More Like this

1. iDFlakies: A Framework for Detecting and Partially Classifying Flaky Tests

   https://doi.org/10.1109/ICST.2019.00038  

   Lam, Wing ; Oei, Reed ; Shi, August ; Marinov, Darko ; Xie, Tao ( April 2019 , Proc. of the 12th IEEE International Conference on Software Testing, Verification and Validation (ICST 2019))

   Regression testing is increasingly important with the wide use of continuous integration. A desirable requirement for regression testing is that a test failure reliably indicates a problem in the code under test and not a false alarm from the test code or the testing infrastructure. However, some test failures are unreliable, stemming from flaky tests that can non- deterministically pass or fail for the same code under test. There are many types of flaky tests, with order-dependent tests being a prominent type. To help advance research on flaky tests, we present (1) a framework, iDFlakies, to detect and partially classify flaky tests; (2) a dataset of flaky tests in open-source projects; and (3) a study with our dataset. iDFlakies automates experimentation with our tool for Maven-based Java projects. Using iDFlakies, we build a dataset of 422 flaky tests, with 50.5% order-dependent and 49.5% not. Our study of these flaky tests finds the prevalence of two types of flaky tests, probability of a test-suite run to have at least one failure due to flaky tests, and how different test reorderings affect the number of detected flaky tests. We envision that our work can spur research to alleviate the problem of flakymore »tests.« less
2. Mitigating the effects of flaky tests on mutation testing

   https://doi.org/10.1145/3293882.3330568  

   Shi, August ; Bell, Jonathan ; Marinov, Darko ( July 2019 , ISSTA 2019 Proceedings of the 28th ACM SIGSOFT International Symposium on Software Testing and Analysis)

   Mutation testing is widely used in research as a metric for evaluating the quality of test suites. Mutation testing runs the test suite on generated mutants (variants of the code under test), where a test suite kills a mutant if any of the tests fail when run on the mutant. Mutation testing implicitly assumes that tests exhibit deterministic behavior, in terms of their coverage and the outcome of a test (not) killing a certain mutant. Such an assumption does not hold in the presence of flaky tests, whose outcomes can non-deterministically differ even when run on the same code under test. Without reliable test outcomes, mutation testing can result in unreliable results, e.g., in our experiments, mutation scores vary by four percentage points on average between repeated executions, and 9% of mutant-test pairs have an unknown status. Many modern software projects suffer from flaky tests. We propose techniques that manage flakiness throughout the mutation testing process, largely based on strategically re-running tests. We implement our techniques by modifying the open-source mutation testing tool, PIT. Our evaluation on 30 projects shows that our techniques reduce the number of "unknown" (flaky) mutants by 79.4%.
3. Compiling Volatile Correctly in Java

   Liu, Shuyang ; Bender, John ; Palsberg, Jens ( July 2022 , 36th European Conference on Object-Oriented Programming (ECOOP 2022))

   The compilation scheme for Volatile accesses in the OpenJDK 9 HotSpot Java Virtual Machine has a major problem that persists despite a recent bug report and a long discussion. One of the suggested fixes is to let Java compile Volatile accesses in the same way as C/C++11. However, we show that this approach is invalid for Java. Indeed, we show a set of optimizations that is valid for C/C++11 but invalid for Java, while the compilation scheme is similar. We prove the correctness of the compilation scheme to Power and x86 and a suite of valid optimizations in Java. Our proofs are based on a language model that we validate by proving key properties such as the DRF-SC theorem and by running litmus tests via our implementation of Java in Herd7.
4. Wait Wait. No, Tell Me. Analyzing Selenium Configuration Effects on Test Flakiness

   https://doi.org/10.1109/AST.2019.000-1  

   Presler-Marshall, Kai ; Horton, Eric ; Heckman, Sarah ; Stolee, Kathryn T. ( January 2019 , Proceedings of the 14th International Workshop on Automation of Software Test)

   Flaky tests are a source of frustration and uncertainty for developers. In an educational environment, flaky tests can create doubts related to software behavior and student grades, especially when the grades depend on tests passing. NC State University's junior-level software engineering course models industrial practice through team-based development and testing of new features on a large electronic health record (EHR) system, iTrust2. Students are expected to maintain and supplement an extensive suite of UI tests using Selenium WebDriver. Team builds are run on the course's continuous integration (CI) infrastructure. Students report, and we confirm, that tests that pass on one build will inexplicably fail on the next, impacting productivity and confidence in code quality and the CI system. The goal of this work is to find and fix the sources of flaky tests in iTrust2. We analyze configurations of Selenium using different underlying web browsers and timeout strategies (waits) for both test stability and runtime performance. We also consider underlying hardware and operating systems. Our results show that HtmlUnit with Thread waits provides the lowest number of test failures and best runtime on poor-performing hardware. When given more resources (e.g., more memory and a faster CPU), Google Chrome with Angularmore »waits is less flaky and faster than HtmlUnit, especially if the browser instance is not restarted between tests. The outcomes of this research are a more stable and substantially faster teaching application and a recommendation on how to configure Selenium for applications similar to iTrust2 that run in a CI environment.« less
5. Issues in the Reproducibility of Deep Learning Results

   https://doi.org/10.1109/SPMB47826.2019.9037840  

   Jean-Paul, S. ; Elseify, T. ; Obeid, I. ; Picone, J. ( December 2019 , IEEE Signal Processing in Medicine and Biology Symposium (SPMB))

   Obeid, I. ; Selesnik, I. ; Picone, J. (Ed.)

   The Neuronix high-performance computing cluster allows us to conduct extensive machine learning experiments on big data [1]. This heterogeneous cluster uses innovative scheduling technology, Slurm [2], that manages a network of CPUs and graphics processing units (GPUs). The GPU farm consists of a variety of processors ranging from low-end consumer grade devices such as the Nvidia GTX 970 to higher-end devices such as the GeForce RTX 2080. These GPUs are essential to our research since they allow extremely compute-intensive deep learning tasks to be executed on massive data resources such as the TUH EEG Corpus [2]. We use TensorFlow [3] as the core machine learning library for our deep learning systems, and routinely employ multiple GPUs to accelerate the training process. Reproducible results are essential to machine learning research. Reproducibility in this context means the ability to replicate an existing experiment – performance metrics such as error rates should be identical and floating-point calculations should match closely. Three examples of ways we typically expect an experiment to be replicable are: (1) The same job run on the same processor should produce the same results each time it is run. (2) A job run on a CPU and GPU should producemore »identical results. (3) A job should produce comparable results if the data is presented in a different order. System optimization requires an ability to directly compare error rates for algorithms evaluated under comparable operating conditions. However, it is a difficult task to exactly reproduce the results for large, complex deep learning systems that often require more than a trillion calculations per experiment [5]. This is a fairly well-known issue and one we will explore in this poster. Researchers must be able to replicate results on a specific data set to establish the integrity of an implementation. They can then use that implementation as a baseline for comparison purposes. A lack of reproducibility makes it very difficult to debug algorithms and validate changes to the system. Equally important, since many results in deep learning research are dependent on the order in which the system is exposed to the data, the specific processors used, and even the order in which those processors are accessed, it becomes a challenging problem to compare two algorithms since each system must be individually optimized for a specific data set or processor. This is extremely time-consuming for algorithm research in which a single run often taxes a computing environment to its limits. Well-known techniques such as cross-validation [5,6] can be used to mitigate these effects, but this is also computationally expensive. These issues are further compounded by the fact that most deep learning algorithms are susceptible to the way computational noise propagates through the system. GPUs are particularly notorious for this because, in a clustered environment, it becomes more difficult to control which processors are used at various points in time. Another equally frustrating issue is that upgrades to the deep learning package, such as the transition from TensorFlow v1.9 to v1.13, can also result in large fluctuations in error rates when re-running the same experiment. Since TensorFlow is constantly updating functions to support GPU use, maintaining an historical archive of experimental results that can be used to calibrate algorithm research is quite a challenge. This makes it very difficult to optimize the system or select the best configurations. The overall impact of all of these issues described above is significant as error rates can fluctuate by as much as 25% due to these types of computational issues. Cross-validation is one technique used to mitigate this, but that is expensive since you need to do multiple runs over the data, which further taxes a computing infrastructure already running at max capacity. GPUs are preferred when training a large network since these systems train at least two orders of magnitude faster than CPUs [7]. Large-scale experiments are simply not feasible without using GPUs. However, there is a tradeoff to gain this performance. Since all our GPUs use the NVIDIA CUDA® Deep Neural Network library (cuDNN) [8], a GPU-accelerated library of primitives for deep neural networks, it adds an element of randomness into the experiment. When a GPU is used to train a network in TensorFlow, it automatically searches for a cuDNN implementation. NVIDIA’s cuDNN implementation provides algorithms that increase the performance and help the model train quicker, but they are non-deterministic algorithms [9,10]. Since our networks have many complex layers, there is no easy way to avoid this randomness. Instead of comparing each epoch, we compare the average performance of the experiment because it gives us a hint of how our model is performing per experiment, and if the changes we make are efficient. In this poster, we will discuss a variety of issues related to reproducibility and introduce ways we mitigate these effects. For example, TensorFlow uses a random number generator (RNG) which is not seeded by default. TensorFlow determines the initialization point and how certain functions execute using the RNG. The solution for this is seeding all the necessary components before training the model. This forces TensorFlow to use the same initialization point and sets how certain layers work (e.g., dropout layers). However, seeding all the RNGs will not guarantee a controlled experiment. Other variables can affect the outcome of the experiment such as training using GPUs, allowing multi-threading on CPUs, using certain layers, etc. To mitigate our problems with reproducibility, we first make sure that the data is processed in the same order during training. Therefore, we save the data from the last experiment and to make sure the newer experiment follows the same order. If we allow the data to be shuffled, it can affect the performance due to how the model was exposed to the data. We also specify the float data type to be 32-bit since Python defaults to 64-bit. We try to avoid using 64-bit precision because the numbers produced by a GPU can vary significantly depending on the GPU architecture [11-13]. Controlling precision somewhat reduces differences due to computational noise even though technically it increases the amount of computational noise. We are currently developing more advanced techniques for preserving the efficiency of our training process while also maintaining the ability to reproduce models. In our poster presentation we will demonstrate these issues using some novel visualization tools, present several examples of the extent to which these issues influence research results on electroencephalography (EEG) and digital pathology experiments and introduce new ways to manage such computational issues.« less