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
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)
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.
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.
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>
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
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
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).
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.
