commented out call to QApplication.ProcessEvents() in xcsw_is_visible()

[RussBerg]

This is a problem I have had for quite a while and only now getting around to submitting an issue and PR for it.
There are two ways I tested to fix this and both worked and neither caused the plot to behave in a way that did not make sense,
first is to comment this line

https://github.com/PierreRaybaut/guiqwt/blob/56ae061e128bcf95e489c347a475ea83dc22fb5a/guiqwt/plot.py#L1251

and the second was to instead comment out the call to ProcessEvents() in xcsw_is_visible,

https://github.com/PierreRaybaut/guiqwt/blob/56ae061e128bcf95e489c347a475ea83dc22fb5a/guiqwt/plot.py#L1304

I chose the later because it is not ideal to call ProcessEvents directly and it didn't cause the plot to behave strangely.

closes #107

[kalvdans]

Suggested change

#QApplication.processEvents()

Just remove the line instead of commenting it.

[RussBerg]

ok thanks, Truth be told I have been running for months with the first proposed change but opted for what I thought would be the less intrusive option 2 of commenting line 1304 but I just ran into a segfault, unless someone has a better suggestion I am am going to revert back to removing the line that connects SIG_VISIBILITY_CHANGED to the x cross section for this PR

[PierreRaybaut]

Please test the following:

• Comment those lines: https://github.com/PierreRaybaut/guiqwt/blob/56ae061e128bcf95e489c347a475ea83dc22fb5a/guiqwt/plot.py#L826-L827
• And also comment this line: https://github.com/PierreRaybaut/guiqwt/blob/56ae061e128bcf95e489c347a475ea83dc22fb5a/guiqwt/plot.py#L831
• But please leave this line uncommented: https://github.com/PierreRaybaut/guiqwt/blob/56ae061e128bcf95e489c347a475ea83dc22fb5a/guiqwt/plot.py#L789

I think that it should avoid all the segmentation faults you are experiencing.
Please tell me if it does.

[PierreRaybaut]

The next step would be the following.
And I think you may directly jump to this solution and ignore my previous comment.

Import QTimer here, adding it next to Qt: https://github.com/PierreRaybaut/guiqwt/blob/56ae061e128bcf95e489c347a475ea83dc22fb5a/guiqwt/plot.py#L120

Replace QApplication.processEvents() by QTimer.singleShot(0, QApplication.processEvents):

• Here: https://github.com/PierreRaybaut/guiqwt/blob/56ae061e128bcf95e489c347a475ea83dc22fb5a/guiqwt/plot.py#L827
• And there: https://github.com/PierreRaybaut/guiqwt/blob/56ae061e128bcf95e489c347a475ea83dc22fb5a/guiqwt/plot.py#L831

[RussBerg]

as per your suggestion I tested replacing the direct calls to processEvents() with
QTimer.singleShot(0, QApplication.processEvents) and it failed 3/10 times, I also tried
QTimer.singleShot(10, QApplication.processEvents) and this also failed 3/10 times, I then tried
QTimer.singleShot(100, QApplication.processEvents) and that had 0/10 failures,
left at 100ms I don't think that would really affect the performance of cross section updates right?

If we leave this open for the rest of the week I can keep testing and see if 100 is truly the magic delay

[RussBerg]

nope the problem remains if using the QTimer solution, the only thing that cures it is to comment out
https://github.com/PierreRaybaut/guiqwt/blob/56ae061e128bcf95e489c347a475ea83dc22fb5a/guiqwt/plot.py#L1251

It seems to not cause any problems as far as I can tell as everything behaves as I would expect.