No blocking on the main thread

[srcejon]

No blocking on the main thread

[srcejon]

DSPAcquisitionInit can hang as well, the second time time an RTL SDR is started. Similar workaround applies.

And forgot to mention, there needs to be coded added to DSPDeviceSourceEngine::handleInputMessages() to handle those messages.

(RTL SDR driver crash on stop has been worked around by using rtlsdr_read_sync instead of rtlsdr_read_async)

[srcejon]

So, it appears there are cases where we do need to wait some of these DSPDeviceSourceEngine commands to complete.

One such case is in void MainWindow::removeDeviceSet. It starts by calling deviceEngine->stopAcquistion(); which results in RTLSDRInput::stop being called. It later on calls deleteSampleSourcePluginInstanceInput, which calls ~RTLSDRInput which also calls RTLSDRInput::stop(). In this instance, it's on the main thread, so it will hang if the previous stop hasn't completed.

So, we need a way for DSPDeviceSourceEngine and friends to indicate when they have completed, but without blocking. As an experiment, I've tried adding an acquistionStopped signal to DSPDeviceSourceEngine. MainWindow::removeDeviceSet then needs to be broken up in to two parts: First calling stopAcquistion - then when it gets the acquistionStopped signal, it can continue the rest of the device set deletion. Seems to work.

[srcejon]

It does get a little trickier though. In mainwindow.cpp, there are several places where all device sets are removed as part of some other sequential code, such as:

// Do X...
// Remove device sets
while (m_deviceUIs.size() > 0) {
    removeLastDeviceSet();
}
// Do Y...

This needs to be broken up in to separate steps, with a non-blocking wait for a signal after each device set is removed, before trying to remove the next.

Trying to think of the cleanest way to write this code, I came across Qt's State Machine API: https://doc.qt.io/qt-5/statemachine-api.html - which seems like a reasonable way to sequence different bits of code using signals. I will give it a try.

Any thoughts on any of this?

[f4exb]

I only had a quick look at the Qt state machine and I find it quite interesting. When the events and event handlers start to be rather complex it probably helps keep things tidy. State Machine is a well known design pattern.