These are already slightly adjust to be in a flat subdirectory.
Original code can be found at https://github.com/marekdedic/BlendSplitter
and was MIT licensed when converted.

Simply moved the expander.png in our already existing qrc file.
Mainly since I didn't feel like working out how to incorporate 2 qrc
files in our buildsystem :-)

Better coloring when widget disabled.
Highlight current SP address.

Currently the old view is in a splitter visible on first workspace.

The old code had a lot of direct signals/slots and calls between
DebuggerForm and all the *Viewer code To decouple this the
SignalDispatcher was created as a central way to forward signals to all
views that registers. Also the CPU registers state code now also lives
in the SignalDispatcher.

This means that now we can have multiple DisasmViewers since the code
doesn't rely on one and exactly one disasmView to exist.

As a side effect the QAction that would toggle breakpoints are now
disabled, since that code can no longer as the cursorAdress from the
one-and-only disasmView the old code assumed.

After splitting you end up with two SwitchingWidgets of the same type

In effect only BlendSplitter layout in use now

Also tried some fancier popup menu on the SplitterHandle, but this
causes a lot of crashes. I assume that the handler is already free'ed while the
mouseReleaseEvent is still being processed causing these SegFaults.

Also I got divide by zero errors when creating new StackViewer with an open
openMSX connection, it seems to handle the command reply before the widget is
fully initialized???

Also the changing of background (when connecting/disconnecting the emulator) in
the scrollarea isn't eassily triggered. If you resize by hand it will correctly
change the color not covered by the single widget, but simply setting an new
background doesn't cause an update, neither doesn explicitely calling update()
or repaint().

This widget simply contains a richtext label to explain the new workings
of the debugger.

The idea is that this is the default widget for a new SwitchingWidget
until the user is familiar with the workings and we have a setting to
switch this behaviour to a more regular blender like splitting of the
current area.

Use JSon format to save and load the current workspaces
TODO: set the enabledWidget status to current connection status.
Maybe save setting from the widgets in the SwitchingWidgets also?

And prevent direct emitting of the signal so that the SignalDispatcher
is the single source of truth.

In preferences you can now can select a default workspace to load when
starting. Also added a 'new user' setting so that the QuickGuide is
clearly visible for new users.
The CPU view/code debugger is now a workspace that can be added/removed.

Moved code in single class, and improved usage.
ESC will now cancel any ongoing edits.
Clicking on other tabs will now close the lineEdit.
And the tabs are now movable.

Thanks to a quick review of Wouter

Since there is code that doesn't use ui files, we now manually set the
QObject names to make debugging simpler

This is going to replace the PaletteDialog that is now called from the
TileViewer and SpriteViewer, but was not yet used in the bitmapviewer...

As soon as those viewers are adapted, we will remove the PaletteDialog

The code viewer now tracks the PC register again.
Unfortunately the old code had some implicit assumptions about the order in
which the signals arrived (hence the disasmStatus enum)

For now the DisasmViewer::setProgramCounter is called twice with a different
infoLine parameter, which may cause the code displayed to "jump up and down"
during debugging

Because it doesn't make a full copy. That is after the assignment source
and destination aren't equal.

It can be trivially calculated from myPal[].

And not having to keep this array up-to-date allowed to simplify various
other functions.

This way we can 'correctly' perfom a split so that both the old and new
widgets have the same options selected. We can also save some of these
settings when we save layouts.

Unfortunately for stuff like the DebuggableViewer the actual text of the
combobox depends on the debuggables that the selected openMSX presents
to the debugger, so saving the selection in a layout will not work.
(Unless we connect to an openMSX that presents the same debuggables
before we even try to restore the layout? That would be a nice
improvement...)

Trying to keep up to date with the master branch
Mainly signal/slot connections

First steps towards an autoconnect feature.
The idea is that the debugger can autoconnect if there is only one
openMSX running. If there is no running openMSX we should start one.
In case of two or more the dialog should pop up.

When the 'slots:' indicator is removed in the header files the names do
no longer end up in the QMetaObject. This is not needed with the new Qt5
way of hooking up signal/slots, but the ui files use this to autoconnect
the slots to GUI elements by the on_<widgetname>_<signalname> naming
convention.
This no longer worked so we are fixing that here.
The VDPCommandRegViewer will be fixed more correctly later :-)

* Consistent indentation:
We currently have a mix of indentation style: sometimes we use 1 tab,
sometimes we use 4 spaces. I don't have a strong preference. But do use
the same style within a single function (preferably also within the
whole file). I guess some code was edited using an editor which did not
have tab-stops at 8 spaces. This lead to horribly inconsistent
(unreadable) indentation in some functions.

* Various typographical conventions:
** no space before comma, space after comma
** space before and after operators
** no space before or after parenthesis

Avoid repeating the same lookup twice (and risk making a typo between
the 1st / 2nd string literal).

No need to implement an empty default constructor.

No need to explicitly call the default constructor of a base-class.

…e same scope

Commit 2c3fd4c removed 'slot' annotations from header files. But this
(unintentionally) broke automatic (string-based) connection setup via
the 'setupUi()' call.

Commit 205a992 already partially fixed this. This commit does the
rest.

Let's get rid of the old PaletteDialog and use the MSXPalette as it was
intended to be used.

Now that the PaletteViewer has taken over all the functions of the
PaletteDialog we remove all traces of this old dialog.

The autosync is only meant for the VDP palette to sync to the current
openMSX session, so make this more clear in the GUI.

Now we can decode 8 bytes as a character.
And since character definitions do not need to be aligned in
memory the hexviewer address can now be unaligned as well.

Keeping up to date with the latest work on the BreakpointViewer

Take decoded chars width into account.
Cyan line in decoded char to indicate marker address.
Allow shift+left/right when in unaligned mode to shift addresses in
single byte steps.

Since openMSX 18.0-106-ge55a4a904 (commit e55a4a9044b...) the TCL
command 'machine_info device VDP' now also returns the version string
from the machines XML config. This is now used to select the correct
view for the VDPRegViewer when decoding the VDP registers.