Software rebuilding is the process of deriving a deployable software system from its primitive source objects. A build tool helps maintain consistency between the derived objects and source objects by ensuring that all necessary build steps are re-executed in the correct order after a set of changes is made to the source objects. It is imperative that derived objects accurately represent the source objects from which they were supposedly constructed; otherwise, subsequent testing and quality assurance is invalidated. This thesis aims to advance the state-of-the-art in tool support for automated software rebuilding. It surveys the body of background work, lays out a set of design considerations for build tools, and examines areas where current tools are limited. It examines the properties of a next-generation tool concept, redo, conceived by D. J. Bernstein; redo is novel because it employs a purely top-down approach to software rebuilding that promises to be simpler, more flexible, and more reliable than current approaches. The details of a redo prototype written by the author of this thesis are explained including the central algorithms and data structures. Lastly, the redo prototype is evaluated on some sample software systems with respect to migration effort between build tools as well as size, complexity, and performances aspects of the resulting build systems.

Software rebuilding is the process of deriving a deployable software system from its primitive source objects. A build tool helps maintain consistency between the derived objects and source objects by ensuring that all necessary build steps are re-executed in the correct order after a set of changes is made to the source objects. It is imperative that derived objects accurately represent the source objects from which they were supposedly constructed; otherwise, subsequent testing and quality assurance is invalidated. This thesis aims to advance the state-of-the-art in tool support for automated software rebuilding. It surveys the body of background work, lays out a set of design considerations for build tools, and examines areas where current tools are limited. It examines the properties of a next-generation tool concept, redo, conceived by D. J. Bernstein; redo is novel because it employs a purely top-down approach to software rebuilding that promises to be simpler, more flexible, and more reliable than current approaches. The details of a redo prototype written by the author of this thesis are explained including the central algorithms and data structures. Lastly, the redo prototype is evaluated on some sample software systems with respect to migration effort between build tools as well as size, complexity, and performances aspects of the resulting build systems.