Testing concurrent algorithms with Lincheck

Everybody knows that concurrent programming is bug-prone. Moreover, some bugs in complicated algorithms occur rarely and are hard to reproduce; thus, it is difficult to detect them via simple hand-written tests. In this talk, we discuss Lincheck tool for testing and debugging concurrent code. We will talk about both the capabilities and the API of the tool, and implementation details.

Transcript

1. Testing concurrent algorithms with lincheck Nikita Koval, Joker 2019
2. Writing concurrent code is pain 2
3. Writing concurrent code is pain 3 … testing it is not much easier!
4. 4 var i = 0 i.inc() i.inc()
5. 5 var i = 0 i.inc() // 0 // 1 i.inc() // 1 // 0
6. 6 var i = 0 i.inc() // 0 i.inc() // 0
7. 7 We do not expect this! var i = 0 i.inc() // 0 i.inc() // 0
8. Execution is linearizable ⇔ ∃ equivalent sequential execution wrt happens-before order (a bit more complicated) 8
9. 9 val q = ConcurrentQueue<Int>() q.add(1) q.poll(): 2 q.poll(): 1 q.add(2) Execution is linearizable ⇔ ∃ equivalent sequential execution wrt happens-before order (a bit more complicated)
10. val q = ConcurrentQueue<Int>() q.add(1) q.poll(): 2 q.poll(): 1 q.add(2) 10 Execution is linearizable ⇔ ∃ equivalent sequential execution wrt happens-before order (a bit more complicated)
11. 11 var i = 0 i.inc() // 0 i.inc() // 0 This counter is not linearizable
12. How to check whether my data structure is linearizable? 12
13. How to check whether my data structure is linearizable? 13 Formal proofs
14. How to check whether my data structure is linearizable? 14 Formal proofs Model checking
15. How to check whether my data structure is linearizable? 15 Formal proofs Model checking Testing
16. How to check whether my data structure is linearizable? 16 Formal proofs Model checking Testing
17. How does the ideal test look?
18. How does the ideal test look? 18 class ConcurrentQueueTest { val q = ConcurrentQueue<Int>() } Initial state
19. How does the ideal test look? 19 class ConcurrentQueueTest { val q = ConcurrentQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() } Operations on the data structure
20. How does the ideal test look? 20 class ConcurrentQueueTest { val q = ConcurrentQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() } Operation parameters can be non-fixed!
21. How does the ideal test look? 21 class ConcurrentQueueTest { val q = ConcurrentQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() @Test fun runTest() = LinChecker.check(this::class) } JUnit
22. How does the ideal test look? 22 class ConcurrentQueueTest { val q = ConcurrentQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() @Test fun runTest() = LinChecker.check(this::class) } The Magic Button
23. Lincheck Overview Lincheck = Linearizability Checker (supports not only linearizability) https://github.com/Kotlin/kotlinx-lincheck 23
24. Lincheck = Linearizability Checker (supports not only linearizability) https://github.com/Kotlin/kotlinx-lincheck 1. Generates a random scenario 2. Executes it a lot of times 3. Verifies the results 24 Lincheck Overview
25. Invalid Execution Example 25 Init part: [poll(): null, add(9)] Parallel part: | poll(): null | add(4) | | add(3) | add(6) | | poll(): 4 | poll(): 3 | Post part: [add(1)]
26. Invalid Execution Example 26 Init part: [poll(): null, add(9)] Parallel part: | poll(): null | add(4) | | add(3) | add(6) | | poll(): 4 | poll(): 3 | Post part: [add(1)] How to understand the error cause?
27. 27 Failed Scenario Minimization Init part: [poll(): null, add(9)] Parallel part: | poll(): null | add(4) | | add(3) | add(6) | | poll(): 4 | poll(): 3 | Post part: [add(1), poll(): 6] Init part: [add(9)] Parallel part: | poll(): null | add(4) | Lincheck tries to remove actors iteratively see Options.minimizeFailedScenario(..)
28. 28 How to generate scenarios?
29. 29 Scenario Configuration class MySuperFastQueueTest { val q = MySuperFastQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() }
30. 30 class MySuperFastQueueTest { val q = MySuperFastQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() } Scenario Configuration Init part: [poll(), add(9)] Parallel part: | poll() | add(4) | | add(3) | add(6) | | poll() | poll() | Post part: [add(1)]
31. 31 Scenario Configuration Init part: [poll(), add(9)] Parallel part: | poll() | add(4) | | add(3) | add(6) | | poll() | poll() | Post part: [add(1)] @StressCTest(actorsBefore = 2, threads = 2, actorsPerThread = 3, actorsAfter = 1) class MySuperFastQueueTest { val q = MySuperFastQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() }
32. 32 Scenario Configuration Init part: [poll(), add(9)] Parallel part: | poll() | add(4) | | add(3) | add(6) | | poll() | poll() | Post part: [add(1)] @StressCTest(actorsBefore = 2, threads = 2, actorsPerThread = 3, actorsAfter = 1) class MySuperFastQueueTest { val q = MySuperFastQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() @Test fun test() = LinChecker.check(this::class) }
33. 33 Scenario Configuration Init part: [poll(), add(9)] Parallel part: | poll() | add(4) | | add(3) | add(6) | | poll() | poll() | Post part: [add(1)] class MySuperFastQueueTest { val q = MySuperFastQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() @Test fun test() = StressOptions() .actorsBefore(2) .threads(2).actorsPerThread(3) .actorsAfter(1) .check(this::class) }
34. 34 Parameters Generation class MySuperFastQueueTest { val q = MySuperFastQueue<Int>() @Operation fun add(@Param(gen = IntGen::class, conf = "-10:10") x: Int) = q.add(x) @Operation fun poll() = q.poll() @Test fun test() = ... } We use parameter generators!
35. 35 Parameters Generation class MySuperFastQueueTest { val q = MySuperFastQueue<Int>() @Operation fun add(@Param(gen = IntGen::class, conf = "-10:10") x: Int) = q.add(x) @Operation fun addIfEmpty(@Param(gen = IntGen::class, conf = "-10:10") x: Int) = q.addIfEmpty(x) @Operation fun poll() = q.poll() @Test fun test() = ... } Let’s add one more add-like method
36. 36 Parameters Generation @Param(name = "elem", gen = IntGen::class, conf = "-10:10") class MySuperFastQueueTest { val q = MySuperFastQueue<Int>() @Operation fun add(@Param(name="elem") x: Int) = q.add(x) @Operation fun addIfEmpty(@Param(name="elem") x: Int) = q.addIfEmpty(x) @Operation fun poll() = q.poll() @Test fun test() = ... } We can share the configuration!
37. 37 Custom Parameter Generators class RandomIntParameterGenerator(ignoredConf: String) : ParameterGenerator<Int> { override fun generate() = Random.nextInt() } It is very simple to write your own ones!
38. 38 Custom Parameter Generators class RandomIntParameterGenerator(ignoredConf: String) : ParameterGenerator<Int> { override fun generate() = Random.nextInt() } It is very simple to write your own ones! Be careful, the running code can be loaded by another ClassLoader!
39. 39 Constraints class MySuperFastQueueTest { val q = TaskQueue<Int>() @Operation fun add(x: Int) = q.addIfNotClosed(x) @Operation fun poll() = q.poll() @Operation fun close() = q.close() @Test fun test() = ... }
40. 40 Constraints class MySuperFastQueueTest { val q = TaskQueue<Int>() @Operation fun add(x: Int) = q.addIfNotClosed(x) @Operation fun poll() = q.poll() @Operation fun close() = q.close() @Test fun test() = ... } What if we can invoke “close” only once by the queue contract?
41. 41 Constraints class MySuperFastQueueTest { val q = TaskQueue<Int>() @Operation fun add(x: Int) = q.addIfNotClosed(x) @Operation fun poll() = q.poll() @Operation(runOnce = true) fun close() = q.close() @Test fun test() = ... } What if we can invoke “close” only once by the queue contract?
42. 42 Constraints class MySuperFastQueueTest { val q = SingleConsumerTaskQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() @Test fun test() = ... }
43. 43 Constraints class MySuperFastQueueTest { val q = SingleConsumerTaskQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() @Test fun test() = ... } Parallel part: | add(2) | add(4) | | poll(): 2 | poll(): null | SC queue with two concurrent consumers is incorrect, what a surprise!
44. 44 Constraints @OpGroupConfig(name = "consumers", nonParallel = true) class MySuperFastQueueTest { val q = SingleConsumerTaskQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation(group = "consumers") fun poll() = q.poll() @Test fun test() = ... }
45. 45 Constraints @OpGroupConfig(name = "consumers", nonParallel = true) class MySuperFastQueueTest { val q = SingleConsumerTaskQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation(group = "consumers") fun poll() = q.poll() @Operation(group = "consumers") fun poll(timeout: Long) = ... @Test fun test() = ... }
46. 46 Number of Scenarios to Generate @StressCTest(iterations = 100500) class MySuperFastQueueTest { ... @Test fun test() = LinChecker.check(this::class) } class MySuperFastQueueTest { ... @Test fun test() = StressOptions() .iterations(100500) .check(this::class) }
47. 47 Custom Scenarios val s = scenario { initial { actor(MyQueueTest::add, 1) } parallel { thread { actor(MyQueueTest::add, 2) actor(MyQueueTest::add, 3) } thread { actor(MyQueueTest::poll) actor(MyQueueTest::poll) } } } github.com/Kotlin/kotlinx-lincheck/pull/8
48. 48 Custom Scenarios class MyQueueTest { ... @Test fun test() = StressOptions() .addCustomScenario(s) .check(this::class) } val s = scenario { initial { actor(MyQueueTest::add, 1) } parallel { thread { actor(MyQueueTest::add, 2) actor(MyQueueTest::add, 3) } thread { actor(MyQueueTest::poll) actor(MyQueueTest::poll) } } } github.com/Kotlin/kotlinx-lincheck/pull/8
49. 49 Custom Scenarios class MyQueueTest { ... @Test fun test() = StressOptions() .addCustomScenario(s) .check(this::class) } val s = scenario { initial { actor(MyQueueTest::add, 1) } parallel { thread { actor(MyQueueTest::add, 2) actor(MyQueueTest::add, 3) } thread { actor(MyQueueTest::poll) actor(MyQueueTest::poll) } } } Be careful, the running code can be loaded by another ClassLoader! github.com/Kotlin/kotlinx-lincheck/pull/8
50. 50 How to run scenarios?
51. 51 Init part: [poll(), add(9)] Parallel part: | poll() | add(4) | | add(3) | add(6) | | poll() | poll() | Post part: [add(1)] Sequential parts
52. 52 Init part: [poll(), add(9)] Parallel part: | poll() | add(4) | | add(3) | add(6) | | poll() | poll() | Post part: [add(1)] How to run the parallel part?
53. Stress Testing 53 active synchronization poll() add(3) poll() add(4) add(6) poll() Thread 1 Thread 2
54. Stress Testing 54 active synchronization poll() add(3) poll() add(4) add(6) poll() Thread 1 Thread 2 Additional work
55. Stress Testing 55 class MySuperFastQueueTest { ... @Test fun test() = StressOptions() .invocationsPerIteration(100500) .check(this::class) } active synchronization poll() add(3) poll() add(4) add(6) poll() Thread 1 Thread 2
56. Model Checking 56 poll() add(3) poll() add(4) add(6) poll() Thread 1 Thread 2 • Sequential Consistency (no races) • Bounded by number of interleavings • Increases the number of context switches • Brute forces interleavings evenly github.com/Kotlin/kotlinx-lincheck/pull/5
57. Model Checking 57 class Counter { @Volatile private var value = 0 fun getAndInc(): Int { val cur = value // line 28 value = cur + 1 // line 29 return cur } fun get() = value } github.com/Kotlin/kotlinx-lincheck/pull/5
58. Model Checking 58 class Counter { @Volatile private var value = 0 fun getAndInc(): Int { val cur = value // line 28 value = cur + 1 // line 29 return cur } fun get() = value } github.com/Kotlin/kotlinx-lincheck/pull/5 class CounterTest : VerifierState() { private val c = Counter() @Operation fun getAndInc() = c.getAndInc() @Operation fun get() = c.get() @Test fun test() = ModelCheckingOptions() .check(this::class) }
59. Model Checking 59 java.lang.AssertionError: Invalid interleaving found: = Invalid execution results: = Parallel part: | getAndInc(): 0 | getAndInc(): 0 | Parallel part execution trace: | | getAndInc(): 0 | | | Counter.getAndInc(CounterTest.kt:28) | | | SWITCH | | getAndInc(): 0 | | | SWITCH | | | | Counter.getAndInc(CounterTest.kt:29) | | | RESULT: 0 | | | FINISH | github.com/Kotlin/kotlinx-lincheck/pull/5
60. 60 How to check results?
61. Results Verification 61 Simplest solution: 1. Generate all possible sequential histories and produce all possible results in advance 2. On each invocation: check whether the current results are among the generated ones
62. Results Verification Simplest solution: 1. Generate all possible sequential histories and produce all possible results in advance 2. On each invocation: check whether the current results are among the generated ones 62 2 threads x 15 operations ⇒ OutOfMemoryError
63. Results Verification Simplest solution: 1. Generate all possible sequential histories and produce all possible results in advance 2. On each invocation: check whether the current results are among the generated ones Smarter solution: State Machine (LTS) 63
64. 64 LTS-Based Verification val q = MSQueue<Int>() q.add(4) q.poll(): 9 q.poll(): 4 q.add(9)
65. 65 LTS-Based Verification 4 add(4) val q = MSQueue<Int>() q.add(4) q.poll(): 9 q.poll(): 4 q.add(9)
66. 66 LTS-Based Verification 4 add(4) poll(): 4 val q = MSQueue<Int>() q.add(4) q.poll(): 9 q.poll(): 4 q.add(9) Result is different
67. 67 LTS-Based Verification val q = MSQueue<Int>() q.add(4) q.poll(): 9 q.poll(): 4 q.add(9) 4 add(4) poll(): 4
68. 68 LTS-Based Verification 9 add(9) val q = MSQueue<Int>() q.add(4) q.poll(): 9 q.poll(): 4 q.add(9) 4 add(4) poll(): 4
69. 69 LTS-Based Verification poll(): 9 val q = MSQueue<Int>() q.add(4) q.poll(): 9 q.poll(): 4 q.add(9) 9 add(9) 4 add(4) poll(): 4
70. 70 LTS-Based Verification poll(): 9 val q = MSQueue<Int>() q.add(4) q.poll(): 9 q.poll(): 4 q.add(9) 9 add(9) 4 add(4) poll(): 4 A path is found ⇒ correct
71. Lazy LTS Creation 71 • We build LTS lazilly, like on the previous slides • We use sequential implementation
72. Lazy LTS Creation 72 • We build LTS lazilly, like on the previous slides • We use sequential implementation 4 add(4) poll(): 4
73. Lazy LTS Creation 73 • We build LTS lazilly, like on the previous slides • We use sequential implementation • Equivalence via equals/hashcode implementations 4 add(4) poll(): 4 class MyQueueTest: VerifierState() { val q = MSQueue<Int>() // Operations here override fun generateState() = elements(q) }
74. 74 Sequential Specification class MySuperFastQueueTest { val q = MySuperFastQueue<Int>() @Operation fun add(x: Int) = q.add(x) @Operation fun poll() = q.poll() @Test fun test() = StressOptions() .sequentialSpecification(SequentialQueue::class.java) .check(this::class) } class SequentialQueue : VerifierState() { val q = ArrayDeque<Int>() fun add(x: Int) { q.add(x) } fun poll() = q.poll() @Override fun generateState() = q }
75. 75 What if my data structure is blocking by design?
76. 76 val c = Channel<Int>() c.send(4) c.receive() // S + 4 send waits for receive and vice versa Rendezvous Channels
77. Client 1 val task = Task(...) tasks.send(task) 77 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } val tasks = Channel<Task>() Rendezvous Channels
78. Client 1 val task = Task(...) tasks.send(task) 78 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } val tasks = Channel<Task>() Have to wait for send 1 Rendezvous Channels
79. Client 1 val task = Task(...) tasks.send(task) 79 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } val tasks = Channel<Task>() 1 Rendezvous Channels
80. Client 1 val task = Task(...) tasks.send(task) 80 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } val tasks = Channel<Task>() 1 Rendezvous Channels
81. Client 1 val task = Task(...) tasks.send(task) 81 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } val tasks = Channel<Task>() Rendezvous! 1 2 Rendezvous Channels
82. Client 1 val task = Task(...) tasks.send(task) 82 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } 1 val tasks = Channel<Task>() 3 2 Rendezvous Channels
83. Client 1 val task = Task(...) tasks.send(task) 83 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } 1 val tasks = Channel<Task>() 3 2 4 Have to wait for receive Rendezvous Channels
84. Client 1 val task = Task(...) tasks.send(task) 84 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } 1 val tasks = Channel<Task>() 3 2 4 Rendezvous Channels
85. Client 1 val task = Task(...) tasks.send(task) 85 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } 1 val tasks = Channel<Task>() 3 2 4 5 Rendezvous! Rendezvous Channels
86. 86 val c = Channel<Int>() c.send(4) c.receive() // S + 4 Non-linearizable because of suspension
87. 87 val c = Channel<Int>() c.send(4) c.receive(): // S + 4 register as a waiter suspend return element
88. 88 val c = Channel<Int>() c.send(4) c.receive(): // S + 4 register as a waiter suspend return element Dual Data Structures* * “Nonblocking Concurrent Data Structures with Condition Synchronization” by Scherer, W.N. and Scott, M.L. request follow-up
89. 89 Rendezvous Channel Test Example class RendezvousChannelTest: LinCheckState() { val c = Channel() @Operation suspend fun send(x: Int) = c.send(x) @Operation suspend fun receive(): Int = c.receive() override fun generateState() = Unit }
90. 90 Rendezvous Channel Test Example class RendezvousChannelTest: LinCheckState() { val c = Channel() @Operation suspend fun send(x: Int) = c.send(x) @Operation suspend fun receive(): Int = c.receive() override fun generateState() = Unit } Why “Unit”?
91. 91 State Equivalence 1. List of suspended operations 2. Set of resumed operations 3. Externally observable state
92. 92 State Equivalence 1. List of suspended operations 2. Set of resumed operations 3. Externally observable state Specified via equals/hashcode Maintained by lincheck
93. 93 Rendezvous Channel Test Example class RendezvousChannelTest: LinCheckState() { val c = Channel() @Operation suspend fun send(x: Int) = c.send(x) @Operation suspend fun receive(): Int = c.receive() override fun generateState() = Unit } Why “Unit”? Suspended and resumed operations define the channel state
94. Client 1 val task = Task(...) tasks.send(task) 94 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } val tasks = Channel<Task>(capacity = 1) One element can be sent without suspension Buffered Channels
95. Client 1 val task = Task(...) tasks.send(task) 95 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } val tasks = Channel<Task>(capacity = 1) 1 Does not suspend! Buffered Channels
96. Client 1 val task = Task(...) tasks.send(task) 96 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } val tasks = Channel<Task>(capacity = 1) 1 The buffer is full, suspends 2 Buffered Channels
97. Client 1 val task = Task(...) tasks.send(task) 97 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } val tasks = Channel<Task>(capacity = 1) 1 Receives the buffered element, resumes the 2nd client, and moves its task to the buffer 3 2 Buffered Channels
98. Client 1 val task = Task(...) tasks.send(task) 98 Client 2 val task = Task(...) tasks.send(task) Worker while(true) { val task = tasks.receive() processTask(task) } val tasks = Channel<Task>(capacity = 1) 1 Retrieves the 2nd task, no waiters to resume 2 3 4 Buffered Channels
99. 99 Buffered Channel Test Example class BufferedChannelTest: LinCheckState() { val c = Channel() @Operation suspend fun send(x: Int) = c.send(x) @Operation suspend fun receive(): Int = c.receive() override fun generateState() = bufferedElements(c) }
100. 100 Buffered Channel Test Example class BufferedChannelTest: LinCheckState() { val c = Channel() @Operation suspend fun send(x: Int) = c.send(x) @Operation suspend fun receive(): Int = c.receive() override fun generateState() = bufferedElements(c) } Externally observable state = buffered elements + waiting senders elements (optionally)
101. 101 https://www.youtube.com/watch?v=hwbpUEGHvvY
102. 102 Are all “correct” data structures linearizable?
103. 103 Sequential consistency Quiescent consistency Quasi-linearizability Quantitative relaxation Local linearizability
104. 104 Sequential consistency Quiescent consistency Quasi-linearizability Quantitative relaxation Local linearizability We actually test for it
105. 105 Sequential consistency Quiescent consistency Quasi-linearizability Quantitative relaxation Local linearizability We support this formalism, and use it in Kotlin Coroutines* * https://github.com/Kotlin/kotlinx.coroutines/blob/1.3.2/kotlinx-coroutines-core/common/src/internal/LockFreeTaskQueue.kt
106. 106 Sequential consistency Quiescent consistency Quasi-linearizability Quantitative relaxation Local linearizability Decided to remove these contracts from lincheck* * Got best decision award :)
107. 107 Sequential consistency Quiescent consistency Quasi-linearizability Quantitative relaxation Local linearizability Not supported Never have Never will
108. 108 Summary
109. • It is easy to check concurrent data structures with lincheck • We support various popular contracts ◦ single reader/writer, dual data structures ◦ serializability, quiescent consistency • We use lincheck in Kotlin Coroutines to test our algorithms and student assignments 109
111. Hydra Conference hydraconf.com Summer Schools in SPb (2017 & 2019) neerc.ifmo.ru/sptcc sptdc.ru Homework Assignments @ITMO (Koval & Elizarov) github.com/ITMO-MPP 111 Useful Materials
112. PPoPP. Principles and Practice of Parallel Programming PODC. Symposium on Principles of Distributed Computing SPAA. Symposium on Parallelism in Algorithms and Architectures Others: DISC, OPODIS, Euro-Par, IPDPS, PACT, ... 112 Main Research Conferences
113. Questions? https://github.com/Kotlin/kotlinx-lincheck
114. 114 Sequential model Concurrent model sequential specification on operations Linearizability (usually)
115. 115 Custom Scenario Generators class MyScenarioGenerator(testCfg: CTestConfiguration, testStr: CTestStructure) : ExecutionGenerator(testCfg, testStructure) { override fun nextExecution() = ExecutionScenario( emptyList(), // init part listOf( listOf( Actor(method = MyQueueTest::add.javaMethod!!, arguments = listOf(1), handledExceptions = emptyList()) ), listOf( Actor(method = MyQueueTest::poll.javaMethod!!, arguments = emptyList(), handledExceptions = emptyList()) ) ), emptyList() // post part ) }
116. 116 Custom Scenario Generators class MyScenarioGenerator(testCfg: CTestConfiguration, testStr: CTestStructure) : ExecutionGenerator(testCfg, testStructure) { override fun nextExecution() = … } class MyQueueTest { ... @Test fun test() = StressOptions() .executionGenerator(MyScenarioGenerator::class.java) .check(this::class) }
117. 117 Custom Scenario Generators class MyScenarioGenerator(testCfg: CTestConfiguration, testStr: CTestStructure) : ExecutionGenerator(testCfg, testStructure) { override fun nextExecution() = … } class MyQueueTest { ... @Test fun test() = StressOptions() .executionGenerator(MyScenarioGenerator::class.java) .check(this::class) } Be careful, the running code can be loaded by another ClassLoader!
118. 118 Can suspending operations be cancellable?
119. TODO: Cancellation support 119