A wide spectrum language is presented, which is designed to facilitate the proof of the correctness of refinements and transformations. Two different proof methods are introduced and used to prove some fundamental transformations, including a general induction rule (Lemma 3.9) which enables transformations of recursive and iterative programs to be proved by induction on their finite truncations. A theorem for proving the correctness of recursive implementations is presented (Theorem 3.21), which provides a method for introducing a loop, without requiring the user to provide a loop invariant. A powerful, general purpose, transformation for removing or introducing recursion is described and used in a case study (Section 5) in which we take a small, but highly complex, program and apply formal transformations in order to uncover an abstract specification of the behaviour of the program. The transformation theory supports a transformation system, called FermaT, in which the applicability conditions of each transformation (and hence the correctness of the result) are mechanically verified. These results together considerably simplify the construction of viable program transformation tools; practical consequences are briefly discussed.

Software reengineering has been described as being "about as easy as reconstructing a pig from a sausage" [11]. But the development of program transformation theory, as embodied in the FermaT transformation system, has made this miraculous feat into a practical possibility. This paper describes the theory...

Abstract not found

Non-Transparent Debugging of Optimized Code by Caroline Mae Tice Doctor of Philosophy in Computer Science University of California at Berkeley Professor Susan L. Graham, Chair Debugging optimized code is a problem for which a widely accepted solution has yet to be found. Over the years many approaches have been suggested, including limiting the compiler optimizations, restricting the debugger functionality, using recompilation or dynamic de-optimization to undo the optimizations, and having the debugger determine the e#ects of optimizations and mask them from the user. All of these approaches have a common thread: they place a barrier between the user and the optimizations, either altering, undoing, or hiding the e#ects of optimizations.

Modern development environments support refactoring by providing atomically behaviour-preserving transformations. While useful, these transformations are limited in three ways: (i) atomicity forces transformations to be complex and opaque, (ii) the behaviour preservation requirement disallows deliberate behaviour evolution, and (iii) atomicity limits code reuse opportunities for refactoring implementers. We present ‘program metamorphosis’, a novel approach for program evolution and refactoring that addresses the above limitations by breaking refactorings into smaller steps that need not preserve behaviour individually. Instead, we ensure that sequences of transformations preserve behaviour together, and simultaneously permit selective behavioural change. To evaluate program metamorphosis, we have implemented a prototype plugin for Eclipse. Our analysis and experiments show that (1) our plugin provides correctness guarantees on par with those of Eclipse’s own refactorings, (2) both our plugin and our approach address the aforementioned limitations, and (3) our approach fully subsumes traditional refactoring.