| FazBrowse GitHub Viewer | Trending | | Home |
| Tools: [Download Repo ZIP] [Original HTTPS Page] |
Sorry, something went wrong.
The coverage gate previously flagged any percentage drop > 0.05% as a regression. This caused false positives when well-covered code was extracted/moved out of a class — the percentage drops due to a ratio shift even though no test coverage was actually lost. Now a regression requires BOTH a percentage drop AND an increase in the absolute number of missed items. This correctly distinguishes real regressions (new uncovered code) from cosmetic ratio changes caused by refactoring. https://claude.ai/code/session_01XVBWUo2VM4B82ssZm1f2mu
Test ReportTest Results
Code Coverage (Java 25)
|
Sorry, something went wrong.
When a class shows a coverage decrease, the PR report comment now includes a collapsible <details> section showing exactly which methods are uncovered or changed. The coverage gate failure message also includes method-level context. The JaCoCo XML parser is enhanced to extract per-method counters (name, descriptor, line/branch coverage) within each class. https://claude.ai/code/session_01XVBWUo2VM4B82ssZm1f2mu
When baseline method data is available, the method-level detail sections now only show methods whose coverage actually changed or that are new/ removed. Previously the hasMissed filter would show every partially-covered method, creating noise that buried the actual regressions. Falls back to showing all uncovered methods when no baseline method data exists yet. https://claude.ai/code/session_01XVBWUo2VM4B82ssZm1f2mu
The buildMethodDetails function now uses the same hybrid check as the coverage gate: percentage drop AND missed count increase. Previously it used a percentage-only check, which would show misleading detail sections for classes where code was merely extracted or moved. https://claude.ai/code/session_01XVBWUo2VM4B82ssZm1f2mu
| Back | FazBrowse Home | New Git URL |
Summary
Files changed
Test plan
https://claude.ai/code/session_01XVBWUo2VM4B82ssZm1f2mu