Seg fault while selecting different rows in a table on macOS

[rmartin16]

Describe the bug

I noticed the table example would eventually crash on macOS with a seg fault while I would click different rows in the table. This would reliably happen in less than 30s. Of note, my macOS environment is virtualized...so, maybe this doesn't happen on Apple hardware.

macOS_error.log

Steps to reproduce

briefcase run in the table example and select one row after another in the table

Expected behavior

No seg fault

Screenshots

No response

Environment

• Operating System: Monterey
• Python version: 3.11.4
• Software versions:
  • Briefcase: 0.3.15.dev361+g02f412be.d20230620
  • Toga: toga_cocoa-0.3.2.dev388+g68ac31aa

Logs

briefcase.2023_06_20-12_31_47.run.log

Additional context

No response

[freakboy3742]

Thanks... I hate it... :-)

I haven't been able to reproduce this myself. To clarify - I can see this is the Table demo, running on a (virtualised) x86_64 Mac; can you clarify the sequence of events that causes the crash? Is it literally "click on row 1, click on row 2, rinse and repeat" for 30 seconds? Is this on the left or right version of the table?

[rmartin16]

Thanks... I hate it... :-)

🙃

Is it literally "click on row 1, click on row 2, rinse and repeat" for 30 seconds?

Yep; this works....although, usually doesn't take near 30s.

Is this on the left or right version of the table?

Either one....or just alternating clicks between them.

I'd make a video....but it's really just any sequence of clicking on the table leads to a crash.

[freakboy3742]

@rmartin16 Can I ask you to take #2011 for a spin? I still haven’t been able to reproduce this (including writing an automated test case that clicked thousands of times) - but I’ve also made a couple of cleanups to the Cocoa Table implementation that might have inadvertently cleaned up some memory issues.

[rmartin16]

Unfortunately, it's the same behavior with #2011. In fact, upon rebuilding and opening the table example, it crashed on the very first click in to the table.

[freakboy3742]

Hmph... well that's annoying... I can't reproduce this on either x86 or M1 Macs - definitely not as explosively as the first selection.

[freakboy3742]

@rmartin16 Thinking outside the box here - is there any possibility what you're seeing is actually #1156? I'm able to reliably cause that error, but simple clicking does nothing. I'm wondering if there's some quirk of your mouse setup (or your wrist action? ;-) ) that is triggering the start of a drag action.

One possible check for this - does briefcase run --test work for you on the testbed app?

If it does - If you add the following test case to testbed/tests/widgets/test_table.py, does it pass?

import random


async def test_click(widget, probe, source, on_select_handler):
    row = 1
    for i in range(0, 100):
        row = (row + 3) % 14
        await probe.select_row(row)
        await probe.redraw(f"{i}: row {row} is selected")
        on_select_handler.assert_called_once_with(widget)
        on_select_handler.reset_mock()

[rmartin16]

@rmartin16 Thinking outside the box here - is there any possibility what you're seeing is actually #1156? I'm able to reliably cause that error, but simple clicking does nothing. I'm wondering if there's some quirk of your mouse setup (or your wrist action? ;-) ) that is triggering the start of a drag action.

ahhh....I didn't know there was an existing issue for seg fault on "dragging" a row. This is likely the same issue then because I wasn't being careful not to move the mouse while I was clicking. This is also likely exacerbated by the fact there's no actual virtualization support for macOS; so, the mouse movement is choppy and delayed.

One possible check for this - does briefcase run --test work for you on the testbed app?

This passes.

If it does - If you add the following test case to testbed/tests/widgets/test_table.py, does it pass?

import random


async def test_click(widget, probe, source, on_select_handler):
    row = 1
    for i in range(0, 100):
        row = (row + 3) % 14
        await probe.select_row(row)
        await probe.redraw(f"{i}: row {row} is selected")
        on_select_handler.assert_called_once_with(widget)
        on_select_handler.reset_mock()

This test fails....although, because of a mock not being called.

[testbed] Starting test_suite...
===========================================================================
Configuring isolated Python...
Pre-initializing Python runtime...
PythonHome: /Users/user/github/beeware/toga/testbed/build/testbed/macos/app/Toga Testbed.app/Contents/Resources/support/python-stdlib
PYTHONPATH:
- /Users/user/github/beeware/toga/testbed/build/testbed/macos/app/Toga Testbed.app/Contents/Resources/support/python311.zip
- /Users/user/github/beeware/toga/testbed/build/testbed/macos/app/Toga Testbed.app/Contents/Resources/support/python-stdlib
- /Users/user/github/beeware/toga/testbed/build/testbed/macos/app/Toga Testbed.app/Contents/Resources/support/python-stdlib/lib-dynload
- /Users/user/github/beeware/toga/testbed/build/testbed/macos/app/Toga Testbed.app/Contents/Resources/app_packages
- /Users/user/github/beeware/toga/testbed/build/testbed/macos/app/Toga Testbed.app/Contents/Resources/app
Configure argc/argv...
Initializing Python runtime...
Installing Python NSLog handler...
Running app module: tests.testbed
---------------------------------------------------------------------------
============================= test session starts ==============================
collecting ... collected 259 items / 258 deselected / 1 selected

tests/widgets/test_table.py::test_click FAILED                           [100%]

=================================== FAILURES ===================================
__________________________________ test_click __________________________________
Traceback (most recent call last):
  File "/Users/user/github/beeware/toga/testbed/build/testbed/macos/app/Toga Testbed.app/Contents/Resources/app/tests/widgets/test_table.py", line 428, in test_click
    on_select_handler.assert_called_once_with(widget)
  File "/Users/user/github/beeware/toga/testbed/build/testbed/macos/app/Toga Testbed.app/Contents/Resources/support/python-stdlib/unittest/mock.py", line 934, in assert_called_once_with
    raise AssertionError(msg)
AssertionError: Expected 'mock' to be called once. Called 0 times.
=========================== short test summary info ============================
FAILED tests/widgets/test_table.py::test_click - AssertionError: Expected 'mo...
====================== 1 failed, 258 deselected in 3.67s =======================

[freakboy3742]

That mode of test failure isn't entirely unexpected - if it selects the same row twice (or there's a long lag in event processing), it could fail due to a mock not being triggered. If the failure is due to the mock, not due to a segfault, I think we've narrowed down the problem here. On that basis, I think I'll close this ticket in favor of #1156; however, I'm no closer to having any idea how to fix that one...