To report a compiler bug, one must often find a small test case that triggers the bug. The existing approach to automated test-case reduction, delta debugging, works by removing substrings of the original input; the result is a concatenation of substrings that delta cannot remove. We have found this approach less than ideal for reducing C programs because it typically yields test cases that are too large or even invalid (relying on undefined behavior). To obtain small and valid test cases consistently, we designed and implemented three new, domain-specific test-case reducers. The best of these is based on a novel framework in which a generic fixpoint computation invokes modular transformations that perform reduction operations. This reducer produces outputs that are, on average, more than 25 times smaller than those produced by our other reducers or by the existing reducer that is most commonly used by compiler developers. We conclude that effective program reduction requires more than straightforward delta debugging.

Aggressive random testing tools ("fuzzers") are impressively effective at finding compiler bugs. For example, a single test-case generator has resulted in more than 1,700 bugs reported for a single JavaScript engine. However, fuzzers can be frustrating to use: they indiscriminately and repeatedly find bugs that may not be severe enough to fix right away. Currently, users filter out undesirable test cases using ad hoc methods such as disallowing problematic features in tests and grepping test results. This paper formulates and addresses the fuzzer taming problem: given a potentially large number of random test cases that trigger failures, order them such that diverse, interesting test cases are highly ranked. Our evaluation shows our ability to solve the fuzzer taming problem for 3,799 test cases triggering 46 bugs in a C compiler and 2,603 test cases triggering 28 bugs in a JavaScript engine.

This paper addresses the problem of creating abstract transfer functions supporting dataflow analyses. Writing these functions by hand is problematic: transfer functions are difficult to understand, difficult to make precise, and difficult to debug. Bugs in transfer functions are particularly serious since they defeat the soundness of any program analysis running on top of them. Furthermore, implementing transfer functions by hand is wasteful because the resulting code is often difficult to reuse in new analyzers and to analyze new languages. We have developed algorithms and tools for deriving transfer functions for the bitwise and unsigned interval abstract domains. The interval domain is standard; in the bitwise domain, values are vectors of three-valued bits. For both domains, important challenges are to derive transfer functions that are sound in the presence of integer overflow, and to derive precise transfer functions for operations whose semantics are a mismatch for the domain (i.e., bit-vector operations in the interval domain and arithmetic operations in the bitwise domain). We can derive transfer functions, and execute them, in time linear in the bitwidth of the operands. These functions are maximally precise in most cases. Our generated transfer functions are parameterized by a bitwidth and are independent of the language being analyzed, and also of the language in which the analyzer is written. Currently, we generate interval and bitwise transfer functions in C and OCaml for analyzing C source code, ARM object code, and AVR object code. We evaluate our derived functions by using them in an interprocedural dataflow analyzer.

Performance guarantees can be given to tasks in an embedded system by ensuring that access to each shared resource is mediated by an appropriate scheduler. However, almost all previous work on CPU scheduling has focused on thread-level scheduling, resulting in systems that are vulnerable to a lower-level form of overload that occurs when too many interrupts arrive. This paper describes three new techniques, two software-based and one hardware-based, for creating systems that delay or drop excessive interrupt requests before they can overload a processor. Our interrupt schedulers bound both the amount of work performed in interrupt context and its granularity, making it possible to provide strong progress guarantees to thread-level processing. We show that our solutions work and are efficient when implemented on embedded processors. We have also taken a description for a microprocessor in VHDL, modified it to include logic that prevents interrupt overload, synthesized the processor, and verified that it works using simulation. By allowing developers to avoid making assumptions about the worst-case interrupt rates of peripherals, our work fills an important gap in the chain of reasoning leading to a validated system. These techniques cannot replace careful system design, but they do provide a lastditch safety guarantee in the presence of a serious malfunction.

Many abstract value domains such as intervals, bitwise, constants, and value-sets have been developed to support dataflow analysis. Different domains offer alternative tradeoffs between analysis speed and precision. Furthermore, some domains are a better match for certain kinds of code than others. This paper presents the design and implementation of cXprop, an analysis and transformation tool for C that implements "conditional X propagation," a generalization of the well-known conditional constant propagation algorithm where X is an abstract value domain supplied by the user. cXprop is interprocedural, context-insensitive, and achieves reasonable precision on pointer-rich codes. We have applied cXprop to sensor network programs running on TinyOS, in order to reduce code size through interprocedural dead code elimination, and to find limited-bitwidth global variables. Our analysis of global variables is supported by a novel concurrency model for interrupt-driven software. cXprop reduces TinyOS application code size by an average of 9.2% and predicts an average data size reduction of 8.2% through RAM compression.