PyQt6 program randomly crashing without any error, falling into despair

[Tropaion]

Hello,

I developed my first bigger application with PyQt6 for my companies production line.
Basically, this software programs an mikrocontroller, tests if pcb components are working and logs it.

Now, the application works, except for some random crashes without any error logs.
The basic structure of the software is:
main thread -> GUI
main worker thread -> button handling, start/stop...
worker thread1 -> testing stuff
worker thread2 -> testing other stuff
The communication between all threads is with signals.
The worker threads class structures are basically testing functions which call themselves with an SingleShotTimer every 500ms.
I'm also logging every function call and exception.

Basically, one test iteration is a big state machine, which repeats itself.
Now interestingly, the testing itself works. But after a few random iterations the program just crashes. No log, not exception, nothing. It happens completely randomly, but always after the first test iteration or later ones. (basically between the 2nd and 5th iteration).

Now, for debugging, I tried trace and pdb, but same, no output.
The only output i got was from pycharm in debugging mode: "Process finished with exit code -1073741819 (0xC0000005)". But as far as I can understand, this doesn't really help.

I hope somone can help me, I have no idea how to debug further and haven't found anything usefull in the web, I'm getting desperate.

Thanks and sincerely,
Tropaion

[JonB]

@Tropaion
First things first:

main worker thread -> button handling, start/stop...

What does "button handling" mean? Assuming "main worker thread" means "a thread which is not the "main thread -> GUI" ", are you aware that no secondary thread is allowed to access any UI stuff, e.g. a button. So what exactly does the secondary thread do for "button handling"?

Second

The only output i got was from pycharm in debugging mode: "Process finished with exit code -1073741819 (0xC0000005)".

Can you run your Python program outside of PyCharm? E.g. from a command line python yourprogram.py? Does that crash? With same message?

[Tropaion]

@JonB Thanks for your reply

@JonB said in PyQt6 program randomly crashing without any error, falling into despair:

What does "button handling" mean? Assuming "main worker thread" means "a thread which is not the "main thread -> GUI" ", are you aware that no secondary thread is allowed to access any UI stuff, e.g. a button. So what exactly does the secondary thread do for "button handling"?

Yes, I know that secondary threads are not allowed to access UI stuff.
The "main worker thread" has many slot functions which do long lasting stuff.
But everything is handled via signals. So a button click sends a signal from the GUI to the main worker.

It basically looks like this:

# Function to initialize main worker thread
    def _init_main_worker(self) -> None:
        logging.debug("called")

        # Create worker object and thread
        self.main_worker = _main_worker.Worker(self.configuration)
        self.main_worker_thread = QThread()

        # Move worker to thread
        self.main_worker.moveToThread(self.main_worker_thread)

        # Stop handler
        self.frontend.main.close_event.connect(self.main_worker.stop)

        # ----- TESTER SELECTION DIALOG -----
        self.main_worker.enable_tester_selection_button.connect(self.frontend.main.bttn_change_pruefer.setEnabled)
        self.main_worker.set_tester_selection_label.connect(self.frontend.main.label_selected_pruefer.setText)
        self.main_worker.start_tester_selection_dialog.connect(self.frontend.tester_selection_dialog.exec)
        
        self.frontend.tester_selection_dialog.selected_tester.connect(self.main_worker.set_tester)

        # ----- DUT SELECTION DIALOG -----
        self.main_worker.enable_dut_selection_button.connect(self.frontend.main.bttn_change_dut.setEnabled)
        self.main_worker.periphery_status.connect(self.frontend.dut_selection_dialog.update_periphery_status)

        self.frontend.dut_selection_dialog.selected_dut_key.connect(self.main_worker.set_selected_dut)        

        # ----- TESTED DEVICE DIALOG -----
        self.main_worker.enable_tested_device_button.connect(self.frontend.main.bttn_tested_device.setEnabled)```

@JonB said in [PyQt6 program randomly crashing without any error, falling into despair](/post/814353):
> Can you run your Python program outside of PyCharm? E.g. from a command line python yourprogram.py? Does that crash? With same message?

I have tried to run it in many different ways.

* embedded x64 python -> no error output -> crashes faster
* installed x64 python in console -> no error output -> crashes a slower
* installed x64 python in console with trace -> no error output 
* installed x64 python in console with pdb -> no error output
* only pycharm did throw an error: "Process finished with exit code -1073741819 (0xC0000005)"

[Tropaion]

Sorry, rest of my answer is in the code section...and somehow I can't edit it because if akismet spam warning.
Here the second answer:
@JonB said in PyQt6 program randomly crashing without any error, falling into despair:

Can you run your Python program outside of PyCharm? E.g. from a command line python yourprogram.py? Does that crash? With same message?

I have tried to run it in many different ways.

• embedded x64 python -> no error output -> crashes faster
• installed x64 python in console -> no error output -> crashes a slower
• installed x64 python in console with trace -> no error output
• installed x64 python in console with pdb -> no error output
• only pycharm did throw an error: "Process finished with exit code -1073741819 (0xC0000005)"

[jsulm]

@Tropaion said in PyQt6 program randomly crashing without any error, falling into despair:

self.main_worker.enable_tester_selection_button

Is this button a UI button like QPushButton? If so what you are doing is not allowed as @JonB pointed out.
All UI stuff must be done in main thread.

[JonB]

@Tropaion
If you can get it to crash outside of PyCharm, do you have a C++ or system debugger available to you? If so you might run python yourprogram.py inside that and let it crash. Although this is not ideal you would get a stack trace of where it is in the Python --- but not in your code --- which may give a clue as to where the problem lies.

[Tropaion]

Hello @JonB, that would be a beginning.
Can you recommend me an C++/system debugger? Not sure which one I should install and is supporting python.
I heard MinGW works, but not sure about it on Windows. And I don't really know how to use it.

I now tried this one I found: https://marketplace.visualstudio.com/items?itemName=benjamin-simmonds.pythoncpp-debug
It did stop because of an exception but I'm not sure what to do here now:

[Tropaion]

Now, I enabled a few debugging modes I found and added many many prints and got this output:

EMITTTING DICT
EMITTED DICT
<SDM> [reset_safety_circuit scope] overwrite:  False
<SDM> [start_backend scope] self._cryptokey:  b'\xdb\xe3\x15\xef\xffL\xb0%s%\x1bC\xb9\xe4DH'
<SDM> [start_backend scope] _worker_params:  {'dut': '200101', 'mac_address': '986D35C0F889', 'serial_number': '2001012411120027', 'vendor_serial_number': '0000000000000000', 'cryptokey': b'\xdb\xe3\x15\xef\xffL\xb0%s%\x1bC\xb9\xe4DH', 'inital_test_index': 0}
<SDM> [_get_meter_measurements scope] measurements:  {}
Windows fatal exception: access violation

Thread 0x00002e1c (most recent call first):
  File "C:\Users\Pruefplatz\AppData\Local\Programs\Python\Python310\lib\site-packages\pymodbus\client\tcp.py", line 227 in recv
  File "C:\Users\Pruefplatz\AppData\Local\Programs\Python\Python310\lib\site-packages\pymodbus\transaction.py<SDM> [_get_meter_measurements scope] measurements[U_in]:  230.9
", line 417 in _recv
  File "C:\Users\Pruefplatz\AppData\Local\Programs\Python\Python310\lib\site-packages\pymodbus\transaction.py", line 325 in _transact
  File "C:\Users\Pruefplatz\AppData\Local\Programs\Python\Python310\lib\site-packages\pymodbus\transaction.py", line 215 in execute
  File "C:\Users\Pruefplatz\AppData\Local\Programs\Python\Python310\lib\site-packages\pymodbus\client\base.py", line 236 in execute
  File "C:\Users\Pruefplatz\AppData\Local\Programs\Python\Python310\lib\site-packages\pymodbus\client\mixin.py<SDM> [_get_meter_measurements scope] measurements[U_out]:  0.0
", line 92 in read_holding_registers
  File "C:\Users\Pruefplatz\AppData\Local\Programs\Python\Python310\lib\site-packages\\drivers\modbus.py", line 68 in read_register
  File "c:\Users\Pruefplatz\Documents\GitHub\teststation-V3\software\src\interface\_main_worker.py", line 1237 in _get_meter_measurements

Current thread 0x000037d0 (most recent call first):
  File "c:\Users\Pruefplatz\Documents\GitHub\teststation-V3\software\src\main.py", line 57 in main
  File "c:\Users\Pruefplatz\Documents\GitHub\teststation-V3\software\src\main.py", line 61 in <module>
<SDM> [_get_meter_measurements scope] measurements[I_in]:  0.02
<SDM> [_get_meter_measurements scope] measurements[P_in]:  0.0
<SDM> [_get_meter_measurements scope] measurements[P_in_9s]:  0.0
<SDM> [_get_meter_measurements scope] measurements[U_in_9s]:  0.0
<SDM> [_get_meter_measurements scope] measurements[I_in_9s]:  0.0
EMITTTING DICT
EMITTED DICT
<SDM> [_get_meter_measurements scope] measurements:  {}
<SDM> [_get_meter_measurements scope] measurements[U_in]:  230.8
<SDM> [_get_meter_measurements scope] measurements[U_out]:  0.0
<SDM> [_get_meter_measurements scope] measurements[I_in]:  0.02
<SDM> [_get_meter_measurements scope] measurements[P_in]:  0.0

[JonB]

@Tropaion
Does this indicate/suggest your problem lies in receiving data from some "measurement device", quite outside of anything Qt?

[Tropaion]

I don't think so, this function only gets some measurements from an device via the internet.
This function is called every 500ms by an QSingleShotTimer, gets the measurements and writes them in an dict.
The dict is then emitted to all other threads.

@pyqtSlot()
    def _get_meter_measurements(self):
        """! Function periodically reads measurements from meters and sends a signal."""
        try:
            # Create empty dict
            measurements = {}

            # Read acthor values
            measurements["U_in"] = self._meters.read_register(0, 2, 2)
            if measurements["U_in"] is False:
                measurements["U_in"] = "??"

            measurements["U_out"] = self._meters.read_register(7022, 2, 4, True)
            if measurements["U_out"] is False:
                measurements["U_out"] = "??"

            measurements["I_in"] = self._meters.read_register(6, 2, 2)
            if measurements["I_in"] is False:
                measurements["I_in"] = "??"

            measurements["P_in"] = self._meters.read_register(12, 2, 2)
            if measurements["P_in"] is False:
                measurements["P_in"] = "??"

            # Read 9s values
            measurements["P_in_9s"] = self._meters.read_register(12, 2, 1)
            if measurements["P_in_9s"] is False:
                measurements["P_in_9s"] = "??"

            measurements["U_in_9s"] = self._meters.read_register(7022, 2, 3, True)
            if measurements["U_in_9s"] is False:
                measurements["U_in_9s"] = "??"

            measurements["I_in_9s"] = self._meters.read_register(6, 2, 1)
            if measurements["I_in_9s"] is False:
                measurements["I_in_9s"] = "??"

            # Measuring done, emit signal
            self.update_meter_measurements.emit(measurements)
            
            # Check if meter measurement should be called again
            if self._meter_measurement_timer:
                QTimer.singleShot(self._config["general_periphery"]["modbus"]["interval"], self._get_meter_measurements)
        except Exception as e:
            logging.exception(e)

Truthfully, the logs are not very helpfull.
In the image I posted previously for example it says: "(Windows) Attach: Unhandled exception at 0x00007FF919D03A20 (Qt6Qui.dll) in python.exe:
0xC0000005: Access Violation reading location 0x0000000000000171"

[jsulm]

@Tropaion said in PyQt6 program randomly crashing without any error, falling into despair:

line 1237 in _get_meter_measurements

Which line in _get_meter_measurements is that?

[Tropaion]

@jsulm It's an empty line: