Full text
| |
| Author(s): |
Fernando Jose Castor de Lima Filho
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | Campinas, SP. |
| Institution: | Universidade Estadual de Campinas (UNICAMP). Instituto de Computação |
| Defense date: | 2006-11-30 |
| Examining board members: |
Cecília Mary Fischer Rubira;
Paulo Henrique Monteiro Borba;
Avelino Francisco Zorzo;
Luiz Eduardo Busato;
Eliane Martins
|
| Advisor: | Cecília Mary Fischer Rubira |
| Abstract | |
Exception handling mechanisms were conceived as a means to help managing the complexity of fault-tolerant software. They promote an explicit textual separation between normal code and the code that deals with abnormal situations, in order to support the construction of programs that are more concise, evolvable, and reliable. Several mainstream programming languages and most of the existing component models implement exception handling mechanisms. In spite of its many bene?ts, exception handling can be a source of many design faults if used in an ad hoc fashion. Recent studies show that developers of large-scale software systems based on component infrastructures have habits concerning the use of exception handling that make applications vulnerable to faults and hard to maintain. Software components introduce new challenges which are not addressed by traditional exception handling mechanisms and increase the chances of problems occurring. Examples include unavailability of source code and architectural mismatches. In this work, we propose two complementary techniques centered on exception handling for the construction of fault-tolerant component-based systems. Both of them emphasize system structure as a means to reduce the impactof fault tolerance mechanisms on the overall complexity of a software system and the number of design faults that stem from complexity. The ?rst one is an approach for the architectural design of a system?s error handling capabilities. It addresses the problem of verifying whether a software architecture satis?es certain properties of interest pertaining the ?ow of exceptions between architectural components, e.g., if all the exceptions signaled at the architectural level are eventually handled. The proposed approach is based on a set of existing tools that automate this process as much as possible. The second one consists in applying aspect-oriented programming (AOP) to better modularize exception handling code. We have conducted a through study aimed at improving our understanding of the efects of AOP on exception handling code and identifying the situations where its use is advantageous and the ones where it is not (AU) | |