Glade Virtual Control Panel

GladeVCP is an LinuxCNC component which adds the ability to add a new user interface panel to LinuxCNC user interfaces like Axis or Touchy. Unlike PyVCP, GladeVCP is not limitied to displaying and setting HAL pins, as arbitrary actions can be executed in Python code - in fact, a complete LinuxCNC user interface could be built with GladeVCP and Python.

GladeVCP users the Glade WYSIWYG user interface editor, which makes it easy to create visually pleasing panels. It relies on the PyGTK bindings to the rich GTK+ widget set, and in fact all of these may be used in a GladeVCP application - not just the specialized widgets for interacting with HAL and LinuxCNC, which are documented here.

Both support the creation of panels with HAL widgets - user interface elements like LED’s, buttons, sliders etc whose values are linked to a HAL pin, which in turn interfaces to the rest of LinuxCNC.

PyVCP:

GladeVCP:

GladeVCP panel windows may be run in three different setups:

If your using an installed version of LinuxCNC the examples shown below are in the configuration picker in the Sample Configurations > sim > gladevcp branch.

The following instructions only apply if your using a git checkout. Open a terminal and change to the directory created by git then issue the commmands as shown.

Run the sample GladeVCP panel integrated into Axis like PyVCP as follows:

Run the same panel, but as a tab inside Axis:

To run this panel as a standalone toplevel window besides Axis, just start Axis in the background and start gladevcp as follows:

To run this panel inside Touchy:

Functionally these setups are identical - they only differ in screen real estate requirements and visibility. Since it is possible to run several GladeVCP components in parallel (with different HAL component names), mixed setups are possible as well - for instance a panel on the right hand side, and one or more tabs for less-frequently used parts of the interface.

While running Axis, explore Show HAL Configuration - you will find the gladevcp HAL component and may observe their pin values while interacting with the widgets in the panel. The HAL setup can be found in configs/gladevcp/manual-example.hal.

The example panel has two frames at the bottom. The panel is configured so that resetting ESTOP activates the Settings frame and turning the machine on enables the Commands frame at the bottom. The HAL widgets in the Settings frame are linked to LEDs and labels in the Status frame, and to the current and prepared tool number - play with them to see the effect. Executing the T<toolnumber> and M6 commands in the MDI window will change the current and prepared tool number fields.

The buttons in the Commands frame are MDI Action widgets - pressing them will execute an MDI command in the interpreter. The third button Execute Oword subroutine is an advanced example - it takes several HAL pin values from the Settings frame, and passes them as parameters to the Oword subroutine. The actual parameters received by the routine are displayed by (DEBUG, ) commands - see configs/gladevcp/nc_files/oword.ngc for the subroutine body.

To see how the panel is integrated into Axis, see the [DISPLAY]GLADEVCP statements in gladevcp_panel.ui, and the [DISPLAY]EMBED* and [HAL]POSTGUI_HALFILE statements in gladevcp_tab.ini respectively.

The user interface is created with the glade UI editor - to explore it, you need to have glade installed. To edit the user interface, run the command

The center window shows the appearance of the UI. All user interface objects and support objects are found in the right top window, where you can select a specific widget (or by clicking on it in the center window). The properties of the selected widget are displayed, and can be changed, in the right bottom window.

To see how MDI commands are passed from the MDI Action widgets, explore the widgets listed under Actions in the top right window, and in the right bottom window, unter the General tab, the MDI command property.

See how a Python callback is integrated into the example:

The is just touching upon the concept - the callback mechanism will be handled in more detail in the GladeVCP Programming section.

To view or modify Glade UI files, you need glade installed - it is not needed just to run a GladeVCP panel. If the glade command is missing, install it with the command:

Verify the version number to be greater than 3.6.7:

This section just outlines the initial LinuxCNC-specific steps. For more information and a tutorial on glade, see http://glade.gnome.org. Some glade tips & tricks may also be found on youtube.

Either modify an existing UI component by running glade <file>.ui or start a new one by just running the glade command from the shell.

This will look like so:

Glade tends to write a lot of messages to the shell window, which mostly can be ignored. Select File→Save as, give it a name like myui.ui and make sure it’s saved as GtkBuilder file (radio button left bottom corner in Save dialog). GladeVCP will also process the older libglade format correctly but there is no point in using it. The convention for GtkBuilder file extension is .ui.

You’re now ready to give it a try (while LinuxCNC, e.g. Axis is running) it with:

GladeVCP creates a HAL component named like the basename of the UI file - myui in this case - unless overriden by the -c <component name> option. If running Axis, just try Show HAL configuration and inspect its pins.

You might wonder why widgets contained a HAL_Hbox or HAL_Table appear greyed out (inactive). HAL containers have an associated HAL pin which is off by default, which causes all contained widgets to render inactive. A common use case would be to associate these container HAL pins with halui.machine.is-on or one of the halui.mode. signals, to assure some widgets appear active only in a certain state.

To just activate a container, execute the HAL command setp gladevcp.<container-name> 1.

The suggested way of linking HAL pins in a GladeVCP panel is to collect them in a separate file with extension .hal. This file is passed via the POSTGUI_HALFILE= option in the HAL section of your ini file.

Place the GladeVCP panel in the righthand side panel by specifying the following in the ini file:

The HAL component name of a GladeVCP application started with the the GLADEVCP option is fixed: gladevcp. The command line actually run by Axis in the above configuration is as follows:

This means you may add arbitrary gladevcp options here, as long as they dont collide with the above command line options.

To do so, edit your .ini file and add to the DISPLAY and HAL sections of ini file as follows:

Note the halcmd loadusr way of starting the tab command - this assures that POSTGUI_HALFILE will only be run after the HAL component is ready. In rare cases you might run a a command here which uses a tab but does not have an associated HAL component. Such a command can be started without halcmd loadusr, and this signifies to Axis that it does not have to wait for a HAL component since there is none.

When changing the component name in the above example, note that the names used in -Wn <component> and -c <component> must be identical.

Try it out by running Axis - there should be a new tab called GladeVCP demo near the DRO tab. Select that tab, you should see the example panel nicely fit within Axis.

To do add a GladeVCP tab to Touchy, edit your .ini file as follows:

Note the following differences to the Axis tab setup:

See also man gladevcp . These are the gladevcp command line options:

Usage: gladevcp [options] myfile.ui

Options:

The integration steps outlined above look a bit tricky, and they are. It does therefore help to understand the startup process of LinuxCNC and how this relates to gladeVCP.

The normal LinuxCNC startup process does the following:

The introduction of gladeVCP brought the following issue:

This is the purpose of the POSTGUI_HALFILE. This ini option is inspected by the GUIs. If a GUI detects this option, it runs the corresponding HAl file after any embedded gladVCP panel is set up. However, it does not check whether a gladeVCP panel is actually used, in which case the HAL cmd file is just run normally. So if you do NOT start gladeVCP through GLADEVCP or EMBED_TAB etc, but later in a separate shell window or some other mechanism, a HAL command file in POSTGUI_HALFILE will be executed too early. Assuming gladeVCP pins are referenced herein, this will fail with an error message indicating that the gladeVCP HAL component is not available.

So, in case you run gladeVCP from a separate shell window (i.e. not started by the GUI in an embedded fashion):

GladeVcp includes a collection of Gtk widgets with attached HAL pins called HAL Widgets, intended to control, display or otherwise interact with the LinuxCNC HAL layer. They are intended to be used with the Glade user interface editor. With proper installation, the HAL Widgets should show up in Glade’s HAL Python widget group. Many HAL specific fields in the Glade General section have an associated mouse-over tool tip.

HAL signals come in two variants, bits and numbers. Bits are off/on signals. Numbers can be "float", "s32" or "u32". For more information on HAL data types see the HAL manual. The GladeVcp widgets can either display the value of the signal with an indicator widget, or modify the signal value with a control widget. Thus there are four classes of GladeVcp widgets that you can connect to a HAL signal. Another class of helper widgets allow you to organize and label your panel.

Most HAL widgets have a single associated HAL pin with the same HAL name as the widget (glade: General→Name).

Exceptions to this rule currently are.

HAL widgets are instances of GtKWidgets and hence inherit the methods, properties and signals of the applicable GtkWidget class. For instance, to figure out which GtkWidget-related methods, properties and signals a HAL_Button has, lookup the description of GtkButton in the PyGtk Reference Manual.

An easy way to find out the inheritance relationship of a given HAL widget is as follows: run glade, place the widget in a window, and select it; then choose the Signals tab in the Properties window. For example, selecting a HAL_LED widget, this will show that a HAL_LED is derived from a GtkWidget, which in turn is derived from a GtkObject, and eventually a GObject.

HAL Widgets also have a few HAL-specific Python attributes:

The are several HAL-specific methods of HAL Widgets, but the only relevant method is:

As a general rule, if you need to set a HAL output widget’s value from Python code, do so by calling the underlying Gtk setter (e.g. set_active(), set_value()) - do not try to set the associated pin’s value by halcomp[pinname] = value directly because the widget will not take notice of the change!.

It might be tempting to set HAL widget input pins programmatically. Note this defeats the purpose of an input pin in the first place - it should be linked to, and react to signals generated by other HAL components. While there is currently no write protection on writing to input pins in HAL Python, this doesn’t make sense. You might use setp pinname value in the associated halfile for testing though.

It is perfectly OK to set an output HAL pin’s value with halcomp[pinname] = value provided this HAL pin is not associated with a widget, that is, has been created by the hal_glib.GPin(halcomp.newpin(<name>,<type>,<direction>) method (see GladeVCP Programming for an example).

Event-driven programming means that the UI tells your code when "something happens" - through a callback, like when a button was pressed. The output HAL widgets (those which display a HAL pin’s value) like LED, Bar, VBar, Meter etc, support the hal-pin-changed signal which may cause a callback into your Python code when - well, a HAL pin changes its value. This means there’s no more need for permanent polling of HAL pin changes in your code, the widgets do that in the background and let you know.

Here is an example how to set a hal-pin-changed signal for a HAL_LED in the Glade UI editor:

The example in configs/gladevcp/examples/complex shows how this is handled in Python.

This group of widgets are derived from various Gtk buttons and consists of HAL_Button, HAL_ToggleButton, HAL_RadioButton and CheckButton widgets. All of them have a single output BIT pin named identical to the widget. Buttons have no additional properties compared to their base Gtk classes.

Check button: Radio buttons: Toggle button: .

HAL_HScale and HAL_VScale are derived from the GtkHScale and GtkVScale respectively. They have one output FLOAT pin with name equal to widget name. Scales have no additional properties.

To make a scale useful in Glade, add an Adjustment (General→Adjustment→New or existing adjustment) and edit the adjustment object. It defines the default/min/max/increment values. Also, set adjustment Page size and Page increment to zero to avoid warnings.

Example HAL_HScale: .

HAL SpinButton is derived from GtkSpinButton and holds two pins:

To be useful, Spinbuttons need an adjustment value like scales, see above.

Example SpinButton: .

HAL_Label is a simple widget based on GtkLabel which represents a HAL pin value in a user-defined format.

Compared to their Gtk counterparts they have one input BIT pin which controls if their child widgets are sensitive or not. If the pin is low then child widgets are inactive which is the default.

The hal_led simulates a real indicator LED. It has a single input BIT pin which controls it’s state: ON or OFF. LEDs have several properties which control their look and feel:

As an input widget, LED also supports the hal-pin-changed signal. If you want to get a notification in your code when the LED’s HAL pin was changed, then connect this signal to a handler, for example on_led_pin_changed and provide the handler as follows:

This will be called at any edge of the signal and also during program start up to report the current value.

Example LEDs: .

The HAL_ProgressBar is derived from gtk.ProgressBar and has two float HAL input pins:

It has the following properties:

Example HAL_ProgressBar: .

HAL_ComboBox is derived from gtk.ComboBox. It enables choice of a value from a dropdown list.

It exports two HAL pins:

It has the following property which can be set in Glade:

In default mode this widgets sets the pins to the index of the chosen list entry. So if your widget has three labels, it may only assume values 0,1 and 2.

In column mode (column > -1), the value reported is chosen from the ListStore array as defined in Glade. So typically your widget definition would have two columns in the ListStore , one with text displayed in the dropdown, and an int or float value to use for that choice.

There’s an example in configs/gladevcp/by-widget/combobox/combobox.{py,ui} which uses column mode to pick a float value from the ListStore.

If you’re confused like me about how to edit ComboBox ListStores and CellRenderer, see http://www.youtube.com/watch?v=Z5_F-rW2cL8.

HAL Bar and VBar widgets for horizontal and vertical bars representing float values. They have one input FLOAT hal pin. Both bars have the following properties:

Horizontal bar: Vertical bar: .

HAL Meter is a widget similar to PyVCP meter - it represents a float value and has one input FLOAT hal pin. HAL Meter has the following properties:

Example HAL Meters: .

Gremlin is a plot preview widget similar to the Axis preview window. It assumes a running LinuxCNC environment like Axis or Touchy. To connect to it, inspects the INI_FILE_NAME environment variable. Gremlin displays the current .ngc file - it does monitor for changes and reloads the ngc file if the file name in Axis/Touchy changes. If you run it in a GladeVCP application when LinuxCNC is not running, you might get a traceback because the Gremlin widget can’t find LinuxCNC status, like the current file name.

Gremlin does not export any HAL pins. It has the following properties:

Example:

For some applications it might be desirable to have background image - like a functional diagram - and position widgets at appropriate places in that diagram. A good combination is setting a bitmap background image, like from a .png file, making the gladevcp window fixed-size, and use the glade Fixed widget to position widgets on this image.

The code for the below example can be found in configs/gladevcp/animated-backdrop:

GladeVcp includes a collection of "canned actions" called EMC Action Widgets for the Glade user interface editor. Other than HAL widgets, which interact with HAL pins, EMC Actions interact with LinuxCNC and the G-code interpreter.

EMC Action Widgets are derived from the Gtk.Action widget. The Action widget in a nutshell:

The appearance of EMC Actions in Glade is roughly as follows:

Tooltip hovers provide a description.

EMC Action widgets are one-shot type widgets. They implement a single action and are for use in simple buttons, menu entries or radio/check groups.

These are bi-modal widgets. They implement two actions or use a second (usually pressed) state to indicate that currently an action is running. Toggle actions are aimed for use in ToggleButtons, ToggleToolButtons or toggling menu items. A simplex example is the ESTOP toggle button.

Currently the following widgets are available:

The following toggle actions have only one associated command and use the pressed state to indicate that the requested operation is running:

These widgets provide a means to execute arbitrary MDI commands. The Action_MDI widget does not wait for command completion as the Action_MDI Toggle does, which remains disabled until command complete.

configs/gladevcp/mdi-command-example/whoareyou.ui is a Glade UI file which conveys the basics:

Open it in Glade and study how it’s done. Start Axis, and then start this from a terminal window with gladevcp whoareyou.ui. See the hal_action_mdi1 Action and it’s MDI command property - this just executes (MSG, "Hi, I’m an EMC_Action_MDI") so there should be a message popup in Axis like so:

You’ll notice that the button associated with the Action_MDI action is grayed out if the machine is off, in E-Stop or the interpreter is running. It will automatically become active when the machine is turned on and out of E-Stop, and the program is idle.

Optionally, MDI command strings may have parameters substituted before they are passed to the interpreter. Parameters currently may be names of HAL pins in the GladeVCP component. This is how it works:

The example UI file is configs/gladevcp/mdi-command-example/speed.ui. Here’s what you get when running it:

It’s perfectly OK to call an O-word subroutine in an MDI command, and pass HAL pin values as actual parameters. An example UI file is in configs/gladevcp/mdi-command-example/owordsub.ui.

Place configs/gladevcp/nc_files/oword.ngc so Axis can find it, and run gladevcp owordsub.ui from a terminal window. This looks like so:

The LinuxCNC G-Code interpreter has a single global set of variables, like feed, spindle speed, relative/absolute mode and others. If you use G code commands or O-word subs, some of these variables might get changed by the command or subroutine - for example, a probing subroutine will very likely set the feed value quite low. With no further precautions, your previous feed setting will be overwritten by the probing subroutine’s value.

To deal with this surprising and undesirable side effect of a given O-word subroutine or G-code statement executed with an LinuxCNC ToggleAction_MDI, you might associate pre-MDI and post-MDI handlers with a given LinuxCNC ToggleAction_MDI. These handlers are optional and provide a way to save any state before executing the MDI Action, and to restore it to previous values afterwards. The signal names are mdi-command-start and mdi-command-stop; the handler names can be set in Glade like any other handler.

Here’s an example how a feed value might be saved and restored by such handlers (note that LinuxCNC command and status channels are available as self.linuxcnc and self.stat through the EMC_ActionBase class:

Only the Action_MDI Toggle widget supports these signals.

Many actions depend on LinuxCNC status - is it in manual, MDI or auto mode? is a program running, paused or idle? You cannot start an MDI command while a G-code program is running, so this needs to be taken care of. Many LinuxCNC actions take care of this themselves, and related buttons and menu entries are deactivated when the operation is currently impossible.

When using Python event handlers - which are at a lower level than Actions - one needs to take care of dealing with status dependencies oneself. For this purpose, there’s the LinuxCNC Stat widget: to associate LinuxCNC status changes with event handlers.

LinuxCNC Stat has no visible component - you just add it to your UI with Glade. Once added, you can associate handlers with its following signals:

Most widget sets, and their associated user interface editors, support the concept of callbacks - functions in user-written code which are executed when something happens in the UI - events like mouse clicks, characters typed, mouse movement, timer events, window hiding and exposure and so forth.

HAL output widgets typically map input-type events like a button press to a value change of the associated HAL pin by means of such a - predefined - callback. Within PyVCP, this is really the only type of event handling supported - doing something more complex, like executing MDI commands to call a G-code subroutine, is not supported.

Within GladeVCP, HAL pin changes are just one type of the general class of events (called signals) in GTK+. Most widgets may originate such signals, and the Glade editor supports associating such a signal with a Python method or function name.

If you decide to use user-defined actions, your job is to write a Python module whose class methods - or in the simple case, just functions - can be referred to in Glade as event handlers. GladeVCP provides a way to import your module(s) at startup and will automatically link your event handlers with the widget signals as set in the Glade UI description.

This is just a minimal example to convey the idea - details are laid out in the rest of this section.

GladeVCP can not only manipulate or display HAL pins, you can also write regular event handlers in Python. This could be used, among others, to execute MDI commands. Here’s how you do it:

Write a Python module like so and save as e.g. handlers.py:

In Glade, define a button or HAL button, select the Signals tab, and in the GtkButton properties select the pressed line. Enter on_button_press there, and save the Glade file.

Then add the option -u handlers.py to the gladevcp command line. If your event handlers are spread over several files, just add multiple -u <pyfilename> options.

Now, pressing the button should change its label since it’s set in the callback function.

What the -u flag does: all Python functions in this file are collected and setup as potential callback handlers for your Gtk widgets - they can be referenced from Glade Signals tabs. The callback handlers are called with the particular object instance as parameter, like the GtkButton instance above, so you can apply any GtkButton method from there.

Or do some more useful stuff, like calling an MDI command!

HAL input widgets, like a LED, automatically associate their HAL pin state (on/off) with the optical appearance of the widget (LED lit/dark).

Beyond this builtin functionality, one may associate a change callback with any HAL pin, including those of predefined HAL widgets. This fits nicely with the event-driven structure of a typical widget application: every activity, be it mouse click, key, timer expired, or the change of a HAL pin’s value, generates a callback and is handled by the same orthogonal mechanism.

For user-defined HAL pins not associated with a particular HAL widget, the signal name is value-changed. See the Adding HAL pins section below for details.

HAL widgets come with a pre-defined signal called hal-pin-changed. See the Hal Widgets section for details.

The overall approach is as follows:

For simple tasks it’s sufficient to define functions named after the Glade signal handlers. These will be called when the corresponding event happens in the widget tree. Here’s a trivial example - it assumes that the pressed signal of a Gtk Button or HAL Button is linked to a callback called on_button_press:

Add this function to a Python file and run as follows:

Note communication between handlers has to go through global variables, which does not scale well and is positively un-pythonic. This is why we came up with the class-based handler model.

The idea here is: handlers are linked to class methods. The underlying class(es) are instantiated and inspected during GladeVCP startup and linked to the widget tree as signal handlers. So the task now is to write:

Here is a minimum user-defined handler example module:

Now, on_this_signal will be available as signal handler to your widget tree.

If during module inspection GladeVCP finds a function get_handlers, it calls it as follows:

the arguments are:

GladeVCP then inspects the list of class instances and retrieves their method names. Qualifying method names are connected to the widget tree as signal handlers. Only method names which do not begin with an _ (underscore) are considered.

Note that regardless whether you’re using the libglade or the new GtkBuilder format for your Glade UI, widgets can always be referred to as builder.get_object(<widgetname>). Also, the complete list of widgets is available as builder.get_objects() regardless of UI format.

It is important to know in which state of affairs your get_handlers() function is called so you know what is safe to do there and what not. First, modules are imported and initialized in command line order. After successful import, get_handlers() is called in the following state:

Once all modules have been imported and method names extracted, the following steps happen:

So when your handler class is initialized, all widgets are existent but not yet realized (displayed on screen). And the HAL component isn’t ready as well, so its unsafe to access pins values in your __init__() method.

If you want to have a callback to execute at program start after it is safe to access HAL pins, then a connect a handler to the realize signal of the top level window1 (which might be its only real purpose). At this point GladeVCP is done with all setup tasks, the halfile has been run, and GladeVCP is about to enter the Gtk main loop.

Within a class, method names must be unique. However, it is OK to have multiple class instances passed to GladeVCP by get_handlers() with identically named methods. When the corresponding signal occurs, these methods will be called in definition order - module by module, and within a module, in the order class instances are returned by get_handlers().

Instead of extending GladeVCP for any conceivable option which could potentially be useful for a handler class, you may use the -U <useroption> flag (repeatedly if you wish). This flag collects a list of <useroption> strings. This list is passed to the get_handlers() function (useropts argument). Your code is free to interpret these strings as you see fit. An possible usage would be to pass them to the Python exec function in your get_handlers() as follows:

This way you can pass arbitrary Python statements to your module through the gladevcp -U option, for example:

This should set debug to 2 and confirm that your module actually did it.

A annoying aspect of GladeVCP in its earlier form and pyvcp is the fact that you may change values and HAL pins through text entry, sliders, spin boxes, toggle buttons etc, but their settings are not saved and restored at the next run of LinuxCNC - they start at the default value as set in the panel or widget definition.

GladeVCP has an easy-to-use mechanism to save and restore the state of HAL widgets, and program variables (in fact any instance attribute of type int, float, bool or string).

This mechanism uses the popular .ini file format to save and reload persistent attributes.

Imagine renaming, adding or deleting widgets in Glade: an .ini file lying around from a previous program version, or an entirely different user interface, would be not be able to restore the state properly since variables and types might have changed.

GladeVCP detects this situation by a signature which depends on all object names and types which are saved and to be restored. In the case of signature mismatch, a new .ini file with default settings is generated.

If you want any of Gtk widget state, HAL widgets output pin’s values and/or class attributes of your handler class to be retained across invocations, proceed as follows:

Then associate an .ini file with this descriptor:

After restore_state(), self will have attributes set if as running the following:

Note that types are saved and preserved on restore. This example assumes that the ini file didn’t exist or had the default values from self.defaults.

After this incantation, you can use the following IniFil methods:

To save the widget and/or variable state on exit, proceed as follows:

This will save state and shutdown GladeVCP properly, regardless whether the panel is embedded in Axis, or a standalone window.

Next time you start the GladeVCP application, the widgets should come up in the state when the application was closed.

By default, the reaction of GladeVCP to a Ctrl-C event is to just exit - without saving state. To make sure that this case is covered, add a handler call on_unix_signal which will be automatically be called on Ctrl-C (actuall on the SIGINT and SIGTERM signals). Example

You can do that, but note that the values in self.defaults override your edits if there is a syntax or type error in your edit. The error is detected, a console message will hint about that happened, and the bad inifile will be renamed to have the .BAD suffix. Subsequent bad ini files overwrite earlier .BAD files.

If you need HAL pins which are not associated with a specific HAL widget, add them as follows:

To get a callback when this pin’s value changes, associate a value-change callback with this pin, add:

and define a callback method (or function, in this case leave out the self parameter):

Since GladeVCP uses Gtk widgets which rely on the GObject base class, the full glib functionally is available. Here is an example for a timer callback:

With glade, widget properties are typically set fixed while editing. You can, however, set widget properties at runtime, for instance from ini file values, which would typically be done in the handler initialisation code. Setting properties from HAL pin values is possible, too.

In the following example (assuming a HAL Meter widget called meter), the meter’s min value is set from an INI file parameter at startup, and the max value is set via a HAL pin, which causese the widget’s scale to readjust dynamically:

Visit linuxcnc/configs/gladevcp for running examples and starters for your own projects.

We believe key handling works OK, but since it is new code, we’re telling about it you so you can watch out for problems; please let us know of errors or odd behavior. This is the story:

Axis uses the TkInter widget set. GladeVCP applications use Gtk widgets and run in a separate process context. They are hooked into Axis with the Xembed protocol. This allows a child application like GladeVCP to properly fit in a parent’s window, and - in theory - have integrated event handling.

However, this assumes that both parent and child application properly support the Xembed protocol, which Gtk does, but TkInter doesn’t. A consequence of this is that certain keys would not be forwarded from a GladeVCP panel to Axis properly under all circumstances. One of these situations was the case when an Entry, or SpinButton widget had focus: in this case, for instance an Escape key would not have been forwarded to Axis and cause an abort as it should, with potentially disastrous consequences.

Therefore, key events in GladeVCP are explicitly handled, and selectively forwarded to Axis, to assure that such situations cannot arise. For details, see the keyboard_forward() function in lib/python/gladevcp/xembed.py.

The LinuxCNC Wiki has information on adding custom widgets to GladeVCP. GladeVCP Custom Widgets