This will be added in a
future commit.
We now use the GtkStleProviderPrivate interface, which hopefully is
faster and more conformant to CSS. Long term, it definitely should be
both.
I would have liked to split this up into multiple commits, but couldn't
find a way.
The GtkScaleMark values are gdouble, a simple a-b compare func would fail for
values with the same integer value. This breaks the sorting and causes random
marker label placement.
When we receive a configure event from Wayland telling us to resize our
surface we check against the geometry constraints for the window to ensure we
do not resize below the minimum and maximum limits.
This also removes setting the FOCUSED state flag when
gtk_window_has_toplevel_focus() since this effect can now be done with the new
WINDOW_UNFOCUSED flag instead which actually works better regarding X grabs
and modal windows.
Setting state flags is actually needed here since this function is called by
GtkButton subclasses which add their specific state flags as a parameter.
This reverts commit e868b8d6ea.
Now that grab and ungrab vfuncs are implemented on GdkDevice then we can use
gdk_device_ungrab to break the implicit grab created by the button press that
triggered the resize and move.
The function definition used a pointer to the enum value rather than the enum
itself.
This broke the build on platforms that don't have an implmentation of these
functions.
This is the interface for GtkWidgets that can be associated with an
action on a GtkAppicationWindow or associated GtkApplication.
It essentially features 'action-name' and 'action-target' properties
with some associated convenience API.
This interface is implemented by GtkButton and GtkToolButton.
https://bugzilla.gnome.org/show_bug.cgi?id=667394
This lets applications block logout and similar actions ahead
of time. Currently only implemented for D-Bus, but Windows has
very similar API since Vista.
This is fairly basic, allowing applications to learn when
the session manager is about to end the session, and possibly
block this. The only implementation at this point is using the
org.gnome.SessionManager D-Bus interface of gnome-session. It should
be straightforward to port the EggSMClient implementations for
Windows and OS X.
An implicit grab is created inside GTK+ when the button is pressed down on a
window. The semantics of wl_shell_surface_resize means that you don't get a
corresponding release event that would ordinarily break the implicit grab. So
we must do it as part of the resize request.
The code for moving the menu into monitor / workarea was duplicated,
once for the push-in scenario and once for without. The problem with
the second case is that we've stored the menu position before adjusting
it. That made us remember an out-of-monitor position that then later
triggered _another_ copy of this code in the size-request implementation.
Unify this to only have one copy of code, and only store the menu
position after adjusting it to be inside the monitor. This fixes both
statusicon menus that get popped up from the panel, outside the workarea,
to not have scroll arrows, and the gedit language menu which was not
placed in the monitor at all after the initial workarea commit.
As a side-effect of this change, we now make large scrolling menus
occupy the full height of the workarea. Before this change, we were
keeping either the top or bottom edge put while shrinking the menu
to fit in the monitor.
https://bugzilla.gnome.org/show_bug.cgi?id=667249
Instead of "attribute with later start index wins, make sure the
attribute list that is merged from takes precedence. This now gives the
multiple attribute lists we use in the label an order:
1) gtk_label_set_attributes()
These attributes override everything. It's what the function's there
for after all.
2) markup of label
Other user-specified attributes come next.
3) attributes for links
When we apply custom attributes on parts of the text, we put them
last. We don't want to mess with what the user does. Also, we change
color and underline, so we usually have something to show.
- Don't compute link color attributes until layout creation
This is useful as a performance enhancement, because we don't have to
lookup the property after setting the text, so multiple markup sets
don't cost style lookups.
- Don't merge attrs into effective_attrs
We do this when applying link colors now. Keeping them separate allows
invalidating them separately.
Instead of getting confused by applied underline or color tags in the
regular markup, we store the link start/end when we actually parse the
text. As a bonus, we can avoid rescanning links when creating the
markup.
As previously announced, key snooping is something that we should not
support in GTK. In particular because AT-SPI made every key press a
synchronous DBUS call / spawned a nested main loop with the key
snoopers. This breaks responsiveness, order of key presses and internal
consistency when the AT decides to remove key events.
This patch modifies GtkPrinterOptionWidget to support loading them
again. It also allows the user to enter values from the dropdown list
to select the item.
This change follows on from a change in semantics in Wayland where calling
wl_input_device_attach with nil would make the compositor set the pointer
sprite to it's default cursor sprite.
Presently, GtkComboBoxText lets you list the items for it in the
GtkBuilder markup like so:
<items>
<item translatable='yes'>Red</item>
<item translatable='yes'>Blue</item>
</items>
but has no way to set the "id" column for added items. Add an id='' tag
for that purpose so that you can do:
<item id='red' translatable='yes'>Red</item>
https://bugzilla.gnome.org/show_bug.cgi?id=667319
When doing DnD between windows the tab label is set as the child of
the DnD info window. If this is happening the remove method of
GtkNotebook should not unparent the tab label, since it's been already
unparented, belongs to the DnD window and will be properly destroyed
when the DnD is over.
https://bugzilla.gnome.org/show_bug.cgi?id=639875
When the X server does not support the shape extension (as some
vnc implementations seem to), our DND code was always seeing
an empty input shape, so drops always missed their target.
http://bugzilla.gnome.org/show_bug.cgi?id=620240
Apparently the deprecation warning macros are placed into gdkconfig.h
during the configure stage, so put these in the pre-configured
gdkconfig.h.win32 as well, as their definitions are needed for all builds.
When switching to the compact table format, we inadvertedly lost
the 'tentative match' handling which allows to complete longer
compose sequences even if a prefix is a complete sequence.
This was pointed out in bug 666710.
This way we don't need to compute them every lookup. (That's not the
real reason though - the real reason is that I want to add new APIs that
require the caching because they return consts).
Reading a card32 property into a long may lead to undefined high
bits, so mask them off. Also, make the conditions for setting and
unsetting the stick flag opposites, to avoid unintended changes.
Patch by John Lindgren, bug 666842
The current EWMH has added 'source indication' fields to a number
of client messages. Set these to 1 to indicate a regular client.
Also fill the button field of _NET_WM_MOVERESIZE.
Previously we expect that a proxy widget holds a reference to the
action via gtk_activatable_do_set_related_action(). However, it
is possible for the widget to still be in a floating state when
it adds itself to the the action.
This is a problem when gtk_action_get_proxies() gets called, because
we return floating objects back to the user. And language bindings
aren't going to be expecting that.
Fix this by calling ref_sink() and unref().
https://bugzilla.gnome.org/show_bug.cgi?id=657367
Since we allocate the standard CSS border to the button now, and center
the child accordingly, there's no need for an additional inner-border
style property. Deprecate it and ignore its values.
https://bugzilla.gnome.org/show_bug.cgi?id=666600
==23282== 64 bytes in 2 blocks are definitely lost in loss record 8,069 of 13,389
==23282== at 0x4A074CD: malloc (vg_replace_malloc.c:236)
==23282== by 0x39A1C3E2EA: cairo_region_create (cairo-region.c:196)
==23282== by 0x6D9AF3D: recompute_visible_regions_internal (gdkwindow.c:964)
==23282== by 0x6D9B4B8: recompute_visible_regions (gdkwindow.c:1126)
==23282== by 0x6DA3450: gdk_window_hide (gdkwindow.c:5689)
==23282== by 0x6D9CED9: _gdk_window_destroy_hierarchy (gdkwindow.c:2042)
==23282== by 0x6D9D040: gdk_window_destroy (gdkwindow.c:2109)
==23282== by 0x655B5E4: gtk_entry_unrealize (gtkentry.c:3012)
==23282== by 0x7068BF3: g_cclosure_marshal_VOID__VOID (gmarshal.c:85)
==23282== by 0x706710B: g_type_class_meta_marshal (gclosure.c:885)
==23282== by 0x7066DF9: g_closure_invoke (gclosure.c:774)
==23282== by 0x7080585: signal_emit_unlocked_R (gsignal.c:3340)
==23282== by 0x707F619: g_signal_emit_valist (gsignal.c:3033)
==23282== by 0x707FB71: g_signal_emit (gsignal.c:3090)
==23282== by 0x679E243: gtk_widget_unrealize (gtkwidget.c:4458)
==23282== by 0x64E83C7: gtk_bin_forall (gtkbin.c:172)
==23282== by 0x6548BBD: gtk_container_forall (gtkcontainer.c:2014)
==23282== by 0x67A966D: gtk_widget_real_unrealize (gtkwidget.c:10253)
==23282== by 0x672D002: gtk_tool_item_unrealize (gtktoolitem.c:474)
==23282== by 0x7068BF3: g_cclosure_marshal_VOID__VOID (gmarshal.c:85)
https://bugzilla.gnome.org/show_bug.cgi?id=666552
This allows drawing calendar arrows in all possible states the main widget may
be in.
The arrow_state array is converted into a bit field since it only really needs
to store boolean information about prelight for each arrow.
We add the app-menu and menubar public APIs to GtkApplication while
leaving the implementation in GApplication.
The actual implementation will be moved soon.
Deal with remote action invocations correctly by implementing
GRemoteActionGroup in the same way that GApplication does and pushing
remote activations through the before/after_emit functions of the
GApplication associated with the window.
This is the last part of getting the threading situation right.
https://bugzilla.gnome.org/show_bug.cgi?id=665737
When we have incoming activations or action invocations we should
acquire the GDK lock, just in case the program in question is using gdk
threads.
https://bugzilla.gnome.org/show_bug.cgi?id=665737
When parsing GMenu xml, use the facility for pick up named submenus.
Note that <section id='foo'>...</section> is _not_ a named submenu.
It is a named item with an unnamed submenu :-(
We want accelerators to work, even if the menus are not shown
locally, so we can't rely on the GtkMenu code to set them up
for us. Currently, this code only installs accelerators when the
window is realized.
This allows to specify accelerators in a convenient way, by
adding accel="..." attributes to the menu items in the
XML for the menu.
Currently, the accelerators are extracted only once, when the
menu property is changed on the application.
Move internal accel map API there and update all users.
Also, add an internal function to create an accel path for
an action and parameter, and use it in gtkapplication.c and
gtkmodelmenuitem.c instead of duplicating that code.
Put this in a separate file and substantially refactor it.
Move handling of submenu creation into gtkmodelmenuitem where it
belongs.
Improve our handling of when to show separators or not.
GtkModelMenuItem, being based on GtkCheckMenuItem, was always reserving
space for a check/radio indicator. This ends up looking pretty strange
in a GtkMenuBar.
Don't reserve space for the indicator unless it's appropriate.
This GtkMenuItem subclass (and GActionObserver implementation) contains
all the knowledge necessary for converting a GMenuModel item description
into a GtkMenuItem.
Remove much of the code that used to do this from
gtkapplicationwindow.c.
This feels premature; we do have the fallback situation covered
adaequately with the menubar, and people can do their own creative
solutions with gtk_application_window_get_menu(), so we don't have
to offer a widget for this right now.
By making "about" a per-window action, we can pop the about dialog up on
top of the correct window instead of trying to guess from the
application list of windows.
This is a GtkWindow subclass that "application windows" will use. Each
is associated with a GtkApplication, has the ability to show menus and
will have its own associated set of actions.
These were destined for GLib, but they don't really make sense as a
public API. Instead, we'll copy/paste them around between the various
codebases that need to render menus.
This is a boolean property that will be set to TRUE if the current
desktop environment is capable of displaying the application menu as
part of the desktop shell.
If it is FALSE then the application will need to display the menu for
itself.
This should be redone to show the label together with the
separator line, somehow. For now, just put the label below
the separator, as a separate item.
This should be redone to show the label together with the
separator line, somehow. For now, just put the label below
the separator, as a separate item.
This makes GtkBuilder accept a GMenuMarkup tree at the toplevel
(ie with <menu id='foo'> being a child of <interface>) and the resulting
GMenu object can be obtained via gtk_builder_get_object (builder, "foo").
We need to connect to items-changed on _every_ menu
model, which is somewhat icky. For some reason, this
works fine with a local model, but not with D-Bus in
between. Debugging needed.
We want to avoid handling focus events for the private focus window,
otherwise the keyboard grab taken by for example buttons will cause a
spurious FOCUS_OUT/FOCUS_IN on the toplevel.
The code that did this seems to have been lost in the XI2 transition for
GTK3.
https://bugzilla.gnome.org/show_bug.cgi?id=657578
I expect them to be used a lot, so this approach seems better than
requiring signals that connect to "notify::widget". Also, we can't use
regular functions (like dispose or constructed), becaiuse those assume
that (un)setting of the widget only happens once and with the current
design (a puble set_widget() function) we can't really guarantee that.
Also, I split them into two separate functions as one function is part
of construction and the other part of destruction of the object. And it
doesn't sound like a good idea to have that both be part of one
function.
I wanted to use weak refs to ensure this instead, but it's a performance
problem when used in the treeview and each and every cell adds a weak
ref to the treeview.
The function returns the part of a monitors area that should be
used for positioning popups, menus, etc. The only non-trivial
implementation atm is in the X backend, all the other backends
just return the full monitor area. The X implementation is
currently suboptimal, since it requires roundtrips to collect
the necessary information. It should be changed to monitor
the properties for changes, when XFixes allows to monitor
individual properties.
https://bugzilla.gnome.org/show_bug.cgi?id=641999
Dealing with disabled devices may turn into hierarchy/device
changed events on device IDs with no backing GdkDevice yet,
so protect against that. The device attachment will be handled
correctly when the device is enabled later.
Even if we can't change our sensitivity because the parent is insensitive we
should still flip the sensitive flag. Otherwise, with and insensitive parent,
child.set_sensitive(True)
...
parent.set_sensitive(True)
would result in child still being insensitive.
https://bugzilla.gnome.org/show_bug.cgi?id=666392
Instead of in the key_press handler, use a signal handler. The signal
handler runs before the default handler, and before other signal
handlers. In particular it runs before the signal handler installed via
gtk_entry_set_completion() which pops down the entry completion, and we
don't want that to happen.
The code does not change the code of the handler in any significant way,
it just refacotrs it to not call the parent anymore.
https://bugzilla.gnome.org/show_bug.cgi?id=663573
This was removed previously, but is necessary to support the case where
no base folder is in use, which should cause an error. This can be
experienced in save mode in the recent files list.
https://bugzilla.gnome.org/show_bug.cgi?id=663573
That way, we can use the regular filtering features of the
filesystemmodel instead of having our own filtering function that
duplicates much of the matching code.
This also removes the broken-on-windows feature where files strting with
a dot were not autocompleted.
Use all the clever functions we invented in recent refactorings to
compute the file and dir part and the current folder.
This also fixes the bug where "./" was not taken as the current
directory but interpreted as a file named "./".
This returns the text that should be completed on. As this is somewhat
tricky to compute (and in fact one place did it wrong), let's make it a
function.
Now that we don't need to know anymore why insertion failed, we can
simplify the append function quite a lot.
A simple rule now: If we did not insert anything, beep.
Too much special code for no gain.
I did a totally unscientific questionnaire on GNOME IRC (assuming
autocompletion is for advanced users) and nobody even knew what this is.
So I suspect it's useless. Also, it's positioned completely wrong anyway
and nobody noticed so far. Last but not least, I'm trying to imitate
bash here and bash doesn't show feedback.
Use dispatch_properties_changed() instead of GtkEditable to watch for
changes to the to-be-completed text. This is necessary because various
functions in GtkEntry don't use the interface vfuncs and one only
becomes aware of them via notifications. I'm not sure this is intended
behavior, but it's how it works today.
Also, use the same code for triggering in all situations.
What the code does is this: It looks at the part of the text in front of
the cursor (or selection) and completes for it. Once the directory has
been enumerated, inline completion is activated. Note that popping up of
the completion popup is completely handled by GtkEntry.
Instead, pass the text to use to refresh_current_folder_and_file_part().
This also gets rid of the problem introduced earlier that the position
is not properly updated in do_insert_text() and therefor the completion
is wrong.
... from GtkEntryCompletion instead of implementing a poor copy
ourselves. This also makes the file chooser entry behave a lot closer to
normal entries.
This is identical to the display name when not having a path that
changes the folder. Otherwise it will have the full path that was
entered in the entry. Say when from your home dir, you type
"../../usr/li", the full path for "lib" and "lib64" will be
"../../usr/lib" and "../../usr/lib64" respectively. This value isn't
used yet, but will be soon.
The new version does not need to muck with the entry, it just extracts
the required information. It returns a reference to the folder though,
as we extract the information now instead of returning something stored.
Setter functions the way I code it look like this (in order):
1) Figure out if the value changed. If not, exit early.
2) Clear the old value (ie unref stuff, disconnect signals
3) Copy the new value
4) Set up things about the new value
This reorganization does that.
And by doing that, it even reduces the amount of code and the amount of
branches (and with it, nesting) needed.
Instead, fall through to the default handler after closing the
completion. This has the advantage of letting the file chooser entry
capture the tab key properly, so one can't accidentally move out of the
entry by pressing tab while the completion is popped up.
I also suspect it fixes bugs with weird tab keys and shift/ctrl
oddities. But who knows...
1) always emit signals
Previously, newly constructed cells would be told to not emit events.
However, we can ensure that nothing is connected to the signals, so
they will not actually emit anything.
2) don't return anything
The return value is unused anyway.
No more signal handler is needed, therefore the code can also get rid of
tracking the treemodel. And we use a faster approach for iterating the
changed cellrenderers: We just iterate all columns instead of over all
cell accessibles, as that number is likely quite a bit smaller.
We first set a NULL model and then reset the old model, just to get the
effect of clearing and then resetting.
We reset the cursor and selection afterwards, so the reset doesn't
destroy all the work we did.
Multiple changes:
- actions are now available unconditionally, but only work in the right
state. This mirrors other actions.
- instead of adding actions manually, they invoke the action on the
GtkCellAccessibleParent interface.
Tests have been updated to reflect the changes
Treat the empty string as if it was a NULL icon name instead of
rendering GTK_STOCK_MISSING_IMAGE.
This used to be the previous behavior of GtkCellRendererPixbuf and
applications might still rely on it.
We need to save the icon size locally for GtkCellRendererPixbuf, since
its properties are set again every time it's drawn (resetting the icon
size on the helper by means of _gtk_icon_helper_clear() each time).
The code is very similar; having two slightly different code paths is
bad and can lead to bugs. Refactor the code to use the same height
request function.
For maximized windows, titlebars cannot be used to reposition or
scale the window, so if an application does not use it to convey
useful information (other than the application name), the screen
space occupied by titlebars could be put to better use.
Add a new window property which requests from the window manager
to hide titlebars when windows are maximized to account for this.
https://bugzilla.gnome.org/show_bug.cgi?id=665616
If the scale has an origin (it will have one by default), GtkRange will
render the two sides before/after the current value with different style
classes, making it possible for themes to use different colors and
properties for the two areas.
This was possible in GTK 2 with style details, but got lost during the
road to 3.0.
https://bugzilla.gnome.org/show_bug.cgi?id=665140
Setup listener functions for the drag and drop events as well as the selection
events. Then create and save a data structure representing the data offer from
the other client.
This implementation is based on a hash table. The hard coded GtkSelection
atoms are preloaded into the hash table at the correct values. User generated
atoms start after the last preloaded atom.
The data device manager is a global object that provides the support
infrastructure around data devices. These data device objects are the basis
for handling drag and drop as well as selections in Wayland.
Reset state of CUPS requests correctly during authentization and
check CUPS requests for errors.
Don't initialize variables holding password and username with empty
string (#664640).
As pointed out in bug 665999, these were just not right.
Before this commit, the nicks were 'output' and 'only'.
After this commit, they are 'input-output' and 'input-only'.
In many cases we used to set focus_child to NULL all the way up
to the top and then to the right value, even if there was
a common ancestor, meaning these see a temporary NULL value for
focus_child. Only when the new focus widgets direct parent was
in the previous ancestor list did we stop early.
This fixes that by always stopping propagation when reaching
the common ancestor.
The casting used to calculate the child allocation is confusing MAX().
As a result, width and height end up with negative values, which makes no sense.
https://bugzilla.gnome.org/show_bug.cgi?id=666109
There was a corner case where the changed signal was not emitted.
If rows were built like this:
1 (not selected)
+ 2 (selected)
+ 3 (not selected)
And row 1 was removed, no signal would be emitted.
I like it when writing tests actually finds bugs that have been around
since 2003 - introduced by 4a03ea2334
actually. :)
We get certain cases, in particular with SELECTION_MULTIPLE, where we
cannot figure out in advance of real_set_cursor() if the selection will
actually change.
Previously, the cursor would just become invalid, which used to
reselect the first row in the treeview later on (without a
cursor-changed signal). This leads to a crash now with the recent
refactorings.
The patch is longer than I'd like it to be, but the situation is kinda
complicated, because we want to make sure to move the cursor to a good
row. It works like this:
1) From the deleted row, search forward for the first row that is not
going to be deleted and is not a separator.
2) If no such row exists, search backwards for a row that is not a
separator.
3) If no such node exists, clear the cursor.
Previously the code used a GtkTreeRowReference, which was (a) less
performant and more importantly (b) hiding errors.
The errors being hidden were the referenced row becoming invalid or
collapsed, because such rows would not be valid cursor rows and it would
be necesary to select a new row and emit a "cursor-changed" signal.
So if a crash bisects down to this commit, it is very likely that the
cursor has not correctly been updated and the cursor row is invalid.
This means we don't have to have hardcoded "/usr/share/wayland" to find the
cursors.
This change also fixes up the warning messages for when loading fails.
Calling gdk_keymap_add_virtual_modifiers causes _all_ virtual
modifiers to be added, which causes problem when they are co-located
on the same real modifier (as Super and Hyper often are). Effectively,
this made it impossible to enter key combinations involving Super,
since they all turn into Super+Hyper.
Move code for child allocation into the child allocation function. Don't
keep it in the overlay allocation code.
See the next commit for why this is useful.
When an implicit paint is flushed during expose, e.g. because a
non-double buffered widget is painting, make sure to copy the existing
data from the window surface we rendered before flushing back to the
paint surface, instead of using an empty base.
Code was already handling that (and said so in the comment), but only
when no implicit paint was used at all, and not in the case when it's
flushed mid-expose.
We used to only set the override color or font if it was different
than the existing value. However, that means you can't change it to
an override that is the same as the default value for the property.
With this fixed you can e.g. override with a color of 0,0,0,0 which
you couldn't before.
Previously all the commands that acted on the shell took the surface that was
to be acted on as parameter. Now we retrieve an object from the shell that
represents its state for the surface. With that wl_shell_surface object we can
then call methods on that.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
When gtk_render_arrow() is called, always add an ARROW style class to
the GtkStyleContext before rendering, so themes can specify a different
color for it.
https://bugzilla.gnome.org/show_bug.cgi?id=665420
Just always render the pixbuf ourselves and set it on the GtkImage in
the X11 case.
Code for other backends was already rendering the pixbuf manually before
translating it into a native type, so this greatly simplifies the code.
https://bugzilla.gnome.org/show_bug.cgi?id=665409
Instead of painting the window background on the grip_window we now
only paint it on the GtkWindow->window, and we make the grip_window
have a transparent background.
We can't really make transparent window handle background optional
via css atm, because the handle color is actually based on the
background color, so if that is set to transparent we won't draw
anything.
gdk_window_get_update_area is supposed to get the area where things
need painting, and remove them from the update areas. However, if
some area is covered by other windows with an alpha background we
can't just expect whatever the app choses to render in the update
area as correct, so we don't actually remove these areas, meaning
they will get correctly rendered when we get to the expose handlers.
gdk_window_move_region doesn't move children, so we can't copy
transparent child window regions with copyarea, so we remove these
from the copy region.
We track the areas that have alpha coverage so that we can
avoid using these as sources when copying window contents.
We also don't remove such areas from the clipping regions so
that they are painted both by parent and child.
This cleans up the expose handling a bit by using the existing
clip regions, and it allows us later to use painters algorithm
to do transparent windows.
If no icon size is provided (e.g. when setting icon-name or stock-id
using the GObject property directly), use the default icon size. This
matches the previous GtkImage behavior.
The new function provides an API that takes the PangoLayout and index
as input params, this way it handles strong and weak cursors internally
factoring out all code duplicated in the widgets that need to render
cursors.
https://bugzilla.gnome.org/show_bug.cgi?id=640317
Explicitely check the split-cursor setting when drawing the textview
insertion cursor instead of relying on the cursor_direction set in the
textlayout. This makes the cursor drawin code more uniform with other
widgets in preparation to refactoring the cursor drawing code in a
shared function.
https://bugzilla.gnome.org/show_bug.cgi?id=640317
Move the handling of primary/secondary cursors to gtktextdisplay, which
makes code simpler and more consistent to how GtkLabel and GtkEntry
draw cursors, which is useful in preparation to further refactoring.
https://bugzilla.gnome.org/show_bug.cgi?id=640317
GtkIconHelper is a helper object to easily obtain a pixbuf from
different icon sources (e.g. a GIcon, an icon name, a stock id, ...).
Code is ported from GtkImage, which will be adapted in the next commit.
It used to be that _gtk_path_bar_set_file() would return an error if
it wasn't able to switch to the specified file, but that hasn't been
the case for a long while now, since the file chooser became async.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
Previously, when the cursor was a descendant of the collapsed row,
the cursor path was set to the collapsed row, but this was not
communicated via cursor-changed events.
_gtk_tree_view_find_path() was not a name that not really described what
the function does. And I kept forgetting it. Also, it took the tree view
as an argument and that was completely unnecessary.
Rewrite GtkSpinButton to use a touchscreen friendly horizontal layout.
Other features include:
- support for theming buttons with nth-child
- full support for RTL
- use + and - symbolic icons instead of arrows
- general streamlining and cleanup of a lot of related code
https://bugzilla.gnome.org/show_bug.cgi?id=663359
From the css docs at http://www.w3.org/TR/CSS2/box.html:
8.5.1 Border width: 'border-top-width', 'border-right-width', 'border-bottom-width',
'border-left-width', and 'border-width'
Computed value: absolute length; '0' if the border style is 'none' or 'hidden'
So, if i specify border-style none and a border-width > 0 that should give the
same result as border-width 0.
From the css docs at http://www.w3.org/TR/CSS2/box.html:
8.5.1 Border width: 'border-top-width', 'border-right-width', 'border-bottom-width',
'border-left-width', and 'border-width'
Computed value: absolute length; '0' if the border style is 'none' or 'hidden'
This is a private type, and the CSS default value is "stretch", which
is how we were parsing and printing NONE anyway.
With this fix we can use the enum stuff for printing/parsing
Various tiny semantic changes in here.
The big fix is that we consider every cell as visible and as showing if
the treeview is mapped.
This was also fixed in the tree dump test.
Also, *do* add the file to the recently-used list, even if the file does not
exist yet. This is used from the Save dialog, so even shitty apps which don't
add the file to GtkRecentManager, will get the file added from the Save dialog.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
_gtk_rbtree_reorder() was moving the node's data while reordering. As we
use the node pointer in the a11y code as a hash key, this didn't work.
So this rewrite changes that. As a bonus, it is less code and faster.
Woohoo!
- Make sure the rotated nodes aren't nil
- Use existing functions for complex computations
- Don't use NULL checks for variables guaranteed to not be NULL/nil
This has one major caveat: The new value is const, so read-only memory.
Any attempt to write to it will cause a crash. Note that we are not ever
supposed to write to it, but bugs happen...
The code used to set nil->parent, which could cause segfaults. Don't do
that. We also need to pass the parent explicitly to the fixup code,
because the node during fixup may be the nil node.
So instead of copying the children and height to the new node, we keep
the old node and copy all the old stuff to it.
This is necessary so the accessibility code can use the node as a key in
the hash table or store the node as a reference to the row instead of
GtkTreeRowReference. And because it already does that (oops), this fixes
a bunch of segfaults with a11y enabled.
gtk_rbtree_adjust() will adjust the summed values of a node and all its
parents in the tree. Currently only implemented by splitting out the
function from gtk_rbtree_free().
Instead of taking the first page and trying to adjust the rect with
random padding values, take the current page, as it's always guaranteed
to be at least as tall as inactive tabs.
This fixes some annoying 1px drawing artifacts while switching tabs when
the theme disables notebook padding.
https://bugzilla.gnome.org/show_bug.cgi?id=664494
This turns GdkWin32DragContext into a proper GdkDragContext subclass.
Because we now correctly initialize GdkWin32DragContext in
gdk_drag_context_new, we no longer crash immediatly when a DnD
operation is initialized (the find_window, drag_status, ... vfuncs
where all pointing to 0x0 instead of their proper win32 implementations).
We now try to consistently refer to GdkDragContext as "context",
GdkWin32DragContext as "win32_context" and the ole2 related
target_drag_context and source_drag_context as "ctx".
Members of GdkWin32DragContext only used by the ole2 DnD codepaths
are now explicitly marked with a ole2_dnd_ prefix.
Similar to GtkNotebook, GtkFrame was only allocating space for the
padding width, and not the border.
This could be seen by just running tests/testframe. With a theme that
renders frame borders, setting xthickness = 0 in the test draws the
button border over the frame border, which is wrong.
https://bugzilla.gnome.org/show_bug.cgi?id=664342
Turned assertion into silent return.
This assertion is only hit when dragging from an empty tree view. In
this case, gtk_tree_view_begin_drag() is triggered from gtkdnd.c and not
from gtk_tree_view_maybe_begin_dragging_row(). We actually want to
cancel the drag at this point, but that is not possible with the GTK+
API as far as I can see.
The alternative is to not allowing the drag to start. This could be
done by simply unsetting the tree view as drag source when it is empty
and setting it as drag source again when rows are added. I didn't
choose to go with this for now, since this will likely break third party
code.
There was some regressions from the recent display line fix;
while fixing it improve the test coverage and make GtkEntry
and GtkTextView return identical results.
gdk_cairo_region_create_from_surface doesn't work correctly on PPC.
This is most prominently seen with the GTK window resize grip, the
shape of which is mirrored every eight pixels horizontally.
At the same time, use an A1 surface for the resize grip shape to
eliminates an A8->A1 surface conversion.
Add GDK_MODIFIER_INTENT_SHIFT_GROUP to enum GdkModifierIntent
and handle it in gdk_keymap_get_modifier_mask(). Add an X11
impl of the method and return keymap_x11->group_switch_mask.
Return 0 from the default impl because we don't know.
If the keyboard group shifting modifier is *also* a normal
accelerator modifier, we need to special case it when calling
gdk_keymap_translate_keyboard_state(), so we get the right
key symbol for accelerators (for example we want Option-O,
not Option-Ø displayed in menu items). This patch should only
affect quartz where the Alt key both shifts the group and can
be used as accel modifier, and not X11 or Win32 where AltGr
is not used for accelerators.
- fix quartz' gdk_keymap_translate_keyboard_state() to return
the right consumed_modifiers
- add _gtk_translate_keyboard_accel_state() which does the
special casing
- use it everywhere instead of gdk_keymap_translate_keyboard_state()
The default theme engine draws a fallback check/radio image, but
doesn't let you replace this. We now check if a background image
is set and if so render that instead of the default fallbacks.
We now support -gtk-win32-theme-part(class,part,state) in background
and border-image CSS properties. This renders the corresponding
theme part using DrawThemeBackground() and acts as a base for a
CSS based windows theme.
Note that we build the parsing code even on non-win32 so that
all themese will parse the same on all arches. We draw pink instead
of the actual theme parts on non-win32 though.
We want to move to a world where getting style properties can
allocate a new value, depending on the style property context.
In this world we can't expose "peeking" a property, as we neet
to return a ref to the property that is newly created.
So, we move the peek code into get_property and use get_property
from get_valist.
At the toplevel we have _gtk_theming_engine_get, which lets us pass
in a property context with (atm) the size to get the property for.
Then there is a lot of plumbing to push this down into the lower
layers of the style property code until finally hitting
the property resolvers.
I need this because I will be adding a property resolver for win32
theme parts, and they render differently depending on the size
(i.e. they don't scale linearly). The idea is that the code
to get the background properties will pass in the final size
and we will resolve the theme part specification to that particular
size.
If the old non-context calls are used we just hardcode a size
of 100x100.
We used to special handle gradient, but we want to be able to store
other sources (that will eventually resolve to cairo patterns).
For instance, this is needed to handle win32 theme part sources.
In gtk_tree_view_state_flags_changed() we were setting the background on
the bin_window without the necessary "view" style class, making the
treeview render with the wrong color in some circumstances (such as when
adding an empty treeview in a window, as spotted by Brian Smith).
We used to explicitly map and unmap the child GtkRecentChooserWidget when
mapping and unmapping the dialog, respectively. Now that GtkContainer actually
unmaps child widgets (instead of avoiding that), we can assume that the
child GtkRecentChooserWidget will be unmapped when we want it to be.
This fixes a warning from gtk_widget_verify_invariants(), as we were mapping
our child widget before calling our parent class' ::map() handler. Bug #659257.
Try to find keycodes for keysyms at level-0 in the keymap, but don't
fail to parse if we can find those in level-1. This fixes the
parsing of numerical keys on the keypad, which require a shift level
through Num-Lock.
https://bugzilla.gnome.org/show_bug.cgi?id=606727
This way, the a11y code knows if a column was reordered, added or
removed and can do the right things instead of trying to guess and
getting it wrong.
Also, this patch finalizes the changes so that only visible columns
exist to the accessibility interface.
In fact, invent our own function get_n_columns() which actually only
counts the visible functions, because that's the only ones we care
about.
The places where it's not used yet will be changed in the coming
patches.
.. as a replacement for _gtk_rbtree_node_find_parity(). Instead of 1 or
0, the function now returns the index of node in the complete tree
(counting from the root). And this is of course identical to the row
number.
Track the RBNode/RBTree instead of keeping a TreeRowReference. This is a
whole lot faster and less error-prone.
Also, notify the accessible of removal of rows before actually removing
them, so we have a chance to clean up.
This function returns the accessible if it already exists. This way we
can call functions on the accessible from the widget itself instead of
having to rely on signals.
Instead of requiring a special functio to be called before just removing
from the hash table. This simplifies code a lot that used to remove from
the hash table.
- Name it properly
- Make it just a set of cell_infos
Currently it's using direct hash, but as long as we don't actually
lookup stuff from it, who cares...
We implement get_cell_index() now, so it's no longer necessary.
As a side effect, we need a different index for our hash table (which
now has a wrong name, but that will soon change).
Ahem.
This is in preparation for future changes and is not meant to stay this
way. But I want to change the hash table's keys and this is way easier
when nobody is using them.
I could have tried to make GtkContainerCellAccessible implement
GtkCellAccessibleParent, but the current implementation of that
interface doesn't make sense for it.
Instead of just storing the least significant bit of the row number,
store the full row number. This will soon be useful for accessibility.
But CSS could like it, too.
Add missing GDK linking to GIR build and examples:
GISCAN Gtk-3.0.gir
CCLD gtk-query-immodules-3.0
./.libs/libgtk-3.so: undefined reference to `gdk_keymap_get_modifier_mask'
./.libs/libgtk-3.so: undefined reference to `gdk_modifier_intent_get_type'
./.libs/libgtk-3.so: undefined reference to `gdk_window_begin_resize_drag_for_device'
./.libs/libgtk-3.so: undefined reference to `gdk_event_triggers_context_menu'
collect2: ld returned 1 exit status
CCLD grid-packing
../gtk/.libs/libgtk-3.so: undefined reference to `gdk_keymap_get_modifier_mask'
[...]
https://bugzilla.gnome.org/show_bug.cgi?id=664027
The default (out) transfer mode is "full", but the passed "model" pointer gets
set to gtk_tree_view_get_model() which is "transfer none". This caused Python
programs to free the model after calling this, causing crashes.
g_checksum_get_digest checks to ensure that the passed digest_len is long
enough to hold the digest, before setting it to the actual length of the
digest returned. Digest_len is uninitialized in the code, so if you're
lucky it will be larger than 20 and everything will work fine. If you're
unlucky, g_checksum_get_digest will return either -1 or some number less
than 20, and the g_assert(digest_len==20) will fail.
Wake up the run loop unconditionally (don't check if it is waiting) because
it might go into waiting state right after we checked for it. Fixes GIMP
startup (which has a lot of GIOChannel I/O but zero NSEvents) from several
minutes to a few seconds.
(cherry picked from commit 0729cdc9a1)
We now do proper handling of existing invalid regions, and
we use ScrollDC which allows us to specify the right clip
region rather than just the bounding rect.
Positioning windows at 0,0 post creation failed, because it
was mapped with CW_USEDFAULT, but private->x/y still said 0,
so moving it to 0,0 did nothing. We now always position the
window at the right place, even when not mapped, but we
create it at CW_USEDEFAULT initially and store that position
before moving it to the right place.
This fixes the window sizing test in testgtk and the inital
position for the gimp toolbar.
The synaptics trackpad driver has some weird behaviour on scroll.
It pops up a window over the mouse pointer (looking like a scrollbar).
This has two problems:
* We get extra enter/leave events for the trackpad window
* We get back the trackpad window when we look for the window
under the mouse to deliver the mousewheel message.
So, we add some trackpad specific hacks to avoid this (sigh) based
on the trackpad window window class.
This fixes bug #542777 and was partially based on a patch there
from Peter Clifton.
gdk_flush() should gdk_display_sync() on all open displays.
Both for display_flush and display_sync it seems useful to call
GdiFlush, but we don't have anything extra to do for display_sync,
as there is no inherent roundtrip on win32.
This should close bug #84314
Windows with transients: center on parent
Splash screens: center on monitor
Also properly ignores initial moves of unmapped
windows that are not override redirect or HINT_POS
Fixes bugs #324254 and #612359
There were still cases where we didn't get a WINDOWPOSCHANGED after
a SetWindowPos() call, like e.g. with a larger minimum size than
the set size (bug #574935)
So, we revert the previous fix and now just always manually emit
a configure notify after the move_resize call. Also, we inhibit
the WINDOWPOSCHANGED configure event during the move_resize operation
to avoid multiple Configures.
There are some cases where we don't get a WINDOWPOSCHANGE such that
we generate a configure event, even if we called gdk_window_move_resize()
or similar. For instance:
* The window is fullscreen
* The window is maximized
* The specified pos/size is the same as the current one
However, as per X11 ConfigureNotify semantics we *always* want one, or
we could run into issue like e.g. bug #537296 where we're waiting for
the CONFIGURE to call gdk_window_thaw_toplevel_updates_libgtk_only().
We always get the WM_DESTROY message anyway, and we remove it there.
Bug #336416 even claims this could be a leak if the WM_DESTROY
message was not seen before the DestroyWindow call returned, as
the WM_DESTROY message could not be handled later without the
window in the handle table. I'm not sure this can happen, but we
might as well remove it.
There is no particular reason to special case this, we want to handle all
sort of normal events. The only special thing we keep is that
as an optimization we pump the message loop extra during a WINPOSCHANGED
in a modal operation as that will cause us to repaint faster.
Also, bump the arbitrary number of mainloop iterations for the timer.
I don't see why we need it at all, but at least doing more than one
iteration if needed should be nice.
When you start a window resize or move via the window menu and
don't actually change anything we're not getting an exitsizemove.
In order to work around this we also look for WM_CAPTURECHANGED.
This moves all the code from WM_SIZE, WM_MOVE, and WM_SHOWWINDOW into
one place, cleans up the code and makes sure we only send a single
configure event even if both size and position changes.
We don't pass in raise anymore, but already_mapped.
Also, already_mapped must be used rather than MAPPED, as we already
synthesize the MAPPED in the generic code (and thus we don't have
to synthesize it again).
Calling PeekMessage can cause reentrant calls into the window procedure
for sent (as opposed to posted) messages, so its not safe to call
when we're not expecting reentrancy. Instead we call GetQueueStatus
when we're just looking for availible messages.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=552041
By reverting gdk_drag_find_window_for_screen logic to what it was
before eb21a7df29.
The old logic knew how to ignore drag_window when searching
for dest_window, but that code was removed (I guess by accident).
https://bugzilla.gnome.org/show_bug.cgi?id=616544
Commit 5ebb32d1ff didn't add the correct
code to find the toplevel window. The WindowFromPoint() function does
not return the toplevel window in the hierarchy, it returns the deepest
non-disabled, non-invisible child. As we don't use invisible or disabled
windows, we don't actually need to use the ChildWindowFromPoint walk for
the non get_toplevel case, so we can remove that code path.
To find a toplevel, we need to start from the desktop and work up, using
ChildWindowFromPointEx (to ignore invisible and disabled windows). If we
don't ignore invisible and disabled windows (as is the case with the
ChildWindowFromPoint call, we are liable to get returns of hidden or
disabled children of the desktop which don't belong to us, but notionally
occupy the same area under the pointer.
An alternative might be to start our walk with one of the children of the
desktop owned by our process and thread - which we can enumerate using,
the EnumThreadWindows call, or (presumably) determine internally. This
would not work when we are inside a GtkSocket though, as the children of
the desktop would belong to the process owning the GtkPlug - we would
have to rely on our own list of windows.
For correctness, this commit adds tests to ensure that we don't try to
return either x or y window coordinates if that corresponding pointer is
NULL.
https://bugzilla.gnome.org/show_bug.cgi?id=658842
The button highlighting in testgtk works again, even with
GDK_NATIVE_WINDOWS. Unfortunately testgtk:menus still does
not work for the forced-native-window-case.
In _gdk_window_move_resize_child it tries to decide whether to pass
SWP_NOSIZE and SWP_NOMOVE based on whether the new size and position
is different from the window's existing position. However it seems
that GDK now ends up updating the window's position before calling
_gdk_window_move_resize_child so this would mean it would think the
window never changes size or position so SWP_NOSIZE|SWP_NOMOVE would
always be set. This causes child windows to never be resized.
This patch changes it so that it never passes either flag to
SetWindowPos. I don't know whether this will cause any side effects
but you'd think it shouldn't do any harm to reassert the current size.
https://bugzilla.gnome.org/show_bug.cgi?id=628049
Signed-off-by: Hans Breuer <hans@breuer.org>
This way, we can include them without accidentally including deprecated
code. Which means we can still use the recently added turning-off tricks
for deprecation warnings.
Allows more modern browsers eg. firefox 5+ to use gtk/broadway
Auto-detects protocol version, and can switch between them at
as you connect a different browser.
This works to some extent, but seems to hang sometimes, for
instance the "button box" test in testgtk never shows up.
Don't try to remember the current keyboard modifier and mouse button
states from the last event, because that isn't always right, and don't
set event.state = 0 for generated events. Instead, add private functions
to get the current states, and implement them with API that retrieves
these states independently from an event.
Move the remaining struct definition into gtkimcontextinfo.h and include
that header in gtk.h. gtkimmodule.h is now an empty header. We should
probably deprecate it somehow.
This is also necessary so headers used in gtk .c files don't include
gtk.h which in turn includes all the deprecated headers which we want to
avoid so we can include them with deprecation warnings turned off.
Make includes not rely on gtkimmodule.h including gtk.h. This will be
important once we remove gtkimmodule from the include files.
That way, we can achieve not pulling deprecated headers automatically.
_NET_WM_STATE_FOCUSED is a new _NET_WM_STATE hint which allows us to
implement a meaningful GDK_WINDOW_STATE_FOCUSED under X11. If the window
manager doesn't support this hint we keep GDK_WINDOW_STATE_FOCUSED set since
that is what gtk+ implicitly assumed historically.
https://bugzilla.gnome.org/show_bug.cgi?id=661428
This state means that the toplevel window is presented as focused to the user,
i.e with active decorations under an X11 window manager.
If the GDK backend doesn't implement this flag, it will just remain set after
mapping the window.
https://bugzilla.gnome.org/show_bug.cgi?id=661428
By default, a background image is stretched. Instead, it is worth to
have a tiled background.
This patch allows background surfaces to be repeated or not, and should
be compatible with future extensions and CSS.
https://bugzilla.gnome.org/show_bug.cgi?id=663522
If a zero-sized NSImage calls lockfocus it throws an exception. Make sure that the image isn't zero-sized before calling lockfocus, and log a warning and return NULL if it is zero-sized.
Have the quartz version of gtk_drag_begin_idle return FALSE if the returned image is NULL.
If _gtk_quartz_create_image_from_pixbuf is given a pixbuf with size 0, 0
or which produces an NSImage with size 0.0, 0.0, it throws an exception
which Gtk doesn't handle.
your changes. Lines starting
which effectively nails down the MOD1 == ALT assumption that is valid
in all other parts of GTK+. After the modifier abstraction fixes for
OSX, the virtual mapping is now (correctly) used in more places, and
caused problems with the common default PC keyboard layout on X11 that
colocates ALT and META on the same key.
Fixes e.g. crashs when dropping from finder.
Turn the "getting_events" boolean into a counter to handle poll_func()
being called recursively, and track the loop depth correctly by
changing its counter before bailing out in run_loop_observer_callback().
This way we reallocate our autorelease pool at the right time, and
don't kill memory that is still in use by outer run loops.
Also drain, not release the pool, just for some defensive forward
compatibility.
(cherry picked from commit ef9a92d225)
Since constructors like gtk_radio_button_new_from_widget take an
instance of their type as the first argument, the gi scanner interprets
them as methods.
https://bugzilla.gnome.org/show_bug.cgi?id=661858
When an NSEvent does not have the window field set, we already assumed
the event was not for us and discarded it. But for NSMouseMoved events
we now make an exception, because such events generated after
using/clicking the main menu bar have the window field set to NULL while
the application window still has focus.
We used to experience a loss of motion events after using the menu bar,
this could be seen in buttons that stopped prelighting and first
clicks often being ignored unless you clicked somewhere else first.
These issues are fixed by this patch.
Submenus are not actually attached to a menubar, but styling them
as if they were imposes some problems (at least with Adwaita).
Just using GDK_WINDOW_TYPE_HINT_POPUP_MENU for submenus instead
fixes the issue.
https://bugzilla.gnome.org/show_bug.cgi?id=662691
Since the wmspec_check_window doesn't have a corresponding GdkWindow we can't
rely on the get_event_window() return value to get the XID from. Just use the
XID from the XEvent directly.
https://bugzilla.gnome.org/show_bug.cgi?id=662953
When multiple pointers are in play, we need to be careful
not to loose track of the device between receiving a button
press and popping up a menu.
https://bugzilla.gnome.org/show_bug.cgi?id=663396
We have an event, so the correct thing to do is to pass
the device into the function that we are calling. GDK
just grew a variant that takes a device, for this purpose.
https://bugzilla.gnome.org/show_bug.cgi?id=663444
Which handle accelerators with keycodes as well as keyvals,
so we can use it in applications that use GtkCellRendererAccel's
"Other" mode of operations (namely gnome-control-center and
gnome-settings-daemon).
https://bugzilla.gnome.org/show_bug.cgi?id=662755
I was trying to avoid including gtk.h here, but this breaks
some of our tests, and there's a good chance that 3rd party
code also relies on this include.
Instead of undefining the DISABLE_DEPRECATED guards,
define the GDK_DISABLE_DEPRECATION_WARNING macro where needed.
Also replace INCLUDES by AM_CPPFLAGS to shut up automake.
The new file defines GDK_DISABLE_DEPRECATION_WARNINGS so it can happily
use deprecated APIs.
This commit moves those functions there that use deprecated functions
and currently cause warnings.
With this commit, GDK compiles without deprecation warnings.
After consulting with Carlos, we agreed that it should be enough to grab
the core pointer instead of doing a full grab. If it turns out that's
wrong, we need to adapt the internal API for resizes to take the device
doing the resize.
As the draw handler expects the items to be laid out already, we cannot
queue a layout here to avoid a race condition with the resize that is
queued immediately after, which in turn would lead to a segfault later
in the paint_item() implementation.
https://bugzilla.gnome.org/show_bug.cgi?id=663138
Everything that deals with a group has to be fixed.
Plus, we have NULL checks without allowing NULL arguments; since NULL
arguments are perfectly fine, we can relax the argument checks.
https://bugzilla.gnome.org/show_bug.cgi?id=662870
This is kind of a hack to get rid of infinite loops that occur when
child accessibles try to set their parent upon creation but the parent
accessible creates its children in the initialize vfunc. Because in that
case, the parent will not have an accessible set when the child tries to
access it, because it is still initializing itself. Which will cause a
new accessible to be created.
https://bugzilla.gnome.org/show_bug.cgi?id=660687
When the active page is switched, it's important
gtk_notebook_pages_allocate() is called after the bulk of
gtk_notebook_real_switch_page() is run, as the former allocates a
different space according to the currently active tab, which is set by
the latter.
gtk_notebook_pages_allocate() already calls gtk_notebook_redraw_tabs()
when the allocation changes, so just move its call down to
gtk_notebook_real_switch_page() to fix the bug.
https://bugzilla.gnome.org/show_bug.cgi?id=662629
If we get a composited-changed signal before the tooltip is ever realized we
have nothing to do.
This actually prevents a segfault in maybe_update_shape() because calling
gtk_widget_get_window() on a non-realized widget returns NULL.
https://bugzilla.gnome.org/show_bug.cgi?id=662467
- add gtkmodulesprivate.h and move stuff there from gtkprivate.h
- add gtkprivate.c and move stuff there from gtkmain.c
- add gtkwin32.c and move stuff there from gtkmain.c
- don't redefine GTK_DATADIR and friends in gtkprivate.h
- have _gtk_get_datadir() and friends on all platforms
- remove the horrid hacks where gtkprivate.h can't be included,
or must be included later due to redefinition of the compile-time
directories
GtkRange needs to check if its allocation intersects with the resize
grip allocation (trimming its own allocation if it does).
In order to do that, it needs to translate its allocation into window
coordinates, and before that, find the window to whose the allocation
is relative; code goes all the way finding the right parent widget, but
then doesn't actually use it when translating the coordinates, leading
to using the wrong rectangles for the intersection check.
https://bugzilla.gnome.org/show_bug.cgi?id=662308
It turns out that simply using XIGrabKeycode instead of XGrabKey
makes the DND keyboard support mostly work (there seem to be some
minor issues with modifiers). This means we no longer grab the
keyboard actively during DND, which in turn makes Alt-Tab and
other window manager shortcuts work again during DND.
At the same time, bring the DND key handling code into the
multi-backend work, by checking for X11 and XI2 at runtime,
in addition to compile time.
Rounded corners now will always work, using XShape in case we're not
running a composite manager.
Also, setting an RGBA visual (if available) on the tooltip toplevel
enables them to be transparent if the theme specifies so.
https://bugzilla.gnome.org/show_bug.cgi?id=599617
kvm(3) is considered a deprecated interface, so make
GMountOperation::show-processes use the recommended sysctl(3) interface
instead. This also removes the need to link with libkvm.
https://bugzilla.gnome.org/show_bug.cgi?id=661835
Bid farewell to G_DISABLE_DEPRECATED and the build errors it causes as
warnings are now used to deter people from using deprecated GLib items
in a more subtle manner
_fstat32 is only introduced with msvcrt80.dll (i.e. Visual C++ 2005), in
which using this function will break compilation with MinGW, which links
against msvcrt.dll. The msvcrt.lib in the Windows DDK which links to
a later incarnation of the msvcrt.dll in later Windows systems may have
this symbol defined, but that needs to be checked upon to be sure.
Thanks to Dieter Verfaillie for pointing out this problem.
Create a synthetic NSMouseLeftDown to store in the GtkQuartzDragSourceInfo
rather than relying on the NSWindow's latest event being the right one (or the
right kind).
Clear the Drag paste board just before the info->context is released.
This way the GtkDragSourceOwner is released just before the drag context
is and thus can pasteboard:provideDataForType: not accidentally access
an already released drag context
When compiled with older SDKs, the original change for this bug caused a
compiler warning about NSWindow not being able to handle a setStyleMask
message. This tricks the compiler into thinking that it can.
and finish the port to using the new modifier abstraction API.
This commit has some evilness, it uses the default display for
the lack of a widget context, and the change to gtkstock.c
is very ugly, but I can't think of anything better given
GtkStockItem needs an accel mask instead of a proper accel
string.
In 2.x, the !HAVE_XCONVERTCASE fallback of keyval_convert_case() was
implicitly used as implementation for all !X11 backends.
In 3.x, when this function was virtualized in GdkDisplayManager,
this fallback was moved to the X11 backend and the other backends
"equipped" with /* FIXME implement */ implementations of
keyval_convert_case() which don't convert anything.
Move the fallback code back to gdk/ as default implementation
of GdkDisplayManager::keyval_convert_case() and remove its
implementations is all backends but X11. Also remove the
implementation in Wayland which was a plain copy of what
is now the default implementation.
(cherry picked from commit f46c1b76d8)
in gtkimcontextsimple and gtktreeview, which are the easy places,
but also true bugfixes, because on OSX we can't use the virtual META
here, we needs MOD2 as delivered in key events.
gobject_introspection's G-ir-scanner doesn't like the -xobjective-c
option needed to compile ObjectiveC features in quartz implementations
of certain gtk functions. This rearranges the compiler flag environment
variables in Makefile so that G-ir-scanner doesn't see them.
This fixes bug 653289, in gtk+-2 gtk_tree_view_stop_editing() was previously
responsible for notifying with gtk_cell_editable_editing_done(), this detail
was missed when porting the cell renderer management to GtkCellArea.
This bug is resolved by fixing two things in
gtk_tree_model_filter_row_deleted():
(1) It is possible for an elt to have elt->visible_siter == NULL, when
it is deleted. Only call g_sequence_remove() if this pointer is
non-NULL.
(2) For the case len (level->seq) > 1, free the elt->children level
if non-NULL. Failing to do this means the level will stick around.
If this child level was not referenced, it will still have a zero
ref count on its parent which cannot be removed!
For both bugs unit tests have been added in the preceding commit.
This tests a case where gtk_tree_model_filter_row_changed() is run
for a node with visible_siter != NULL, however its parent has
visible_iter == NULL. For this case, no valid paths can be generated
as the node cannot be accessed (because the parent is invisible).
Thanks to Kean Johnston for pointing this out.
There are a few places in GTK that use "struct stat",
and then g_stat(), rather than using GStatBuf.This breaks things on
Windows. Since the size of struct stat can vary depending on other
flags specified, this has the potential to cause overwrites and is
trivial to fix.
Based on patch submitted by Kean Johnston
gdk_x11_device_manager_core_list_devices returns a new allocated
list, which has to be freed.
valgrind output:
==18686== 160,176 (80,088 direct, 80,088 indirect) bytes in 3,337 blocks are definitely lost in loss record 25,347 of 25,378
==18686== at 0x4C256DD: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==18686== by 0x6CD7752: g_malloc (in /lib64/libglib-2.0.so.0.3000.0)
==18686== by 0x6CEE2B6: g_slice_alloc (in /lib64/libglib-2.0.so.0.3000.0)
==18686== by 0x6CCB37D: g_list_prepend (in /lib64/libglib-2.0.so.0.3000.0)
==18686== by 0x654CADA: gdk_x11_device_manager_core_list_devices (gdkdevicemanager-core-x11.c:836)
==18686== by 0x6531489: gdk_display_pointer_is_grabbed (gdkdisplay.c:1270)
==18686== by 0x5162E1E: filter_func (ui.c:140)
==18686== by 0x6558B50: gdk_event_apply_filters (gdkeventsource.c:83)
==18686== by 0x6558CB3: _gdk_x11_display_queue_events (gdkeventsource.c:197)
==18686== by 0x6530680: gdk_display_get_event (gdkdisplay.c:311)
==18686== by 0x65589F1: gdk_event_source_dispatch (gdkeventsource.c:356)
==18686== by 0x6CD0A0E: g_main_context_dispatch (in /lib64/libglib-2.0.so.0.3000.0)
https://bugzilla.gnome.org/show_bug.cgi?id=660676
Rather than keeping it and enabling deprecated code in testgtk I decided
to delete the example. The warnings from testgtk about deprecated code
are too useful IMO.
Should we need it again, we can make it into its own test, or add a
testgtkdeprecated.
Shut up valgrind complaining about uninitialized values
==5330== Conditional jump or move depends on uninitialised value(s)
==5330== at 0x624FF4C: gdk_window_move_resize_internal
(gdkwindow.c:6037)
==5330== by 0x62502C3: gdk_window_move_resize (gdkwindow.c:6274)
==5330== by 0x5E4FBB5: _gtk_tree_view_column_allocate
(gtktreeviewcolumn.c:2169)
==5330== by 0x5E3846B: gtk_tree_view_size_allocate_columns
(gtktreeview.c:2631)
==5330== by 0x5E43D76: gtk_tree_view_size_allocate
(gtktreeview.c:2686)
==5330== by 0x7F03959: g_closure_invoke (gclosure.c:774)
==5330== by 0x7F16446: signal_emit_unlocked_R (gsignal.c:3202)
==5330== by 0x7F20256: g_signal_emit_valist (gsignal.c:3003)
==5330== by 0x7F20421: g_signal_emit (gsignal.c:3060)
==5330== by 0x5E6DFBE: gtk_widget_size_allocate (gtkwidget.c:4886)
==5330== by 0x5D94ADD: gtk_scrolled_window_allocate_child
(gtkscrolledwindow.c:1506)
==5330== by 0x5D95299: gtk_scrolled_window_size_allocate
(gtkscrolledwindow.c:1704)
==5330== by 0x7F03959: g_closure_invoke (gclosure.c:774)
==5330== by 0x7F16446: signal_emit_unlocked_R (gsignal.c:3202)
==5330== by 0x7F20256: g_signal_emit_valist (gsignal.c:3003)
==5330== by 0x7F20421: g_signal_emit (gsignal.c:3060)
==5330== by 0x5E6DFBE: gtk_widget_size_allocate (gtkwidget.c:4886)
==5330== by 0x5C62708: gtk_box_size_allocate (gtkbox.c:658)
==5330== by 0x7F03959: g_closure_invoke (gclosure.c:774)
==5330== by 0x7F16446: signal_emit_unlocked_R (gsignal.c:3202)
==5330== Uninitialised value was created by a stack allocation
==5330== at 0x5E4FA8D: _gtk_tree_view_column_allocate
(gtktreeviewcolumn.c:2139)
Those if() blocks don't have any reason being there, as x and y are not
pointers. If the window is destroyed, just set the out values to zero
and return.
As seen in valgrind:
==3306== Conditional jump or move depends on uninitialised value(s)
==3306== at 0x624C74F: gdk_window_get_root_coords (gdkwindow.c:6933)
==3306== by 0x5E193C3: gtk_tooltip_show_tooltip (gtktooltip.c:1160)
==3306== by 0x5E19C05: tooltip_popup_timeout (gtktooltip.c:1282)
==3306== by 0x623B102: gdk_threads_dispatch (gdk.c:754)
==3306== by 0x8592F3A: g_timeout_dispatch (gmain.c:3907)
==3306== by 0x859174C: g_main_context_dispatch (gmain.c:2441)
==3306== by 0x8591F47: g_main_context_iterate (gmain.c:3089)
==3306== by 0x8592494: g_main_loop_run (gmain.c:3297)
==3306== by 0x5D2E501: gtk_main (gtkmain.c:1362)
==3306== by 0x5C5652F: gtk_application_run_mainloop
(gtkapplication.c:115)
==3306== by 0x7C47C9D: g_application_run (gapplication.c:1323)
==3306== by 0x447B5F: main (nautilus-main.c:102)
==3306== Uninitialised value was created by a stack allocation
==3306== at 0x624D48A: gdk_window_get_device_position
(gdkwindow.c:4952)
gdk_unicode_to_keyval(uc) returning (uc | 0x01000000) is not an
error return value but simply the way to encode 24-bit unicode
characters directly as keyvals.
When we use the style context to get information for the tab region,
also add a style class to indicate its position, so that the relevant
information is pulled off from the theme.
https://bugzilla.gnome.org/show_bug.cgi?id=659777
This is useful to e.g. theme notebook tabs differently according to
their position directly from the CSS sheet.
GtkNotebook support in a separate commit.
https://bugzilla.gnome.org/show_bug.cgi?id=659777
The setter for this was deprecated in 2.x and removed in 3.0.
I don't see any reason why we should hardcode 2px for this value;
instead, deprecated the getter and make it always return zero, and stop
using the variables internally.
https://bugzilla.gnome.org/show_bug.cgi?id=659777
The code before was basically adding and removing the same padding value
in two different places during the allocation cycle.
Instead, what we want to do is to offset the inactive tab allocation by
the difference with the active tab padding, to ensure the tab content is
always drawn centered and in the right position.
https://bugzilla.gnome.org/show_bug.cgi?id=659777
We want to enable the use of different padding values between active and
inactive tabs, so that the two are completely separated (but limited by
the active tab size).
This way themes can decide how bigger the active tab is drawn compared
to the normal one just specifying a different padding value from the
CSS, like this:
.notebook tab {
padding: 2;
}
.notebook tab:active {
padding: 4;
}
As a first step, fetch the padding values with the right state flags
from GtkStyleContext.
https://bugzilla.gnome.org/show_bug.cgi?id=659777
Add enum GdkModifierIntent which identifies use cases for modifier masks
and GdkKeyMap::get_modifier_mask(). Add a default implementation which returns
what is currently hardcoded all over GTK+, and an implementation in the
quartz backend. Also add gtk_widget_get_modifier_mask() which simplifies
things by doing widget->display->keymap->get_modifier_mask().
Add _gtk_button_event_triggers_context_menu() and use it instead
of checking for event->button == 3, so context menus are invoked
correctly on the Mac.
which are SHIFT and MOD2 on the Mac, and SHIFT and CONTROL otherwise.
Use the new define all over the place and rename variables and
members to not say "shift" or "control".
Introduce <Primary> in accelerator strings, which resolves to
GDK_CONTROL_MASK on X11/Win23, and to GDK_META_MASK on quartz.
Also serialize CONTROL/META as <Primary> depending on the platform.
Use foreach() instead of forall() to find the child's siblings
because internal children of subclasses have no sibling relation
to the box' children. Also deal with the subclass failing to
implement get_path_for_child(). This caused an infinite widget
path invalidation loop of sorts with GimpMessageBox, which is a
vertical box with a decorative icon to the left.
Fixes#640741 - gtk_tree_view_column_get_cell_position() seems to be broken
It is supposed to return the offset within the column, but returned
the offset within the tree, changed by allocation.x of the column's
button (which I don't really unserstand and was clearly not working).
The handling of the module lists is a bit confusing here. There is
a global list that is populated with initial modules from GTK_MODULES,
and then there is a second list that is suppposed to hold the modules
loaded from the gtk-modules xsetting, which we attach to the GtkSettings
object. The blacklisting caused us to put the wrong modules into the
second list, and mess up the module refcounts, eventually double-freeing
memory when the setting changes repeatedly.
https://bugzilla.gnome.org/show_bug.cgi?id=659966
The following assertion was triggered in model_row_inserted() because
iterate_thru_children() modifies the parameter tree_model before passing it to
traverse_cells().
Gtk-CRITICAL **: gtk_tree_path_compare: assertion `b->depth > 0' failed
The stack with the bug was:
#0 gtk_tree_path_compare at gtktreemodel.c
#1 traverse_cells at gtktreeviewaccessible.c
#2 model_row_inserted at gtktreeviewaccessible.c
This patch calls iterate_thru_children() with a copy of the path so that the
original is not modified.
This way, all rows have the same height. It has 3 advantages:
1) No actual computation is necessary to compute the size of the cell.
This speeds up the list tremendously (filling out the list goes from
25s to 3s).
2) Buggy fonts don't mess up the list anymore with their weird sizes.
Instead, they are clipped / empty space is added.
3) Buggy fonts look more buggy. So their use is hopefully discouraged.
With absolute sizes, Pango is way better at getting the actual sizes of
the fonts to match up. It's a bit harder to compute a proper value for
this, whcih makes the code ugly, but as long as it works better...
This way, we can find fonts way quicker as we only need to create font
descriptions for fonts with matching families. Most importantly, we're
rather quick in the "the font doesn't exist" case.
Name it gtk_font_chooser_widget_load_fonts(). Also, don't take any
arguments, they were the same everywhere and they're member variables of
the font chooser anyway.
Instead of reloading the font list, we now just queue a redraw. This
works, because the preview text is added using a cell data func instead
of a custom column.
... which looks up the font in the list of fonts. This then can be used
to select an actual font upon changes.
Also fixes cases where the get_family() and get_face() functions would
return outdated data when set_font() had been called.
They now go through gtk_font_chooser_widget_take_font_desc(). The end
goal is to make all changes go through this function, so that all
updates that are happening are easy to track.
Another change is that the code now merges the font description instead
of just using the new one. This avoids weird situations when people set
the font "Bold" for example, which has neither a size nor a font family.
... instead of rereading all the fonts every time.
With this change, the liststore now contains every font face known to
GTK, so we can actually walk it for matching fonts.
- Don't do unnecessary casts
- Fix weird variable declaration indenting
- Fix loop indentation
- Use a for loop for iterating over a list, instead of a while loop
- Casefold font name only once, instead of every iteration
- Remove needless true_var = true_var && TRUE assignment
Render GtkIconView cell items with the prelight state flag when they're
being mouse hovered.
This works basically in the same way it's done for GtkTreeView cells,
and e.g. GtkCellRendererPixbuf will need to have its follow-state
property to opt in to prelight rendering.
https://bugzilla.gnome.org/show_bug.cgi?id=615501
The setter for the "font" property returned a boolean that indicated
whether the given font has been found. Instead, fall back to the default
font name when the given font doesn't exist.
GtkFontButton already has a property named "font-name" which may
conflict / be an ABI break when moving GtkFontButton to implement
GtkFontChooser. Also, this is more in line with how other parts in
gtk (e.g. GtkCellRendererText) call a font string property.
Make the GtkFontChooser API similar to the Gtk{File,Recent,App}Chooser
APIs by introducing GtkFontChooser as an interface, that has a default
implementation in GtkFontChooserWidget.
https://bugzilla.gnome.org/show_bug.cgi?id=657627
We simply use the Tracker DBus api here, caching and direct
access that come with libtracker-sparql are probably not needed
here. Based on a patch by Martyn Russell.
https://bugzilla.gnome.org/show_bug.cgi?id=658272
gtk_tree_model_filter_free_level() should always release ref on parent
unless the parent has been removed in the child model. The unit tests
added in the previous commit test this.
If the level to be reordered did not contain an elt with offset == 0
(for example if that node was hidden), the first reference was not
correctly transferred.
A unit test for this has been added in a previous commit.
Handle dead keys in special_ucs_table and have them converted by
UCKeyTranslate(), so all dead key combinations can be entered.
Later, this should be handled in the input method, just as it's
done for X11/Win32.
This is due to the special handling for filter functions that depend
on child level state. If the parent level of a level is the root level,
the level cannot be removed from the cache. Any change in this level
can cause its parent in the root level to become visible again.
The most important change is that the function now properly releases
any external ref count and propagates these changes to the child model.
If a node is removed due to a filtering action, we now properly release
all external reference counts for its hierarchy of child nodes.
Apart from that, the function has been restructured to remove code
duplication.
Finally, there is still some doubt whether there are more calls to
gtk_tree_model_filter_free_level() which need unref_external set to TRUE.
Should be helpful when debugging this thing. Perhaps this needs to
be properly integrated in G_DEBUG/GTK_DEBUG, but currently I do not see
the need to.
There was still a discrepancy between the comments and the code about when
exactly a level was to be removed. The rule is now set on that for a level
to be removed its zero_ref_count as well as the zero_ref_count of its
parent has to be zero.
The code sets old_adj_ptr to the location containing the old weak ref,
but then wants to remove a weak ref from &view->old_hadj, causing warnings
when disposing the widget.
Fixes bug #658379 - Disabled devices still added to list on startup,
spotted by Bastien Nocera. Do not create GdkDevices for disabled
devices on device manager construction, leading to a confusing initial
state.
This will take you to the file's folder and select the file in question.
The menu item is only available in Recently-used and Search modes, so that
you can go from files in them to the 'normal' browsing mode.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
We get rendering artifacts that make tests fail. Not good.
The code is only commented out so far, so it should be trivial to
reenable if someone wants to. (Or this commit could just be reverted.)
row-spacing behaved like column-spacing should have and vice versa.
Also update tests that erroneously checked this behavior.
Thanks to Joanmarie Diggs for finding this.
For god-knows-what reason, at-spi is trying various formats
of strings when registering listeners, triggering g_warnings()
from gailutil code. Stop doing that.
Also, don't leak temporary string arrays that are a side-effect
of passing parameters around as formatted strings.
https://bugzilla.gnome.org/show_bug.cgi?id=658168
Make page titles left-aligned again. Changing from misc alignment
to GtkWidget::halign here was a mistake. We can't do that for labels
in a size group, as long as labels still pay attention to misc
alignment.
https://bugzilla.gnome.org/show_bug.cgi?id=658008
The gtk_render_line() code has a FIXME (coming from the 2.x migration I
suppose) about using xthickness, and it actually hardcodes two lines of
different color instead of drawing a single one.
Since we don't support xthickness anymore, make gtk_render_line() do
what it says, i.e. just draw an 1px line colored with the current
color.
https://bugzilla.gnome.org/show_bug.cgi?id=657963
-Tell people about the GNOME Live! page which gives a more detailed
outline on building the GTK+ stack with Visual C++
-Update README.win32 as GAIL is now a standard part built into GTK+
for its a11y functions, and GAIL-Util is now built with the project
files too.
-Tell people about the VS2010 support that has been available for a
while
Otherwise the default values are only used if no vfunc is set,
making that default unavailable to vfunc implementations as a fallback,
and unavailable to language bindings (such as gtkmm) that always
have a vfunc implementation, even if that often only tries to call
the base vfunc implemntation.
This was making gtkmm's ToolPalette's (and probably ToolBar's)
items look like buttons instead of being flat.
https://bugzilla.gnome.org/show_bug.cgi?id=656223
We used to get an implicit -lm from gdk-pixbuf, but that has
recently been changed to not inject extraneous libraries into
link lines, so we have to do it ourselves now. This is more
correct, anyway.
This fixes a regression introduced in commit
b1a1685a33eca16979e63d79915395af0b36. gdk_visual_get_*_pixel_details()
should not be called with a NULL visual.
https://bugzilla.gnome.org/show_bug.cgi?id=649588
When using an editable toolbar in evince, we can end up
in gtk_toolbar_get_visible() position with item being the
highlight_tool_item, but not one of the regular children.
So, handle that.
There's no real asynchronicity going on, anyway, so let's do both within
a single iteration of the idle handler.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
The mtime from GtkRecentManager may not the same as the file's actual
mtime, so the final result could appear unsorted to the user. Instead,
we will let the view do the sorting.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
It's possible the GtkLabel receives drag'n'drop related signals
if it was setup to receive them by a container or other external
code, just avoid dereferencing priv->select_info in this case.
Also create a separate section for values to associate the values with.
Finally, use this section information when delayed-parsing a widget
style property.
This struct keeps track of an area of text in a CSS file and uses it
when specifying information. Also, the cssprovider keeps track of
sections when parsing a file.
gtk_menu_shell_insert() is a virtual function that was being directly
invoked from the class vtable.
Turn it into a proper signal and emit it in the usual way.
https://bugzilla.gnome.org/show_bug.cgi?id=656565
This makes the behavior identical to other widgets which
handle button presses, and it avoids problems when placing
switches into a windows main toolbar.
https://bugzilla.gnome.org/show_bug.cgi?id=656986
We have meaningful labels for the Ok buttons in all other choosers,
so just use the same string we use in the app chooser and the
font chooser.
Also move the mnemonic from s to a in 'Saturation', to avoid a
mnemonic conflict. String change !
https://bugzilla.gnome.org/show_bug.cgi?id=657006
This is done in the same way as GArray was replaced with GSequence in
GtkTreeModelFilter, in a patch done by Xavier Claessens.
All sorting code has been reworked to use the GSequence sorting
and insert sort functions, instead of doing a lot on our own.
Significantly improves performance when e.g. removing (filtering) a lot
of rows from the filter model. Fixes bug 616871.
This commit includes changes by Kristian Rietveld to make the patch apply
on top of the treemodel-fix branch and pass all newly written unit tests.
This allows for more thorough testing of "has child" filter functions.
We also test a has child filter function with a sort model as
child model, to verify that we receive enough signals to function
properly.
The filter model is now more strict about the signals which are emitted,
so the base tests have been expanded to test with both the root level
nodes collapsed and expanded.
- Before we kept a reference on all nodes in non-root levels. This has
been changed, now we keep a reference on the first node of each level.
If, due to changes in the model, another node becomes the first node in
the level, the reference is transferred to this new first node.
- All non-root levels keep a reference on their parent.
- By making use of the external ref count, the filter model now emits less
unnecessary signals.
- GtkTreeModelFilter does support filter functions which decide visibility
of a given node based on the number of or visibility of children.
To accomplish this, a child level of a node is cached when its
parent has an external ref count > 0, because changes to the node might
affect this parent.
- An optimization for not building the root level in case the inserted
node is not visible in gtk_tree_model_filter_row_inserted() has been
removed. In this case, we still need to build the root level and
possibly a child level to monitor for signals which might make
this row visible.
In gtk_tree_model_filter_check_ancestors(), also handle the case when
a node is already in the cache, but invisible, in the root level.
With the upcoming changes to GtkTreeModelFilter's ref counting this
case can occur.
We need to distinguish between the ref count objects have on us versus
the ref count we have on our child model. To keep track of the former,
we introduce the "external ref count" in this commit. The zero_ref_count
needs to be determined from the external ref count, because objects that
have a ref count on us have say in which levels must be cached and which
can be released.
Before the caching in GtkTreeModelFilter was essentially broken and
levels were never released. This was caused because the zero_ref_count
was connected to the ref count the filter model had on its child model.
Now that this depends on the external ref count, this is working fine and
as to be expected.
The row-deleted signal should be emitted after the internal data
structures have been updated. In gtk_tree_model_filter_remove_elt_from_level
and gtk_tree_model_filter_virtual_root_deleted the signal was still being
emitted before the updates were carried out.
Now that we call unref_node in free_level, we have to take care that
free_level may only unref (parent) nodes when these still exist in the
child model. After row-deleted has been received for a node, its
children may no longer unref this node.
Referencing a parent node for each referenced node is overdone. Instead,
we now reference the parent from build_level and unreference in free_level.
Each level keeps a single reference on its immediate parent. This both
alleviates the performence problems and should perfectly serve the purpose.
The bulk of the fix is to walk the chain of ancestors, starting at the
root level, and check if the visibility of any of the ancestors has
changed. If yes, the necessary signals are emitted so that this change
is propagated properly. This walk is done after a node has been
inserted, changed or deleted, see function
gtk_tree_model_filter_check_ancestors().
Bug reported, and initial debugging and analysis, by Xavier Claessens.
Suggested by Xavier Claessens / bug 621076.
(Additional obseration: this should speed up the filter model's
handling of row-inserted as a binary search is now used instead
of a linear scan).
-Only define HAVE_STRINGS_H when MSVC is not used
-Define HVE_STDINT_H for Visual C++ 2010 and later, as Visual C++ 2010
does ship with a "proper" stdint.h by default
-Add check macros HAVE_RINT and HAVE_ROUND--they are only defined when
MSVC is not used
-Remove obsolete check macros
gtk/gtkborderimage.c, gtk/gtklabel.c and gtk/gtkstyleproperty.c call
round() and/or rint(), which was only available in C99 compilers.
This adds the inclusion of the fallback implementation (gtk/fallback-c89.c)
to define these functions if they are not initially made available by the
compiler.
Also remove the rint() implementation in gtk/gtklabel.c as it is now in
the fallback implmentation.
Add an implementation for rint() and round() for compilers that do not
define one or both of them. Note that this file should be included
after config.h was included.
Copy the Win32-specific GDK backend headers during the "install" stage
as well, as they were missed in the install, causing trouble when
building other projects like WebKitGTK+.
Thanks to greg.hellings for pointing this out in Bug 653964.
This commit introduces a new setting, gtk-visible-focus, backed
by the Gtk/VisibleFocus X setting. Its three values control how
focus rectangles are displayed.
'always' is equivalent to the traditional GTK+ behaviour of always
rendering focus rectangles.
'never' does what it says, and is intended for keyboardless
situations, e.g. tablets.
'automatic' hides focus rectangles initially, until the user
interacts with the keyboard, at which point focus rectangles
become visible.
https://bugzilla.gnome.org/show_bug.cgi?id=649567
This way themes can use GtkComboBox.combobox-entry to match that
specific case, and GtkComboBox.combobox-entry .button to match the
button/arrow side of the widget.
Apparently schemas are part of the ABI contract; running apps crash if a
schema disapppears and they happen to access it.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
Functions dealing with native Xlib types were (skip)ed because
gobject-introspection did not have correct Xlib types declarations.
They are corrected now, so these GdkX11 functions can be enabled back
again.
https://bugzilla.gnome.org/show_bug.cgi?id=655495
I tried to suppress compiler warnings on pre-10.6 machines this way,
but it defeats its purpose when you compile for pre-10.6 machines on
a 10.6 machine. For now, we have to live with the warnings when
compiling on/for pre-10.6 machines, there does not seem an easy and proper
way to suppress the warnings.
Years ago, when the location entry only appeared in a popup window, it used to appear
populated with the current folder's path. We had some logic to do this, but since now
the location entry appears in the main dialog, we can do away with that old code.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This way we can re-create the contents of that box easily without disrupting
the rest of the file chooser.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This will be the central function to switch operation modes: stop searches or the recent-files
process; switch widgets, etc.
We factor out the common code from recent_switch_to_browse_mode() and search_switch_to_browse_mode(),
and remove those functions. All the code that switched modes by hand now calls
the central operation_mode_set().
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
That code indeed lives in operation_mode_stop(), so use it instead of
having inlined duplicates.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
We'll now prefix functions to change the impl->operation_mode with 'operation_mode',
for clarity.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
Various paths in SAVE or CREATE_FOLDER return from ::should_respond() but
leave an asynchronous process running. This process checks some things
in the user's selection, for example, 'does the file exist, and if so
do we need to bring up an overwrite-confirmation dialog?'. When these
async processes complete *and* it is indeed time for the dialog to
be terminated (via the response-requested signal), we also need to
save the selection to the recently-used list - as ::should_respond() does
by itself in the cases when it can request a response immediately.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
To make life easier for users, when apps don't properly update the recently-used list
after choosing a file, we now do that directly from the file chooser.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
In GtkClipboardOwner pasteboard:provideDataForType do not call
_gtk_quartz_set_selection_data_for_pasteboard() is selection_data.length
is smaller than 0. The function relies on having a positive length,
since it stores the length in a uint ...
After some searching, it turns out that gtk-doc now interprets
markdown, and can no longer accept '-' in the docs without making
a mess. I hope that this does not affect more places in more
subtle ways.
Note that clamping in rounded_box_grow() is not necessary as that
function cannot lead to overlap unless the rounded box was overlapping
previously.
https://bugzilla.gnome.org/show_bug.cgi?id=655009
-Change ATK dependency back to atk-1.0.lib, and the corresponding include
folder back to atk-1.0 for all projects as ATK-2.x will still retain the
1.0 suffixes. (ATK Commit 01cec72)
-Update corresponding description in the VS README.txt files
-Also fix up the VS2010 README.txt file a bit.
We want the role to be kept in the accessible object. Using
gtk_widget_class_set_accessible_role() is only meant as a quick
workaround to requiring subclassing of the accessibles in the quite
common case where a subclass does not change the accessible
implementation at all and only has a different role.
This is an attempt to explore differences in GTK+'s and gails
notions of which widget currently has the focus. Since gail
does all its focus updates in an idle, we cannot do the
comparisons synchronously, so just check ever so often in
a timeout.
GParamSpec now uses an intern'ed string for 'name', so the code
in gtk_theming_engine_register_property() which prefixes the name
with a namespace has to be adjusted.
https://bugzilla.gnome.org/show_bug.cgi?id=654695
https://bugzilla.gnome.org/show_bug.cgi?id=653947
It could happen that a cookie event has been already allocated/freed
in an event filter, as it can't be allocated a second time, all GDK
can do is skipping the event. Spotted by Guillaume Desmottes.
This function can be used to find the GdkDevice wrapping
an XInput2 device ID. For core devices, the Virtual Core
Pointer/Keyboard IDs (2/3) may be used.
This function can be used to find out the XInput2 device ID
behind a GdkDevice, mostly useful when you need to interact
with say Clutter, or raw libXi calls.
-Added projects to compile the a11y portion of GTK+. This is now necessary
as a11y/GAIL is now integrated into the main GTK+ library, and it must
be built before compiling/linking GTK+. This project is done like the
GDK/GTK+ projects, where the source file listings for the VS2008/2010
projects are fed into templates (.vcprojin, .vcxprojin and
.vcxproj.filtersin) during 'make dist'
-Added projects to compile the libgail-util DLL (no templates for this
as this does not have source files added/removed often)
-Added the new projects into distribution, and headers, DLLs and .LIB files
into the "install" stage
Various paths in SAVE or CREATE_FOLDER return from ::should_respond() but
leave an asynchronous process running. This process checks some things
in the user's selection, for example, 'does the file exist, and if so
do we need to bring up an overwrite-confirmation dialog?'. When these
async processes complete *and* it is indeed time for the dialog to
be terminated (via the response-requested signal), we also need to
save the selection to the recently-used list - as ::should_respond() does
by itself in the cases when it can request a response immediately.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
To make life easier for users, when apps don't properly update the recently-used list
after choosing a file, we now do that directly from the file chooser.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
We use submenuitems for all menuitems now. So they shouldn't be all
advertised as menus. For the ones that do have submenus, we still
advertise ROLE_MENU per the last patch.
The remove handler would not check that removed widgets do indeed have
an accessible associated with them and would happily run all the removal
code with a NULL object. Not good.
Due to tragic misunderstandings at the birth of ATK, AtkUtil
can't actually be properly derived. Instead, each implementation
has to poke its vfuncs directly into the AtkUtilClass struct.
So painful to have shipped this stuff for 10 years...
And we don't support not loading the module anymore, either.
So take out the code that looks at environment variables for
that purpose, and don't print a misleading message about
module loading.
As we're going to merge the two accessibles for submenus, we need to
make sure the code that insisted on submenus can handle the case where
no submenu exists.
1) Allow inserting text to move the cursor to after the inserted text
2) Implement selecting text for GtkTextView. Also assert if it's an
unsupported widget type.
3) Select an inside part of the widget, not the end.
Explicitly set the notebook page object as the accessible parent
of the page, otherwise, AtkObject follows the widget parents chain
and makes the notebook accessible the parent.
This drops the AtkText implementation, and also strips handling
of children out. Instead of listening for enter/leave/press/released,
just listen for state changes on the widget.
The way submenus are handled here isn't really right, since
they can come and go at runtime, but I've left it like this
for now. A side-effect is that we can't use
gtk_widget_class_set_accessible_type here, but have to
override get_accessible.
At the same time, drop the AtkText implementation.
Also, reintroduce a GtkBoxAccessible, since the previous
way of setting the role in get_accessible was affecting
GtkBox subclasses negatively.
Adjust expected test output to match.
We are listening to page-added and emit child_changed::added
in response to that, so don't let the add_gtk handler add
extraneous extra signal emissions. This fixes the child-handling
test for GtkNotebook.
Selection count is tested via assertions and can be inferred from
looking at the number of selected children.
And I'd like to reduce the output of the tests. We're outputting way too
much anyway.
Check that number of selected children == number of children that
actually are selected. If that's not the case, assert, to avoid people
being careless about adding tests.
We used to keep the GtkNotebook, but that causes issues at creation time
of the GailNotebook, when the GtkNotebook does not yet have an
accessible set and so it can't be queried.
Previously, the code tried to track the indexes of the pages and keep
them up to date in a list and tracking the index in the
GtkNotebookPage. Now, we store the widget we are tracking in the
GailNotebookPage and keep a hash table of widget=>GailNotebookPage in the
GailNotebook. This frees us from the burden of tracking page changes.
It is now done by the GailNotebook. Previously it tried to do it itself.
Also, we now use GtkNotebook::page-removed to track the removal, as
opposed to child::parent-set.
Remove idle handler and omit checking if the name really was changed.
ATs will either have to live with the name not really changoing or we
should check in GtkWindow.
Ensure that text and selection boundaries are in a
consistent state when we emit notifications, that ::insert-text
and ::delete-text are emitted before any changes, and that
::changed is emitted after property notifications when inserting
or deleting text.
These tests check that text and selection boundaries are in a
consistent state when we emit notifications, that ::insert-text
and ::delete-text are emitted before any changes, and that
::changed is emitted after property notifications when inserting
or deleting text.
These functions are implementations of the AtkText api on top
of a PangoLayout, and are intended to replace GailTextUtil.
Since gtkpango.h is a private header, also remove the individual
inclusion prohibition.
The new test is much smaller, and the results have been carefully
looked at; GtkLabel gets it right, the others don't.
Also, temporarily include some code that dumps out useful information.
This removes a horrible workaround for bug 126797. To prevent
picking up accidental markup in label texts, the label accessible
is listening for window creation and mapping and defers initializing
its text until then.
This tests a problem that was first noticed in 2003, in
bug 126797: The text of a label accessible is not properly
updated when the label switches from !use-markup to use-markup.
A first experimental conversion from the gail namespace to gtkaccessible.
At the same time, use gtk_widget_class_set_accessible_type() to register
the accessible type for GtkLabel.
Avoid many unnecessary list iterations by using a hash table
to store cell infos, and caching row and column counts. Based
on patches by William Jon McCann, bug 554171.
tree-performance results:
before: (MINPERF:large tree test with a11y: 9.18531sec)
after: (MINPERF:large tree test with a11y: 0.923463sec)
for comparison, without accessibility:
(MINPERF:large tree test: 0.016179sec)
File choosers depend on the files in the current directory and the
settings the user has set for the file chooser, so the output cannot be
deterministic.
This reverts commit f05c9e02c49f857c33e02c3d89483fcccb5df254.
The font chooser depends on the fonts installed on the system, so the
output is nondeterministic.
This reverts commit 08fdc399762c2af07d94e42a1801e691e5d4d6ab.
As set_description is never called and unsupported by the at-spi, we can
omit implementing it.
This means we can also omit get_description calls in various places, as
they'd just return the default value: NULL.
This code was supposed to work around a bad interaction between GOK and
Nautilus from 7 years ago.
If it still exists, the GOK developers may complain to the Nautilus
developers.
https://bugzilla.gnome.org/show_bug.cgi?id=137401
The doc build for that is currently broken, and libgail-util is
undergoing reconstructive surgery anyway, currently.
Or maybe it'll turn out to be an amputation...
It turns out that ATK_DEFINE_TYPE_WITH_CODE() is broken; it
tells GType that the class and instance size for the accessible
type are the same as for its parent type. Which is not true
if your instance struct has members such as 'description' here.
This was causing hard-to-track-down memory corruption, since
description and the GtkAccessible private pointer were sharing
the same memory location.
Note that the spatial aspects of AtkComponent are all dependent
on fonts and other environmental factors, thus we only dump the
little bits outside of that.
Not really the best fix; for now just ignore accessible parents
that are not GtkAccessibles - it seems that something causes
GailToplevels to show up as parents of GailWindows, randomly.
For initial creation of the results, allow to run
accessibility-dump --generate file.ui,
which will dump just the expected output to stdout,
without any extra decorations.
The tool works like this:
./accessibility-dump [FILE ...]
If no files are given, all files with the extension ".ui" in the current
directory are taken. For every file "test.ui", the following steps are
performed:
1) test.ui is loaded using GtkBuilder
2) The accessible for the window is loaded
3) The information of accessible is converted into a string using a
syntax defined in this test file
4) The generated string is diffed with the file "test.txt"
5) If the diff is empty, the test is a success, if not, the test fails.
6) The diff is output when the test runner is run with --verbose
So to add a test named "test", create a file called "test.ui", put it
into this directory. Then create the expected output file "test.txt",
put it into this directory too. You can create the initial version of
this file by invoking "./accessibility-dump --verbose test.ui". The
output will contain the expected text and can be copy/pasted into the
text file.
The function is supposed to bypass the ATK registry. For 2 reasons:
1) We get rid of a lot of boilerplate madness.
2) The registry allows creating multiple accessibles per widget and we
don't.
The old code for registries is still there.
It is now no longer possible to disable it.
This doesn't matter though because GTK will not instantiate a11y
objects until you actually use it. So nothing changes in practice.
Instead of including <gail/gailfoo.h>, include "gailfoo.h"
Private headers should be included like this (to make it esily visible
that the headers are private) and we want to move this directory to a
different name, so hardcoding the directory name is counterproductive to
that.
When tests are larger than the screen size and no compositing is
enabled, the window will be clipped to the screen size and all areas
outside of the screen have undefined contents.
To avoid this, we can use an offscreen window.
Basically, don't ever set the current folder, and only use
gtk_file_chooser_set_filename() for 'File/Save As'. This is so
that the file chooser will be able to present its recently-used
lists as appropriate, giving the user good suggestions by default.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
None of the cases where _gtk_file_chooser_entry_set_base_folder() appear to require
the entry highlighting the file's basename. Doing the highlighting actually makes
things look weird in Save/Recent mode if you
1. type a filename
2. click on a recent-folder,
as right after (2) your filename would get its basename highlighted for
no apparent reason.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
The create-folder machinery doesn't handle that case yet; we may enable it later
once we figure out the implications for the GUI.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
It used to be that every part of the file chooser's code would show/hide the widgets
near the pathbar as needed. Now we have two central functions:
path_bar_update()
path_bar_set_mode()
These take care of all the widget shuffling; setting the visibility of the
pathbar, info bar, and Create Folder button as appropriate; setting the contents
of the info bar, etc. - based on the current operation_mode and action.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
We will centralize the place where all the pathbar-related widgets are created:
the location button, the pathbar itself, the Create Folder button, and in
subsequent commits, the info bar as well. We will deal with the pathbar/infobar
as a unit, instead of swapping them in and out in an ad-hoc fashion.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
Since the GtkFileChooserEntry already gets the recent-folder set upon it when a recent-folder
is selected, it already can give us the correct fully-formed path.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This lets us do proper completion in GtkFileChooserEntry even when no base folder
has been set. Completion for relative paths won't work, as usual, as expected.
They weren't being selected in the shortcuts bar when those modes were
activated programmatically, instead of through the user selecting
them from the user interface.
In RELOAD_EMPTY mode, when no folder has been selected by the calling app, we now
start showing the recently-used list. The rationale is as follows:
- In Open mode, the user is likely to pick a file he has used recently.
- In Save mode, the user is likely to want a destination folder which
he has used recently.
For the Save case, where we want to present the user with recent folders instead
of recent files, we will make the recent-list do so in subsequent commits.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
Now we reparent the browse_path_bar_hbox to that spot in Save mode,
or to be above the file lists in Open mode. The pathbar makes for a very
clear indication of the location to save in.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
And with this we get rid of the craziness of having a separate filter model
for the combobox's model.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This effectively makes the file chooser always be in 'expanded' mode.
Later, we'll move the pathbar to the 'Save in folder:' line.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This extracts the parent folders from the items in the recently-used
list. We'll use it in the file chooser to present a list of
recently-used folders.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
The table was incomplete and out of date. Instead, just
put a list of links in that place, and move all the extra
documentation to the macros. Bug 653785
When we build the sibling path for the order, we do not skip hidden
children (since, quoting the comment, "we cannot reliably detect changes
in widget visibility"). So we need to invalidate the order when hidden
children are reordered and removed.
https://bugzilla.gnome.org/show_bug.cgi?id=652769
This time I realized that I needed to set autocrlf=false on my Windows side
... ugh...
This is one of those files that must have CRLF line endings to work
correctly :|
-Reinstate build/win32/vs10/gtk+.sln with the correct EOL (DOS/Windows), so
that it will be correctly recognized by Windows instead of having the
annoying "Unrecognized Visual Studio Version".
-Update property sheets to reflect on new headers added
-Change the demo program to be gtk3-demo.exe, to be consistent with the
names on other platforms, and updated/renamed related project/solution
files, and added overlay.c to the list of demo sources.
Removing the window from the window list before setting the
application to %NULL avoids gtk_application_remove_window() triggering
another call to gtk_application_window_removed(), which would release
the application a second time.
https://bugzilla.gnome.org/show_bug.cgi?id=653053
As Cairo and Cairo-GObject are often built as two seperate DLLs/modules,
set the property sheets to link to both libraries, instead of using the
previous approach where a monolithic Cairo DLL which contains GObject
support is used.
-Remove unneeded tags from projects
-Seperate intermediate directories for projects to avoid rebuilding/
linking on every rebuild and MSBuild errors (et al.) for not being
able to write into build log files as they are in use.
It turns out there's more places where the toolbar item size is used as
the margin box instead of the content box. Because of that, store the
margin box when allocating and use it whenever calls
toolbar_content_get_allocation() instead of calling
gtk_widget_get_allocation().
size_allocate() allocates the available space for the margin box,
get_allocation() returns the actual space of the content box and those
can be different. And then animations never stop.
If that makes you go "huh?", you might want to read
http://www.w3.org/TR/CSS21/box.html
and the docs for gtk_widget_compute_align().
... and implement the CSS font properties:
- font-size
- font-style
- font-family
- font-weight
- font-variant
This is the second try at this. The first was backed out previously due
to bugginess. Let's hope this one survives a bit longer.
Also makes the font-family CSS test work again.
Fixes Bug 645993 - XIM has wierd behaviors. Some XIM modules
filter every key event, possibly replacing it with their own
one. These events usually have serial=0, so make
GdkDeviceManagerXI2 also listen on these.
At http://dev.laptop.org/ticket/10643 we are seeing that drag-and-drop
within the Sugar shell causes all of Sugar's custom keybindings to be
removed.
This is because gtkdnd tries to unbind XK_KP_Space, which (on my systems)
is resolved to NoSymbol by XKeycodeToKeysym(). NoSymbol has value 0,
the same as AnyKey, and XUngrabKey(AnyKey) is equivalent to unbinding
all possible keycodes.
Fix this by catching NoSymbol before binding/unbinding.
https://bugzilla.gnome.org/show_bug.cgi?id=652402
Use a bunch of tricks to get inset/outset right with a small amount of
code. In particular, fix the hidden sides causing artifacts.
Included is a bunch of code comments explaining what we actually do.
Now instead of invalidating when we create the layout we invalidate
when we realize the widget and we remove the invalidation when
unrealizing. It was pointless too destroying the layout in unrealize
as at the end what we just wanted was to remove the invalidation idles.
This is necessary because we want to use NULL as the default value. But
the default value for borders is { 0, 0, 0, 0 } and not NULL.
Fixes border-image-gradient and border-image-repeat reftests.
This test is copied from the CSS spec. The reference is very sensitive
to the rendering algorithm in use, because the rounded corners have
semi-transparent pixels. It might be necessary to update them in the
future.
Stop documenting the base class as just a base class for the
H and V specializations, copying the useful descriptions from
those H/V classes to the base class. Do not advise the use of
the H/V classes or refer to them unnecessarily.
Now that we are not allocating treeview column buttons anymore
with invisible headers, we can't rely on their allocations for
other things like cell area computations anymore. Use x-offset
and width of the column directly, instead.
GtkDialog changes its size depending on style properties. If
we only do this in response to ::style-updated, it happens during
the initial realization of the dialog and leads to the dialog
'growing' between when we determine the initial window size and
when we allocate it that size. So, do this beforehand.
It's useful to set a slice size != border-width, as backgrounds are
clipped to border-width too.
As slices can be half-transparent and overlap the background,
this would not fill the border box properly if we only use a single
property for specifying the width.
Also, this brings us even closer to CSS3.
It's a custom property with the same semantics of text-shadow, which
works on icons and icon-like UI elements, such as spinners, arrows and
separators.
It's an entry, so it already has a background below, even when the entry
doesn't have a frame. Also, gtk_spin_button_draw_arrow() will render a
background and a frame in the arrow space anyway.
NULL is a valid value for border-image, so if it's NULL when unpacking,
don't try to access the struct fields, but just init the GValues for the
unpacked parameters.
This functionality adds a new 'Printer Profile' entry to the 'Color' page in the
UNIX print dialog if colord support is enabled.
This shows the user what color profile will be used for the settings they have
selected, and if no profile or the default profile is going to be used.
We are deliberately not allowing the user to _change_ the selected profile, as
the ICC profile is an implementation detail, and we should not change the other
print settings based on the characterization state.
The OpenICC group broadly recommend showing the profile that is used, so that
power users can be sure the correct profile is being used at the right time.
Normal users won't care, as they don't know how horrible the color match is
without profiling the printer and media.
Tests that border-image-repeat is rendered correctly.
The reftest is done via toolbars stuffed into a GtkFixed using
background-image and images tuned for it.
The new code is smaller, less crashy and correct(er), but arguably more
complex. I'd have liked to make it simpler, but this border image
algorithm is complex...
It's not useful to cache these surfaces here, as the GtkBorderImage will
be always generated on the fly, being a shorthand property.
This also allows to get rid of the intermediate image surfaces for
rendering the slices; we now use cairo_surface_create_for_rectangle()
to proxy the slices from the source surface to the rendered area, which
should also yield better performance.
It's composed by
- border-image-source: a cairo_pattern_t holding an image from file or a
gradient
- border-image-slice: a GtkBorder containing the slice offsets to apply
on the image
- border-image-repeat: a GtkRepeatType for the image
We deviate from pure CSS3 in the following ways:
* border-image-width is assumed to be always 1, i.e. always equal to
what's specified by border-width. I don't think it's a particularly
useful property to have, but we could add it later if needed.
* border-image-outset is absent, as we can't render outside of the
allocation yet.
https://bugzilla.gnome.org/show_bug.cgi?id=651194
For client-side windows, we need to queue a repaint when the background
changes. For native windows, the windowing system does take care of it,
but client-side windows are our own, so we gotta do it manually.
https://bugzilla.gnome.org/show_bug.cgi?id=652102
It seems to be very broken wrt input handling, causing random icons to
light up etc.
So until this is fixed, better remove composite support. It only removes
animations, so it's not a big loss for functionality.
We need pay attention the the request mode when doing size allocation.
The code was using request mode for requisition, but orientation
for allocation.
Also add a reftest that exhibits the issue, courtesy of
Benjamin Otte.
When GTK+ runs with inputim-ime.dll module, there is NULL
pointer reference. Because "context_ime->client_window" may
be NULL in gtk_im_context_ime_reset.
https://bugzilla.gnome.org/show_bug.cgi?id=644906
This extends the current hacks in gailmenuitem to not only
recognize menuitems with labels in them, but also menuitems
which contain a GtkCellView, since those are used by GtkComboBox
nowadays.
Some of the AtkText apis are not fully supported.
https://bugzilla.gnome.org/show_bug.cgi?id=650302
Doing it unconditionally is not useful when the subclasses actually
using them can only ever do it in the style_updated signal. So do it in
the style_updated handler instead.
This was lost when GailItem was nuked. Not having it is leading
to problems, even though randomly poking strings out of your children
seems to be a less than perfect approach to naming...
The hack in gtk_style_context_get_font() was causing segfaults in
combobox code. This is not acceptable and I'm not awake enough to fix
it, so just reverting until it's fixed sanely is easiest.
This reverts commit cf6bfbdb17.
Inheritance is now done using style property flags, so thereis no ned
anymore to clobber the pspec flags namespace.
Also, there is no need to expose this functionality in the public API
without exposing more of the styleproperty API.
This essentially reverts commit 9bfd3d2eec.
As junctions just modify the border radius and we now have a per-corner
border-radius, we can just set the border-radius to 0 for those corners
and get the same effect.
This way we avoid lots of copying when using widget paths as properties.
Keeping this as a separate patch so bugs where people were modifying a
structure they shouldn't are easily visible.
... and take an optional style property as argument. This way, we can
allow custom parse functions for properties. The style property needs to
be optional so that we can use it for widget style properties, too.
Name it _gtk_style_property_print_value() and actually pass it the style
property. This way, we can later change it to use custom print functions
for different style properties.
Include the child widget path in the returned path now. This allows
customizing the path of the current widgets - like adding flags to child
widgets (and maybe siblings in the future).
This code should be more obvious: The check only needs to check for
scrollable widgets, and only needs to compare to tab_width.
As a side note, for vis_pages == 1, tab_width will be smaller than
tab_max + stuff, so this code will not behave differently.
We don't want tabs to resize when a 2nd page is added, just because the
scroll arrows are now visible. And we do want the arrow size to be
included.
Note: Previosuly this code was never run, as the check
requisition->width < tab_width
was only run when tab_width wasn't computed, so was still equal to 0
(same for the height). So what this patch essentially does it add the
functionality of the if switch.
The call to gtk_border_free() within unpack_border() felt completely
in the wrong place, as the border actually pertains to the GValue
being unpacked. Plus, the GValue itself was also being leaked.
In finalize(), clear all rulesets.
In parse_declaration(), Free the GValue under unhandled error situations.
In gtk_css_provider_load_internal(), Do not leak the file contents.
In rgba_value_parse(), unref the symbolic color once we've resolved it.
In gradient_value_parse(), take the GtkGradient so we leave no dangling
references.
GValues stored in GtkCssRulesets are gslice managed, so don't
g_memdup() GValues from shorthand properties. This fixes
memory corruptions when reloading contents in a GtkCssProvider.
This is already done in gdk_event_source_get_filter_window(), and
could lead to wrong event assignment if an event translator happens
to return a window for an event it doesn't handle.
The old tutorial examples haven't been built in years, and are
not useful to keep around in git unless somebody does the work
to integrate them into the 'Getting started' section of the
current docs.
We want the menu realized so we know the size it's allocating to itself.
And we need that size to position the menu properly.
This is best visible on right-to-left.
Fixes sidebar in evince not showing up. This was caused by the
visiblility of the widget changing and the paned not fixing its child
window's visibility accordingly.
If the user pressed Enter to confirm the file chooser while the filename entry
was empty, then gtk_file_chooser_default_should_respond() would go back and forth
between the cases for handling the filename entry and the file list.
Don't rely on priv->resize_grip_visible as the code comment in the
variable declaration indicates.
This fixes warnings with GtkPlug, which can cause resize_grip_visible to
be TRUE but grid_window to be NULL - running tests/teststatusicon
reproduces this.
This broke with 7ef113ce56
Shorthand properties are basically the same a in CSS. For storage in
style properties or the CSS provider, they are unpacked into the real
values, so it is possible to partially override them.
No properties are yet converted to the new world yet, this is just the
code for supporting them.
... instead of duplicating code. This causes an extra g_value_copy().
If that turns out to be a performance issue, we can invent something
that handles this (like passing a gboolean take_value).
The reason for this duplication deletion is that we want to complicate
the setting code to handle shorthands by unpacking them and storing the
separate values.
Instead of initing the default style properties in the class_init
func of the style properties, init them when they are first needed -
when they are queried or when new ones are registered.
That way, they will always be available.
This will be used as a base both for parsing text-shadow and box-shadow
properties. The type is private, as there's no real use in exporting
this in a public API.
https://bugzilla.gnome.org/show_bug.cgi?id=649314
This has mostly two advantages:
- the most obvious one is the theme can render a border around the
sidebar if it wants to.
- we also can avoid hardcoding a container border width for the sidebar,
and just use a padding from the theme. This also allows different
themes to define a different padding, etc.
The drawback is we must draw the background ourselves, but it's easy
enough.
https://bugzilla.gnome.org/show_bug.cgi?id=650530
Lots of code calls gtk_menu_popup() and we don't want to resize the
window needlessly.
In this particular case, keyboard navigation to submenus caused those
submenus to shrink.
Note: I'm not sure this fix doesn't have nasty side effects, as I'm not
a specialist on menu popup code, so if it does, we'll need to revert it.
Until then, let's keep it, it fixes a bug.
This method can be implemented by event translators so they
return the right window from XGenericEventCookie events, as
ev->xany.window isn't meaningful for these.
GdkEventSource now also uses this to find out the right window
filters to apply.
If props == NULL in gtk_symbolic_color_resolve(), fail sanely for named
colors. The docs used to say it was not allowed to pass NULL for named
color, but that had problems:
1) You do not know if a color was created that way. This is especially
hard for generic users (like language bindings).
2) It wasn't even true. Colors using other symbolic colors would also
fail when trying to resolve their named colors, but the docs didn't
say so.
And because I want to use the function to resolve static colors early
where possible, I changed things.
This provides a huge speedup as we only need to preprocess style
properties when they are indeed inherited. This roughly doubles the
performance of the CSS matcher and brings the time taken by
gtk_css_provider_get_style() from 19% to 7% in my favorite benchmark.
One for the style properties, one for the widget style properties.
This way we can make one hash table by pspec which means we don't have
to repeat the pspec lookup.
Keep rulesets as an on-stack/heap structure instead of allocating all
instances separately.
Also, pass a ruleset to the ruleset parser, so we can make the ruleset
parser do lots of fancy things that might be useful for performance.
The code used the quarked name before, but when we already have the
pspec we want to have a lookup that does not involve quarking. And
lookup is equally fast if we only have the name.
Previously we got the list of all matching rules and then iterated it to
find the first one that had the property. Now we look while matching
rules, so we don't lookup rules that we don't need.
In widget hierarchy like "Foo Bar Bar Baz", we want the selector "Foo >
Bar Baz" to match, because it matches the elements 1, 2 and 4.
Previously, the selector only matches the Bar at position 3 and then
failed because it wasn't preceded by a Foo.
Instead of relying on GScanner and its idea of syntax, code up a parser
that obeys the CSS spec.
This also has the great side effect of reporting correct line numbers
and positions.
Also included is a reorganization of the returned error values. Instead
of error values describing what type of syntax error was returned, the
code just returns SYNTAX_ERROR. Other messages exist for when actual
values don't work or when errors shouldn't be fatal due to backwards
compatibility.
This is pretty important, because otherwise recursions cause crashes.
And if you accidentally change your theme to one that crashes on load,
all your gonna SEGV and then on reboot, gdm tries to load the theme...
Call gtk_css_provider_load_from_file() instead of the internal function.
This has two advantages:
1) It simplifies the code a lot
2) It gets rid of GMappedFile usage. GMappedFile does not work
everywhere, so this is finally portable.
This way, we achieve two things:
1) We can unify file loading to one location
2) We can emit the error from file loading using the parsing-error
signal. This is very useful for @import handling in particular.
Emits the error without the need for a scanner. Also simplifies
gtk_css_provider_take_error() because we now can assert an available
scanner at all times.
Instead of having an error member in the CSS provider's private struct,
connect a signal handler when an error is passed in. This has two
advantages:
1) It makes the code clearer as we don't have to keep track of an error
member anywhere.
2) It causes a non-emission of the g_warning() when an error was passed
in, because it only triggers when no signal handlers are connected.
So we get identical behavior to GTK 3.0 where warnings where only
emitted when no error was passed in.
Instead of aborting a parse whenever we encounter an error, parse to the
end. But if a GError was passed in, reset the provider completely as if
nothing had been parsed.
Value parsing only sometimes emitted errors. Sometimes it didn't emit
errors but ignored the value, sometimes it took a default, sometimes it
converted it to something it deemed suitable.
While refactoring, I moved the whole GValue <=> char * conversion
routines to a separate file, to make navigating the core css provider
easier.
Previously, we only checked for errors after parsing the full
declaration. Now we detect errors with the property before even
attempting to parse its value.
The benefit here is that the error reporting reports the correct line
and position numbers.
Tests in the parser need 1 or 2 files:
1) test.css
2) test.ref.css (optional, defaults to test.css)
The test instantiates a CSS provider, loads test.css, then dumps the
loaded file to test.out.css and then checks that that file matches
test.ref.css. If not, it dumps a diff between those two to the log and
fails.
You want to run the test with --verbose to get the output dumped to
stdout.
XKB and GDK both add "internal" bits to GdkModifierType. In C,
this typically doesn't cause problems as bitfields are just integers,
and there's no validation. However for bindings, it's normal to
convert enumerations to "native" enumeration types, which don't
support unknown bits. See bug 597292.
https://bugzilla.gnome.org/show_bug.cgi?id=634994
See the documentation in the script.
Tests are not yet added as the output from the 2 included scripts
doesn't match and the intended reference output first needs to be agreed
on.
The previous code failed to account for all child visibility and paned
mapedness invariants which could cause stray GDK windows to appear.
Not good.
Credit goes to Xan for triggering it.
GtkArrow and the align properties use different methods (float vs int)
to center the arrow. If the size of the arrow is odd, this will cause a
rendering that differs by half a pixel. So we request an even size for
both the arrow and the container and everything works out.
This is a special-purpose button that can be used together with
GPermission objects to control the sensitivity of system settings.
Suitable permission objects can e.g. be obtained from PolicyKit.
http://bugzilla.gnome.org/show_bug.cgi?id=626457
Background color was not painted on the whole tagged segment after
line breaking, this was due to the default ->prepare_run() function
wiping the internal PangoColors clean... fixed this by unconditionally
setting local rgba colors from prepare_run instead of comparing them
and bailing out.
This now allows text view to render text with alpha values in
the text foreground and backgrounds, the work is almost complete,
currently the error-underline-color is still a GdkColor style property
and since we use only GdkRGBA for rendering it needs to be converted
and applied, probably a new rgba version of the style property should
also be introduced.
This commit adds tests/testtextview that must be run from the tests/
directory to show translucent text in action.
The progressbar is composed by two different rendered areas: the trough
(i.e. the non-filled part of the bar) and the bar itself.
The bar should be able to fill the whole height/width of the trough
without resorting to nasty hacks in the theme, and we can control the
amount of space between the bar and the trough with the padding already.
https://bugzilla.gnome.org/show_bug.cgi?id=649593
GtkAssistant is widely recognized as a butt-ugly widget.
This commit changes its style to look more modern. We
deprecate the sidebar and header image properties and
don't show them anymore. Instead, page titles are arranged
in a sidebar, with the title of the current page shown
in highlighted style.
This commit simply removes some child->parent == container checks,
to add some flexibility for containers with 'inner structure'.
If these checks are considered useful, we can bring them back
with a is_child vfunc that allows container implementations to
decided who they consider legitimate child.
This is a variant of gtk_widget_child_notify() that takes an
explicit container, instead of relying on widget->parent to
be the correct container to use.
With gtk-auto-mnemonics on, we hide mnemonics on focus out. We should also
check if the modifier is pressed on focus in and if so, show mnemonics again.
https://bugzilla.gnome.org/show_bug.cgi?id=618815
This reverts commit 1c46e04f30.
The change broke too many widgets that relied on the size being
constant. A proper fix would require letting themes override the size.
That would probably also require letting themes specify the size
relative to font size.
I was hunting a memory leak and couldn't find it; at least I'm
pretty sure all of these are OK. But document things better
for the future.
Also use g_hash_table_replace in one more case for consistency.
https://bugzilla.gnome.org/show_bug.cgi?id=649457
If we have a toplevel, and not a popup window, do wait an additional
0.5s to give the WM/server enough time to actually create the window.
This is a hack and there should be a better solution. But it works.
Please use POPUP windows for tests unless the test must use toplevel
windows.
It could be the case that gdk_window_set_cursor() is called on
pointers not yet known to the device tracking code in GdkDisplay,
so update the cursor on all master pointers.
The code actually updating the cursor for the given window has
been refactored out to gdk_window_set_cursor_internal(), used
in gdk_window_set_device_cursor() as well, which makes it handle
root/foreign windows too.
https://bugzilla.gnome.org/show_bug.cgi?id=649313
Checks that the size requests for labels are as they should be for
required and natural size given various combinations of wrap, ellipsize,
width-chars and max-width-chars.
See
http://mail.gnome.org/archives/gtk-devel-list/2011-April/msg00036.html
for the discussion spawning this test.
Add a new test runner supposed to do a lot of generic tests. Run it like
this:
./gtk-reftest [OPTIONS] TESTFILE [TESTFILES...]
where FILE is a GtkBuilder ui file to run.
For a general test named "test", you want to have the following files:
1) test.ui
2) test.ref.ui
3) test.css (optional)
The test will then check that test.ui and test.ref.ui are rendered
identically with the provided css.
In detail, for every provided TESTFILE the test runner will:
1) Add the css to the default screen
2) Load the test.ui file and the test.ref.ui file
3) Grab the first GtkWindow subclass widget
4) gtk_widget_show() it and take a snapshot image of its contents into
a cairo surface.
5) Compare the two images to be bitwise identical. If they are not, a
diff image will be created hilighting the differences.
6) Save the images as png files to the output directory named:
- test.out.png (rendering of test.ui)
- test.ref.png (rendering of test.ref.ui)
- test.diff.png (optional, differences from step 5)
7) Fail the test if the two images are not bitwise identical
Credit for the idea of reftests goes to Mozilla and in particular David
Baron. For a larger introduction of why reftests are useful, see
http://weblogs.mozillazine.org/roc/archives/2008/12/reftests.html
And the getter, too: gtk_menu_item_get_right_justified() and the
corresponding property. Also make the only caller use the private
structure (as it did before the recent patches).
Fixes a regression introduced in 07d49ee5.
Libtool by default refuses to link static libraries into shared
libraries. In Windows, libuuid is however a static library and needed
for shared libgtk; as a work around, use "-Wl,-luuid" to pass the option
directly to the linker.
https://bugzilla.gnome.org/show_bug.cgi?id=642214
Also print the parent widget. As the parent's size_allocate
implementation is usually the culprit for this warning happening, it
makes sense to print it.
Glade causes such files when a notebook page has no content. And it's
especially bad because the last tab widget wins, so the label displayed
in Glade will not be what you see when you run the app.
See the code comments for the reasoning behind this. After we don't
force a "guessed" minimum size for labels anymore, a lot of issues
started to surface that this patch attempts to fix. In particular:
1) Tooltips where wrapped as much as possible.
2) The recentchooser submenu displayed only ellipsize dots.
-Update to distribute the VS2010 files.
-Added rules in Makefile.am's of GDK and GTK to fill in the
project/filter files templates with up-to-date source file
listings to simplify maintenace.
Any comments on the usage of the VS2010 files are welcome!
These are the VS2010 Project files to compile GDK and GTK+, using
the Win32 backend, along with a brief README.txt explaining the
process.
The GDK and GTK project/filter files are templates that are filled
in during "make dist" with up-to-date source file listings as far
as possible, to simplify maintenance.
Comments on their usage are most welcome.
Mnemonics for characters that go beyond the baseline (q, y, g) were not
being shown, because they are drawn outside of the label's allocated
size.
This patch just disables the clip-to-size for labels, so that the label
can draw outsize of its allocation. In most cases, that works around
this bug.
https://bugzilla.gnome.org/show_bug.cgi?id=648570
- Passing a NULL ID string should be equivalent to
gtk_combo_box_set_active (combo_box, -1).
- Use g_strcmp0() instead of strcmp() when comparing
ID strings to avoid a crash when an ID is NULL.
- Return a boolean indicating if the ID was found or not.
https://bugzilla.gnome.org/show_bug.cgi?id=647806
We want save, clip, draw1, restore, save, clip, draw2, restore
and not save, save, clip, draw1, restore, clip, draw2, restore
Functionally the same thing, but not as obvious
The zlib compressed xmlhttprequest thing was a nice hack, but it doesn't
really work in production. Its not portable, doesn't have enought API
(missing notification for closed sockets) and having to synchronize
between two different connections in a reliable way is a pain.
So, we're going everything over the websocket. This is a pure switch,
but after this we want to modify the protocol to work better over
the uncompressed utf8 transport of websockets.
Some special key keycode values as seen in keydown actually match
normal keys (like "." has a keyCode 46 on keyPress, which is the same
as Delete, but 190 for KeyDown). So we must match the special keys on
keypress. However, some things must be checked on keydown as they are not
generating keypress events.
Make the VS 2008 build instructions and information
more up-to-date and clearer to people.
Please let me know if I have missed any needed details...
Thanks
Patches by Benjamin Otte.
The "invalidate last column" hack is removed. It is now of no use since
the entire widget will be redrawn when a single column changes.
We have to redraw if we are below a couple of rows that changed height
in do_validate_rows(). This will still require a redraw for a large
amount of cases, can we do better? You would expect that a redraw of
the tree view is not required when the dy changes with the same delta
as the delta of the height accrued when validating the nodes. This
further optimization will likely require changes to the top_row/dy
synchronization code.
Introduces a 'last-folder-uri' GSettings key, where we remember the last-opened
folder from the previous instance of the file chooser.
The idea is that this works globally, across all applications, so it will be
easy to do things like
1. Save an attachment from a mail (or some other file)
2. Open another program
3. Do File/Open and automatically get sent to the folder where (1) happened.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
https://bugzilla.gnome.org/show_bug.cgi?id=644426
Copy over the implementation of _gtk_menu_shell_activate
so that C4013 warning/error (aka gcc's "implicit
declaration of ...") can be avoided when compiling with MSVC
This is in response of the changes for bug 554057 in commit
5f55c42047
We can't really know the client side keymaps, so we use the keysym
as the hardware keycode (essentially claiming to have a keyboard with
one key for all possible keysyms). This is not ideal, but its hard to
do better with no knowledge of the client side keyboard mappings.
(And html keyboard events suck badly...)
We're using the noVNC keyboard even handling model (and some of the
code with permissions). This means we combine data from keydown and
keypress to figure out the translated keysyms according to the keyboard
layout at the users machine.
This symbol needs to be exported for GDK (Win32) so that the
runtime checks for Win32 backend usage can be done on
MSVC-compiled versions of GTK+ too.
I did not add the corresponding symbols for the other backend
though-they are probably exported automatically by GCC AFAIK.
This is done to make commit
9db4accf9c
work on MSVC
Otherwise, we end up doing maximal work with minimal effect, and
bring the X server to 100% CPU. Note that this makes spinner widgets
look really unclear, but they really have to be changed to adapt
to enable-animations == FALSE at the widget level.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
