Turning a failing test or stack trace into a candidate fix
An engineer hands over the failing assertion, the stack trace and the implicated files. The model proposes a localisation — which function is most likely at fault — and a candidate patch. That patch is a hypothesis to run against the suite and read line by line, not a change to merge.
- Effort
- Hours of work
- Skill level
- Some technical skill
- Organisation size
- Small business
- Value
- Time saved
Tools named for this
- A repository-aware agent permitted to run the test suite
- A large language model given the trace, the failing test and the implicated files
- The project's existing fault-localisation and coverage output as supporting context
What to check before you ship it in India
- Production logs pasted into a debugging transcript are a copy of personal data sitting outside your retention pipeline. Section 8(7) requires erasure once the specified purpose is no longer served, and a chat history that no system owns is exactly where that obligation is quietly missed. Note the carve-out the section itself carries: erasure is required UNLESS retention is necessary for compliance with any law in force, so check whether a sectoral or record-keeping rule requires you to keep it before scheduling a delete.
- Repair research has a name for the main failure here: an overfitting patch, one that fails to correctly repair the defect beyond making the tests pass. A green suite is evidence about the suite.
Sources
Every claim on this page traces to one of these, on the date it was read.
- The Digital Personal Data Protection Act, 2023 (No. 22 of 2023) — most obligations commence 13 May 2027 under the DPDP Rules 2025 — s.8(7) · Ministry of Electronics and Information Technology · a rule · read 2026-09-01
- A Systematic Literature Review on Large Language Models for Automated Program Repair · arXiv (Zhang, Fang, Xie, Ma, Sun, Yang, Chen) · that this is done · read 2026-09-01
- Automated Classification of Overfitting Patches with Statically Extracted Code Features · arXiv (Ye, Gu, Martinez, Durieux, Monperrus) · how it is done · read 2026-09-01
- SWE-bench: Can Language Models Resolve Real-World GitHub Issues? · arXiv (Jimenez et al.); ICLR 2024 · how it is done · read 2026-09-01