What’s the problem? interrogating actors to identify the root cause of concurrency bugs

Autor: Carmen Torres Lopez, Louise Van Verre, Elisa Gonzalez Boix
Přispěvatelé: Castegren, Elias, De Koster, Joeri, Fowler, Simon, Informatics and Applied Informatics, Software Languages Lab, Faculty of Sciences and Bioengineering Sciences
Rok vydání: 2021
Předmět:
Zdroj: AGERE!@SPLASH
Popis: Programs written using Communicating Event-Loops (CEL) concurrency model do not suffer from low-level data races by design but are not exempt from other concurrency bugs, such as behavioral deadlocks and message order violations. When programmers need to find the root cause of a bug, they typically ask questions about the application's behavior. However, current debugging tools are mostly operational, offering features at the source code level like breakpoints and watchpoints. Consequently, understanding the program behavior when debugging can take a lot of time for developers since questions on behaviors need to be mapped into operations in the debugger. Inspired by interrogative debugging, this paper proposes an interactive debugging approach for actor-based programs that enable developers to reason about the program output by selecting questions from a set of predefined questions about the code and the program's execution. We present the design of the questions and answers, and we describe a prototype implementation in Apgar, an online debugger for actor-based programs written in SOMns. We define questions based on key concepts of the actor model: actors, turns, messages, and promises. The debugger then computes the answers by analyzing a recorded trace of events about the program execution.
Databáze: OpenAIRE