We need to cleanup state here immediately so that we do not potentially
access the g_class private data after it been finalized. This ensures that
the borrowed reference is dropped by the muxer.
Before this commit, adding GtkWidgetAction to class private data would
require copying the actions to each subclass as they were built or
modified. This was convenient in that it is a sort of "copy on write"
semantic.
However, due to the way that GTypeInstance works with base _init()
functions, the "g_class" pointer in GTypeInstance is updated as each
_init() function is called. That means you cannot access the subclasses
class private data, but only the parent class private data.
If instead we use a singly linked list of GtkWidgetAction, each subclass
has their own "head" yet all subclasses share the tail of the
GtkWidgetAction chain.
This creates one bit of complexity though. You need a stable way to know
which "bit" is the "enabled" bit of the action so we can track enabled
GAction state. That is easily solved by calculating the distance to the
end of the chain for a given action so that base classes sort ahead of
subclasses. Since the parent class always knows its parent's actions, the
position is stable.
A new dynamic bitarray helper also helps us avoid allocations in all the
current cases (up to 64 actions per widget) and dynamically switches to
malloc if that is to ever be exceeded.
I've seen a crash when the action muxer gets
disposed during widget destroy, and tries to
disconnect from widget signals too late.
There is no real need to disconnect, since the
only time an action muxer is going away is when
its widget is destroyed, so just don't do it.
Previously, we would not look any further for
an action once we found a match for the prefix,
defining inheritance by groups. Change this to
inheritance for individual actions, since we
are moving towards individual actions as the
main ingredient in GTKs action support.
The only cases of stateful actions we've seen
so far have been boolean properties, and we
don't really want to add much state handling
API, so lets just go with property actions
for now.
Adapt the only user in GtkText.
Add a facility to register and install actions
at class init time. The intended use for these
actions is for
a) context and other model-based menus
b) key bindings
Most of these actions are going to be stateless,
so add separate apis for the simple and stateful
cases.
We avoid creating an action group for these by
teaching the action muxer about these actions.
The action muxer also maintains the enabled
state for these actions.
Instead of duplicating this code in multiple places,
add an api to look up an action group for a prefixed
name, and return the unprefixed name at the same time.
This adds specific marshallers for all of the locations where a generic
marshaller is being used. It also provides va_marshallers to reduce the
chances that we get stack traces from perf going through ffi_call_unix64.
This is forward ported from gtk-3-24.
# Conflicts:
# gtk/gtkeventcontrollerkey.c
# gtk/gtkeventcontrollermotion.c
# gtk/gtkgesture.c
# gtk/gtkgesturemultipress.c
These functions, while added for use by the GTK inspector, are generally
useful to applications that need to resolve what action groups are
available to a particular GtkWidget.
https://bugzilla.gnome.org/show_bug.cgi?id=741386
Reuse the existing infrastructure in GtkActionMuxer for propagation of
accelerator information: in particular, what accel label ought to appear
on menu items for a particular action and target.
This is a good idea because we want accels to travel along with the
actions that they're tied to and reusing GtkActionMuxer will allow us to
do that without creating another hierarchy of a different class for the
sole purpose of filling in accel labels on menu items.
Doing it this ways also allows those who copy/paste GtkActionMuxer to
insert the accels for themselves.
Add a new method on the GtkActionObserver interface to report changes.
This patch introduces a new concept: "action and target" notation for
actions. This format looks like so:
"'target'|app.action"
or for non-targeted actions:
"|app.action"
and it is used over a number of possible alternative formats for some
good reasons:
- it's very easy to get a nul-terminated action name out of this format
when we need it, by using strrchr('|') + 1
- we can also get the target out of it using g_variant_parse() because
this function takes a pointer to a 'limit' character that is not
parsed past: we use the '|' for this
- it's extremely easy to hash on this format (just use a normal string
hash) vs. attempting to hash on a string plus a GVariant
A close contender was to use detailed action strings here, but these are
not used for two reasons:
- it's not possible to easily get the action name or target out of the
strings without more work than the "action and target" format
requires
- we still intend to use detailed action strings on API (since they are
a lot nicer to look at) but detailed action strings can be given in
non-canonical forms (eg: 'foo::bar' and 'foo("bar")' are equivalent)
so we'd have to go through a normalisation step anyway. Since we're
doing that already, we may as well convert to a more convenient
internal format.
This new "action and target" format is going to start appearing in a lot
more places as action descriptions are introduced.
I suspect that nobody is using '|' in their action names, but in case I
am proven wrong, we can always switch to using something more exotic as
a separator character (such as '\x01' or '\xff' or the like).
Rename our internal GActionMuxer, GActionObserver and GActionObservable
classes and interfaces to have names in our own namespace.
These classes were originally intended for GIO but turned out to be too
special-purpose to be useful there, so we never made them public API but
have just been copying them around (without bothering to properly rename
them). Now that other people will be copying them out of Gtk, it's even
more important to prevent this namespace abuse from spreading further.
