Bugsplainer: Leveraging Code Structures to Explain Software Bugs with Neural Machine Translation

Parvez Mahbub, M. Masudur Rahman, Ohiduzzaman Shuvo, Avinash Gopal. Bugsplainer: Leveraging Code Structures to Explain Software Bugs with Neural Machine Translation. In Proceeding of The 39th IEEE International Conference on Software Maintenance and Evolution (ICSME 2023), pp. 06, Bogota, Columbia, October 2023

Transcript

1. Bugsplainer: Leveraging Code Structures to Explain Software Bugs with Neural Machine Translation Parvez Mahbub Dalhousie University parvezmrobin@dal.ca Masud Rahman Dalhousie University masud.rahman@dal.ca Ohiduzzaman Shuvo Dalhousie University oh599627@dal.ca Avinas Gopal Metabob avi@metabob.com
2. 2 2 Should we explain bugs? 50%
3. We present Bugsplainer — a novel transformer- based generative model Can leverage code structures Trained using both buggy and fixed source code 3
4. 4 In: Bug-fix Commit Generate diffSBT Discriminatory Pre-train diffSBT of Bug-free Nodes diffSBT of Buggy Nodes Fine-tune Commit Message Out: Fine- tuned Model In: Buggy Code Generate diffSBT Fine-tuned Model Out: Generated Explanation In: Line Numbers to Explain diffSBT How does it work? Training Bugsplainer Explanation Generation
5. 5 A B C D E (A (B )B (C(D)D(E)E)C)A Structure-Based Traversal (SBT)
6. 6 DiffSBT
7. Show me the explanation! Ground truth: Fix a bug where the lyricswiki fetcher would try to unescape an empty (None) response and crash CommitGen: Small bug fix for error handling NNGen: fix UnicodeDecodeError with non- ASCII text Fine-tuned CodeT5: Don’t try to get lyrics if we are licensed pyflakes: no error found Bugsplainer: fix crash when lyrics not found
8. Bugsplainer meets ChatGPT 8 Ground truth: Fix a bug where the lyricswiki fetcher would try to unescape an empty (None) response and crash
9. Human evaluation 9 # Developers # Countries Programming Experience Bugfix Experience 20 6 1-10 years 1-7 years
10. Thank You! Questions? Masud Rahman masud.rahman@dal.ca