Drawing nows happens directly when the Dialog or Widget draw methods are
called. This makes it easy to debug why a particular low level draw
method was called, by inspecting the call stack.
This replaces the notion of "buffering" by two independant ways to
control what is drawn and where:
- The active layer is used to select whether the foreground or
background part of the dialogs are rendered by the draw calls.
- The active surface is used to select if the draw calls affect the back
buffer or the screen.
The foreground layer of the active dialog is drawn directly to the
screen. Its background layer is drawn to the back buffer. This way
widgets can restore the back buffer in order to update without having to
redraw the dialog's background.
Dialogs lower in the dialog stack are drawn entirely to the back buffer.
Also remove the unused linesWidth variable and fix the hlLeftPadding and
hlRightPadding widget attributes to actually work.
There are still issues remaining with the caret in the list widget due
to the ellipsis being used to shorten long text. Ellipsis is accounted
for when drawing the text but not when computing the caret position.
The previous combination of a fixed 10 milliseconds delay and time since
last update checks meant that in most cases 20 milliseconds elapsed
between two calls to updateScreen resulting in a 50 fps framerate. On
systems with wait for vsync enabled that meant that some frames were
missed.
The new frame limiter waits for a variable delay equal to the non
consumed time in the slot allocated to the frame.
Previously, a C-style cast was used to convert a
Common::Array<Plugin *>, populated with pointers to StaticPlugin
and DynamicPlugin instances, to a
Common::Array<PluginSubclass<T> *>, but PluginSubclass<T> is a
*sibling* class to StaticPlugin/DynamicPlugin, so this cast was
invalid and the results undefined. The methods for retrieving
subclasses of plugins can't be easily changed to just generate an
array of temporary wrapper objects that expose an identical API
which dereferences to the preferred PluginObject subclass because
pointers to these objects are retained by other parts of ScummVM,
so the wrappers would needed to be persisted or they would need to
just re-expose the underlying Plugin object again. This indicated
that a way to solve this problem is to have the callers receive
Plugin objects and get the PluginObject from the Plugin by
explicitly stating their desired type, in a similar manner to
std::get(std::variant), so that the pattern used by this patch to
solve the problem.
Closes gh-1051.
Updating the list reset the selection in the list widget, however if
a save had previously been selected and the Choose button was enabled,
it remained enabled despite no save being selected. Trying to load
the game resulted in a crash. This was particularly an issue with
cloud enabled as if you are unlucky you could have tried to load a
save just as the cloud sync finished, which updated the list and
unselected the save. This change fixes bug #9766: Assert in
SaveLoadChooser dialog.
In addition to adding a sanity check on the selected index for the
Choose command, this commit also preserves the selection when
updating the list as I think this would be the expected behaviour
in this dialog.
