Abstract Verification is usually performed on a high‐level view of the software, either specification or program source code. However, in certain circumstances verification is more relevant when performed at the machine‐code level. This paper focuses on automatic test data generation from a stand‐alone executable. Low‐level analysis is much more difficult than high‐level analysis since even the control‐flow graph is not available and bit‐level instructions have to be modelled faithfully. The paper shows how ‘path‐based’ structural test data generation can be adapted from structured language to machine code, using both state‐of‐the‐art technologies and innovative techniques. The results have been implemented in a tool named OSMOSE and encouraging experiments have been conducted. Copyright © 2010 John Wiley & Sons, Ltd.

Recent advances in path-based test (data) generation open the way to the systematic testing of large scale programs. However, these technologies are still suffering from two major bottlenecks: efficient constraint solving and the path explosion phenomenon. We focus in this paper on the second issue and propose three complementary heuristics geared toward lowering path explosion. All these heuristics are both easy to implement and lightweight, and each one deals with a distinct source of path explosion. We provide theoretical and experimental evidence that they can achieve a significant reduction in the search space.

Verification is usually performed on a high-level view of the software, either specification or program source code. However in certain circumstances verification is more relevant when performed at the machine code level. This paper focuses on automatic test data generation from a standalone executable. Low-level analysis is much more difficult than high-level analysis since even the control-flow graph is not available and bit-level instructions have to be modelled faithfully. We show how "path-based" structural test data generation can be adapted from structured language to machine code, using both state-of-the-art technologies and innovative techniques. Our results have been implemented in a tool named OSMOSE and encouraging experiments have been conducted.