If a a treeview has frequent periodic additions and removals of
rows, it is possible that a page down keypress moves the cursor
out of the height of the treeview. In some of these cases, we
can be tricked into dereferencing a NULL pointer.
Bug 612919.
https://bugzilla.gnome.org/show_bug.cgi?id=746269
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.
https://bugzilla.gnome.org/show_bug.cgi?id=746064
Rename the install projects as gtk-install, to ease the integration of the
projects in a grand solution that may be used to build, for example, the
entire GTK+ stack.
"Install" the .pdb files to help people use them to debug the GTK+ stack,
or for their GTK-using applications, as they are already generated for all
builds.
Also update the copying of the DLLs, LIBs and EXEs so that we ensure that
we only copy the items from GTK+-2.x, without accidently copying items that
are not meant to be copied, or make extra copies of items in the wrong
places, such as when the projects here are used in parts of grand solutions
used to build the entire GTK+ stack.
The .pdb file name must be specified for Visual Studio 2010+ later in order
fo match the output filename if the output filename is different from the
project name. Update the projects as necessary.
Running gnome-shell under valgrind, I saw the attached invalid write.
Basically we can destroy a window during event processing, and the old
window_remove_filters simply called g_free() on the filter, ignoring
the refcount. Then later in event processing we call filter->refcount--,
which is writing to free()d memory.
Fix this by centralizing list mutation and refcount handling inside
a new shared _gdk_window_filter_unref() function, and using that
everywhere.
==13876== Invalid write of size 4
==13876== at 0x446B181: gdk_event_apply_filters (gdkeventsource.c:86)
==13876== by 0x446B411: _gdk_events_queue (gdkeventsource.c:188)
==13876== by 0x44437EF: gdk_display_get_event (gdkdisplay.c:410)
==13876== by 0x446B009: gdk_event_source_dispatch (gdkeventsource.c:317)
==13876== by 0x4AB7159: g_main_context_dispatch (gmain.c:2436)
==13876== by 0x4AB7957: g_main_context_iterate.clone.5 (gmain.c:3087)
==13876== by 0x4AB806A: g_main_loop_run (gmain.c:3295)
==13876== by 0x8084D6B: main (main.c:722)
==13876== Address 0x1658bcac is 12 bytes inside a block of size 16 free'd
==13876== at 0x4005EAD: free (vg_replace_malloc.c:366)
==13876== by 0x4ABE515: g_free (gmem.c:263)
==13876== by 0x444BCC9: window_remove_filters (gdkwindow.c:1873)
==13876== by 0x4454BA3: _gdk_window_destroy_hierarchy (gdkwindow.c:2043)
==13876== by 0x447BF6E: gdk_window_destroy_notify (gdkwindow-x11.c:1115)
==13876== by 0x43588E2: _gtk_socket_windowing_filter_func (gtksocket-x11.c:518)
==13876== by 0x446B170: gdk_event_apply_filters (gdkeventsource.c:79)
==13876== by 0x446B411: _gdk_events_queue (gdkeventsource.c:188)
==13876== by 0x44437EF: gdk_display_get_event (gdkdisplay.c:410)
==13876== by 0x446B009: gdk_event_source_dispatch (gdkeventsource.c:317)
==13876== by 0x4AB7159: g_main_context_dispatch (gmain.c:2436)
==13876== by 0x4AB7957: g_main_context_iterate.clone.5 (gmain.c:3087)
https://bugzilla.gnome.org/show_bug.cgi?id=637464
Backport of 806c04411d306680353cf90cffee98ce73e122f2 to
the gtk-2-24 branch. 806c0441 was authored by Colin Walters
<walters@verbum.org>
Without this patch, the spotify linux client was crashing during
playback of some songs when using gtk+ 2.24.26
https://bugzilla.gnome.org/show_bug.cgi?id=745536
Use the /MP compiler option, where the build time for release builds can
be cut down by quite a bit. This will however cause a brief warning with
debug builds due to the use of /Gm, but the code will otherwise build
normally. Unlike the Visual Studio 2010+ builds, we can't use /d2Zi+ as
Visual Studio 2008 does not support that, so we can't get a better
debugging experience for release builds here.
Use Multiprocessor compilation which can cut down build times by quite a
bit and use the /d2Zi+ flag to have better debugging info being logged to
the .pdb for release builds.
These are only applicable for Visual Studio 2010/2012 and later.
In order to provide a constant mtime between OS build and deploy time,
while also maintaining a hardlink content-addressed model independent of
timestamps, ostree sets all mtimes to 0.
The icon cache code currently ignores directories with mtime 0, assuming
they don't exist.
Track directory existence in a more precise way.
https://bugzilla.gnome.org/show_bug.cgi?id=745052
If we fail to attach pixel data to a cairo surface, free it immediately.
We still want to cairo_set_source() the surface to popagate the error.
Spotted by Christian Boxdörfer
An event filter may add or remove filters itself. This patch does
two things to address this case. The first is to take a temporary
reference to the filter while it is being used. The second is
to wait until after the filter function is run before determining
the next node in the list to process. This guards against
changes to the next node. It also does not run functions
that have been marked as removed. Though I'm not sure if this
case can arise.
https://bugzilla.gnome.org/show_bug.cgi?id=635380
Backport of 323df2b2800383832ed3c2e43626f2c6821c33ec to
the gtk-2-24 branch by Wolfgang Ulbrich.
When using the implicit paint pixmap, always draw to the same surface
(the standard one for the pixmap) each time rather than creating a new
one each time. This is both more effective and more natural.
It also fixes a redraw issue on win32, where using multiple surfaces
on the same HDC sometimes causes issues. This seems to be due to
leftover state on the HDC from previous surfaces.
https://bugzilla.gnome.org/show_bug.cgi?id=741060
Commit 85f2a721cf introduced an unconditional call to
gtk_widget_get_direction(). This does not work for 'foreign'
uses of the theme engine, e.g. in Qt, where widget is NULL.
This was reported as a crash in Fedora here:
https://bugzilla.redhat.com/show_bug.cgi?id=924683
Since GLib ≥ 2.41, attempting to release an unlocked mutex will abort(),
as it happens on most systems already.
Given the lack of proper documentation on how to use GDK with threads,
there is code in the wild that does:
gdk_threads_init ();
gdk_init ();
...
gtk_main ();
instead of the idiomatically correct:
gdk_threads_init ();
gdk_threads_enter ();
gtk_init ();
...
gtk_main ();
...
gdk_threads_leave ();
Which means that gtk_main() will try to release the GDK lock, and thus
trigger an error from GLib.
we cannot really fix all the wrong code everywhere, and since it does
not cost us anything, we can work around the issue inside GDK itself, by
trying to acquire the GDK lock inside gdk_threads_leave() with
trylock().
https://bugzilla.gnome.org/show_bug.cgi?id=735428
The function gtk_drag_set_icon_pixmap() triggered failing assertions. This was because it called the function gdk_window_get_screen(mask), where "mask" is a pixmap, but not a window.
https://bugzilla.gnome.org/show_bug.cgi?id=735005
As the Visual Studio 2012/2013 are only slightly different from the Visual
Studio 2010 projects, we can provide support for them by using scripts to
copy the Visual Studio 2010 projects, and update the specific parts as
necessary. This is being provided to help people still needing GTK+-2.x
and also to help them to transition to GTK+-3.x easier.
Thus, there would be little maintenance overhead for these as only the 2010
projects need to be kept up-to-date as a result. This might change when we
do get the stack working with WinRT/Metro, but that's going to be another
totally different issue.
We need to enclose paths containing $(BinDir) with double quotes as it
points to something like c:\foo\gtk+-x.yy.zz, which the copy command on
Windows does not like "+" in paths unless enclosed in quotes.
Since large images are in the icon cache, and apps don't tend to use that
many icons anymore, simply don't include image data and instead make apps
load files from disk. Additionally, since they're stored in GdkPixbuf data,
that means that we have to first convert them either to a cairo_surface_t,
which requires converting pixel data to be premulitplied, or an OpenGL
texture, which requires a whole GPU upload anyway.
So, even with the icon cache, the goal of icons through zero-copy, mmap()'d
data from disk just isn't doable with the icon cache format we have. The
icon cache on my disk is nearing 100MB, since we include a bunch of
high-resolution application icons, that I doubt would be used by apps at all.
Removing this inefficient pixel data makes memory usage for all applications
go down, with no speed loss.
The icon cache also, however, has an index of what icons are in each folder,
which prevents a readdir() and allows GTK+ to know what icon is where without
having to do a bunch of stat(); calls. Keeping this data is good for GTK+,
so we should still keep the index.
It doesn't make sense to remove any code for mapping pixel data from the icon
cache. There's a plan in the works to have a symbolic icon cache that does
pixel math on 16x16 icons to prevent slow SVG rendering. 16x16 pixels are
fairly small, and such images are flat colors, which should compress easily,
so the icon cache would be worthwhile here. So let's keep the code around
in preparation for that case.
https://bugzilla.gnome.org/show_bug.cgi?id=721895
