Bugs in code continue to pose a fundamental problem for software reliability and cause expensive failures. The process of removing known bugs is termed debugging, which is a classic methodology commonly performed before code is deployed. Traditionally, debugging is tedious, often requiring much manual effort. A more recent technique that complements debugging is data structure repair, which handles bugs that make it to deployed systems and lead to erroneous behavior at runtime by modifying erroneous program states to recover from errors. While data structure repair presents a promising basis for dealing with bugs at runtime, it remains computationally expensive. Our thesis is that debugging and data structure repair can be integrated to provide the basis of an effective approach for removing bugs before code is deployed and handling them after it is deployed. We present a bi-directional integration where ideas at the basis of data structure repair assist in automating debugging and vice versa. Our key insight is two-fold: (1)a repair action performed to mutate an erroneous object field value to repair it can be abstracted into a program statement that performs that update correctly; and (2)repair actions that are performed repeatedly to fix the same error can be memoized and re-used. We design, develop, and evaluate two techniques that embody our insight. One, we present an automated debugging technique that leverages a systematic constraint-based data structure repair technique developed in previous work and provides suggestions on how to fix a faulty program. Two, we present repair abstractions that are based on the same central ideas as in our automated debugging technique and memoize how an erroneous state was repaired, which enables prioritizing and re-using repair actions when the same error occurs again. The focus of our work is programs that operate on structurally complex data, e.g., heap-allocated data structures that have complex structural integrity constraints, such as acyclicity. Checking such constraints plays a central role in the techniques that lay at the foundation of our work. These techniques require the user to provide the constraints, which poses a burden on the user. To facilitate the use of constraint-based techniques, we present a third technique to check constraint violations at runtime using graph spectra, which have been studied extensively by mathematicians to capture properties of graphs. We view the heap of an object-oriented program as an edge-labeled graph, which allows us to apply results from graph spectra theory. Experimental results show the effectiveness of using graph spectra as a basis of capturing structural properties of a class of commonly used data structures. / text

Contracts and specifications have long been used in object-oriented design, programming and testing to enhance reliability before software deployment. However, the use of specifications in deployed software is commonly limited to runtime checking where assertions form a basis for detecting incorrect program states to terminate the erroneous executions. We present a contract-based approach for data structure repair, which repairs erroneous states in deployed software. The key novelty is the support for rich behavioral specifications, such as those that relate pre-states with post-states of a method to accurately specify expected behavior and precise repair. The approach is based on the view of a specification as a non-deterministic implementation. The key insight is to use any correct state mutations by an otherwise erroneous execution to prune non-determinism in the specification, thereby transmuting the specification to an implementation that does not incur a prohibitively high performance penalty. While invariants, pre-conditions and post-conditions could be provided in different modeling languages, we leverage the Alloy tool-set, specifically the Alloy language and the Alloy Analyzer for systematically repairing erroneous states. Four different algorithms are presented and implemented in our data structure repair framework. These algorithms can repair a medium sized erroneous data structure in a few seconds. We introduce repair guide annotations defined by the user to improve the accuracy and performance of the repair mechanism. Experiments using complex specifications show the approach holds much promise in increasing software reliability. / text

Programs developed with standard techniques often fail when they encounter any of a variety of internal errors. We present a set of techniques that prevent programs from failing and instead enable them to continue to execute even after they encounter otherwise fatal internal errors. Our results indicate that even though the techniques may take the program outside of its anticipated execution envelope, the continued execution often enables the program to provide acceptable results to their users. These techniques may therefore play an important role in making software systems more resilient and reliable in the face or errors. / Singapore-MIT Alliance (SMA)