reschedule before AutoResetEvent.Reset

[aoli-al]

This fixes #495

public static async Task TestConcurrentAccountCreation()
{
    var threadReady = new AutoResetEvent(false);
    var autoResetEvent1 = new AutoResetEvent(false);

    var task = Task.Run(() =>
    {
        threadReady.WaitOne(); 
        autoResetEvent1.Set();
    });

    threadReady.Set();
    Thread.Yield();  // Without this forced interleave, coyote can not identify the deadlock.
    autoResetEvent1.Reset();
    autoResetEvent1.WaitOne();

    await Task.WhenAll(task);
}

Manually tested with the above code. Can confirm coyote successfully identifies the bug.

[akashlal]

Thanks -- can you add the corresponding test for deadlock as well? Best place for it is here:
https://github.com/microsoft/coyote/blob/main/Tests/Tests.BugFinding/Synchronization/AutoResetEventTests.cs

[aoli-al]

@microsoft-github-policy-service agree

[pdeligia]

This is great addition @aoli-al , thanks for adding the test too! Just kicked off CI to run on this

[pdeligia]

Could you also update the hashes in this test script: https://github.com/microsoft/coyote/blob/main/Tests/compare-rewriting-diff-logs.ps1

This is because the new added test changed the expected binary-rewriting hash of the Tests.BugFinding project, which Coyote CI validates to make sure binary-rewriting doesn't break unexpectedly. See:

Error: The 'Tests.BugFinding' project's IL diff hash '786F4C49E4C7361F9D91BA0DC61E05F33E3E04BB284ACF204BB9DBD73498AFEB' is not the expected '0112C27DC0C301BA980E598C484CFABA4F881F5E91FDD74107ADE0002A6901A4'.

in CI: https://github.com/microsoft/coyote/actions/runs/8113588744/job/22183054931?pr=496

[pdeligia]

If everything passes, I will merge this. It was great that you added this feature under the IsLockAccessRaceCheckingEnabled this was the right thing to do!

[aoli-al]

Is it expected that Coyote will generate different .diff.json on different platforms? I've tried to run compare-rewriting-diff-logs.ps1 on Ubuntu 22.04 and Windows 11 and got different results on different platforms.

[pdeligia]

Is it expected that Coyote will generate different .diff.json on different platforms? I've tried to run compare-rewriting-diff-logs.ps1 on Ubuntu 22.04 and Windows 11 and got different results on different platforms.

Great observation, yea exactly what you said happens. To simplify this we just run this specific check on windows only (${{ matrix.platform == 'windows-latest' }}).

[aoli-al]

2024-03-04T18:36:58.0613449Z   Failed Microsoft.Coyote.BugFinding.Tests.ThreadRunTests.TestThreadStartAndJoinStress [74 ms]
2024-03-04T18:36:58.0613628Z   Error Message:
2024-03-04T18:36:58.0614179Z    State of thread '312' is Background, Stopped instead of Stopped.
2024-03-04T18:36:58.0614195Z 
2024-03-04T18:36:58.0614337Z Expected: True
2024-03-04T18:36:58.0614479Z Actual:   False
2024-03-04T18:36:58.0615186Z    at Xunit.Assert.True(Nullable`1 condition, String userMessage) in /_/src/xunit.assert/Asserts/BooleanAsserts.cs:line 132
2024-03-04T18:36:58.0615702Z    at Xunit.Assert.True(Boolean condition, String userMessage) in /_/src/xunit.assert/Asserts/BooleanAsserts.cs:line 116
2024-03-04T18:36:58.0616908Z    at Microsoft.Coyote.Tests.Common.BaseTest.RunSystematicTest(Delegate test, Configuration configuration, Action`1 startIterationCallBack, Action`1 endIterationCallBack) in D:\a\coyote\coyote\Tests\Tests.Common\BaseTest.cs:line 114
2024-03-04T18:36:58.0617059Z Expected: False
2024-03-04T18:36:58.0617200Z Actual:   True
2024-03-04T18:36:58.0617324Z   Stack Trace:
2024-03-04T18:36:58.0618789Z      at Microsoft.Coyote.Tests.Common.BaseTest.RunSystematicTest(Delegate test, Configuration configuration, Action`1 startIterationCallBack, Action`1 endIterationCallBack) in D:\a\coyote\coyote\Tests\Tests.Common\BaseTest.cs:line 121
2024-03-04T18:36:58.0620595Z    at Microsoft.Coyote.Tests.Common.BaseTest.RunSystematicTest(Action test, Configuration configuration, Action`1 startIterationCallBack, Action`1 endIterationCallBack) in D:\a\coyote\coyote\Tests\Tests.Common\BaseTest.cs:line 95
2024-03-04T18:36:58.0622078Z    at Microsoft.Coyote.Tests.Common.BaseTest.Test(Action test, Configuration configuration) in D:\a\coyote\coyote\Tests\Tests.Common\BaseTest.cs:line 47
2024-03-04T18:36:58.0623393Z    at Microsoft.Coyote.BugFinding.Tests.ThreadRunTests.TestThreadStartAndJoinStress() in D:\a\coyote\coyote\Tests\Tests.BugFinding\Threads\ThreadRunTests.cs:line 66
2024-03-04T18:36:58.0624204Z    at System.RuntimeMethodHandle.InvokeMethod(Object target, Void** arguments, Signature sig, Boolean isConstructor)
2024-03-04T18:36:58.0624895Z    at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr)

Here is the log for the failing test. I'm not able to reproduce it on my machine. Is this test flaky?

[pdeligia]

Yea seems that test is somehow flaky. But orthogonal to your PR. Everything passed so I will approve and merge. Thanks for your contribution!