Remove the roots vector again, with selective inlining it's better
to use the "live" data structures instead of a copy of the roots.
It would be nice to centralize that into a helper function that returns
say an `Ref<'_, impl Iterator<Item = &Rc<Component>>>>` but that's not
possible yet.
The same chained iterator pattern is also used in some passes already.
Make sure that doc.root_component.used_types.structs has all the
structs, instead of storing them in used_types for each sub-component.
That should also account for the same structures used by different
components.
The generation is still incomplete, but this change passes them to the C++ generator
when not inlining.
Another option would've been a "inline: bool", but that won't suffice in
the future when we may want to use heuristics to selectively inline some
and others not. And always taking used_types.sub_components is wrong
when inlining as those are ... inlined.
`SIXTYFPS_DISABLE_INLINING=true` will do that and subsequently apply all the optimizations passes
on all components instead of just the root.
In the future we'll flip this around make make inlining opt-in, instead
of opt-out.
Revert part of the previous commit that tries to do all the pass without inlining
Fixup the few first passes so they work without inlining, but still do a full
inlining for most passes
The goal is to make it work pass by pass until we can have everything without
requiring full inlining
Similar to the parent commit, the model tracks changes to the rows and
marks an internal property dirty. Since we have a base class this is a
little less intrusive.
cc #98
This is done by exposing the ModelNotify to the caller via the new
ModelTracker trait, which has a function that allows "hooking" into the
dirty tracking of the size.
By extension, this also works in JavaScript.
cc #98
The Flickable should already have sane default
But we need to make sure that the default binding don't override bindings
set with aliases. So we must use BindingsMap::set_binding_if_not_set which
sets the priority properly
Closes#581
Two problems:
- We were not marking the property as dirty, so dependent property would not
update themselves
- In the generated rust/c++ code, we would not call set_animated_value
So that the `debug_assert!` that fires when a error message ends with a
period does not crash the LSP when typing an incomplete filename that ends
with a '.'
