Testing a PL/I Compiler Using Precomputation-based Program Generation

Jesse Postema,Johan Fabry, Yannick Barthol,Ana Oprescu

2022 IEEE Conference on Software Testing, Verification and Validation (ICST)(2022)

引用 0|浏览9
暂无评分
摘要
In automated compiler testing, the focus typically lies in uncovering bugs caused by optimisations performed by the compiler. However, there is a class of compilers where little to no optimisations are performed: those for migration of legacy software. Therefore, it is not clear to what extent such legacy compilers would benefit from automated compiler testing. We investigated this in the context of the Raincode legacy compiler for PL/I, an industrial compiler targeting the. NET platform. We designed and implemented a framework for automated PL/I compiler testing through precomputation-based program generation and ran it on two versions of the Raincode PL/I compiler: an older with known bugs and the latest release. On the older version, our framework generated around 127.000 programs and found five bugs, two of which were previously unknown to us. For the latest compiler release, after 180 hours of tests and more than 718.000 generated programs, the framework did not reveal any bugs.
更多
查看译文
关键词
precomputation-based program generation,legacy software,Raincode legacy compiler,industrial compiler,program bugs,PL/I compiler testing,compiler release,time 180.0 hour
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要