AI-Written Tests Passed While Hiding a Penny-Off Bug in Every UK Invoice

·1 views

A developer building a .NET invoicing library with AI coding assistant Claude discovered that 202 passing tests failed to catch a rounding error that would have misstated every UK invoice containing three items priced at £3.99. The bug stemmed from the library's rounding helper being tested in isolation, while the invoice calculator itself was never verified to actually call that helper. Manual mutation testing — deliberately breaking code to see which tests fail — exposed the gap, revealing that swapping the correct half-up rounding for .NET's default banker's rounding caused only five failures, none involving real invoice scenarios. Further probing found similar blind spots in discount calculations and a culture-formatting test that used round numbers, making it unable to detect the very formatting trap it was designed to catch. The episode illustrates a key risk with AI-generated test suites: when the same system writes both the code and its tests, passing results may only confirm internal consistency rather than correctness.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Developer Seeks Community Help to Validate React Flex Forms SPFx Sample

A developer has built a SharePoint Framework sample called React Flex Forms, consisting of two web parts: a Form Designer and a Form Renderer that saves submissions to a SharePoint list. The sample has passed local lint, build, and packaging checks, but still requires real-tenant validation and screenshots for the official PnP sample README. Anyone with access to a SharePoint Online tenant is being asked to test the web parts and capture scrubbed screenshots showing the designer and renderer in action. Feedback on provisioning, permissions, accessibility, responsive behavior, and theme support is also welcome. Contributions can be submitted directly to pull request #6473 on the project repository.

0 comments Read more at DEV Community

ProgrammingHacker News ·

Survey: Majority of Britons want private messaging to remain encrypted

A new survey indicates that most people in the United Kingdom want their private messages to remain secure and unmonitored. The findings come amid ongoing government debates over whether tech companies should be required to provide backdoor access to encrypted communications. Authorities have argued that such access is necessary for national security and law enforcement purposes. Privacy advocates, however, warn that weakening encryption would expose ordinary users to significant security risks. The survey results suggest public opinion leans strongly toward protecting message privacy over expanded surveillance capabilities.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

How One Developer Built a 237-Endpoint Vedic Astrology API and the Bugs Found Along the Way

A developer recently published a technical retrospective on building GrahaAPI, a REST API covering 237 endpoints across 23 modules of Vedic astrology with bilingual Hindi and English responses. The project surfaced a subtle thread-safety bug in the underlying C ephemeris library, where FastAPI's threadpool silently defaulted new threads to tropical zodiac mode instead of sidereal, returning plausible but astronomically wrong planetary positions with no errors. To validate accuracy, the team built regression tests anchored to centuries-old classical invariants — such as Vimshottari dasha periods summing to exactly 120 years and fixed choghadiya sequences — treating any test failure as a code defect, never a test flaw. One such test caught a time-formatting bug where 59.57 seconds was rounded to 60 without carrying over into the next minute, displaying 07:59:34 as 07:00. The post highlights how legacy domain knowledge with frozen specifications can serve as unusually reliable test fixtures for modern software systems.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Vincent 0.7.0 Released: AI Coding Agent Now Manages Its Own Development

Developer tool Vincent has reached version 0.7.0, with the key milestone being that the project now uses its own workflows to manage its entire development lifecycle, from issue creation to release. The update introduces structured workflow inputs with validation, an expanded CLI enabling scriptable task management without a terminal UI, and resilience features such as retry backoff, repair agents, and daemon backup and restore. A new workflow-authoring skill prioritizes deterministic automation over AI agents, invoking reasoning models only where genuinely needed. The release also re-adopts the MIT license and is designed to work as a vendor-independent local control plane compatible with tools like Claude Code, Codex, or Cursor.

0 comments Read more at DEV Community