(2017) Debugging with Probabilistic Event Structures.
|
Text
p437-soremekun(1).pdf Download (166kB) | Preview |
Abstract
Debugging is a search process to find, understand and fix the root cause of software defects. Can debugging benefit from probabilistic information? We hypothesize that debugging activities can benefit from probabilistic information that capture the statistical dependence of program features and the minor variations of program behavior. This probabilistic information helps to guide the search for the root cause of the bug and provides detailed diagnostic information (such as failure-inducing inputs and method calls leading to the fault). To realize our hypothesis, we propose to improve debugging activities by guiding bug diagnosis using both probabilistic reasoning and program analysis. The main idea is to mine probabilistic information from program executions, then apply these information to construct probabilistic event structures (e.g. probabilistic call graphs) that guides debugging activities such as fault localization and comprehension. The resulting probabilistic model will guide bug diagnosis towards the most likely paths to the root cause of bugs and provide contextual diagnostic information.
| Item Type: | Conference or Workshop Item (A Paper) (Paper) |
|---|---|
| Divisions: | Andreas Zeller (Software Engineering, ST) |
| Depositing User: | Ezekiel Soremekun |
| Date Deposited: | 15 Feb 2018 08:09 |
| Last Modified: | 14 Apr 2020 10:40 |
| URI: | https://publications.cispa.saarland/id/eprint/1471 |
Actions
Actions (login required)
 |
View Item |