konflux-ci/integration-service/estimating-codecov-patch-coverage

Estimate Codecov Patch Coverage

Use this skill from the repository root before opening or updating a pull request that changes Go code. The repository's hack/check_patch_coverage.py estimates coverage of the added lines in the patch, not coverage of the whole repository.

1. Run:

``bash make codecov-estimate ``

This runs the normal test suite to produce a fresh cover.out, then estimates coverage against the 85% target. Override CODECOV_BASE for a pull request targeting another branch, or CODECOV_TARGET if the policy changes.

1. Review the changed lines listed as uncovered and add tests for those paths.
2. Re-run make codecov-estimate after making changes.

Do not proceed with PR preparation when make test fails, the checker cannot run, the checker warns that the local Go version differs from CI, or the estimate is below 85%. The checker exits 1 below 85% and exits 2 when it cannot produce a reliable estimate.

If the checker warns about the Go version, install the version specified in .github/workflows/pr.yaml and re-run with it, for example:

GOTOOLCHAIN=go1.26.0 make codecov-estimate

The checker compares the working tree with the merge base of origin/main by default. Use make codecov-estimate CODECOV_BASE=origin/<branch> when the pull request targets a different branch, and ensure the selected base ref is current. Run it again after the final Go edit; it rejects a coverage profile older than any changed Go source file.

Without an explicit threshold, the checker reads the target from codecov.yml. The Make target passes 85% explicitly because the repository may temporarily have a lower informational target while it prepares for enforcement.

Changes with no coverable Go lines are reported as not applicable. Generated, vendored and other paths listed in codecov.yml are excluded from the estimate. The checker mirrors Codecov's Go coverage-region merging and Go source-line fixes; it was calibrated against the published result for integration-service PR 1706. It remains a preflight estimate because Codecov uses the uploaded CI report and the provider's server-side pull-request diff. The Codecov PR check is authoritative.

Reading the Codecov result

Codecov's pull-request result is about the changed lines, not the whole repository. A result such as 70.77% <100.00%> (+0.25%) means project coverage, patch coverage, and the project change respectively; the patch value is the one that will be enforced. A local result at or above 85% increases confidence but is not a guarantee that Codecov will pass.