fix(sonar): raise overall coverage floor 80% -> 85% (FAR-835) [partial] - #536
Conversation
farnalabs
added
agent-generated
farnalabs
left a comment
farnalabs
left a comment
There was a problem hiding this comment.
Review feedback (non-blocking): The change is a clean, minimal bump of the overall unit-test coverage floor from 80% to 85% in .github/scripts/coverage-thresholds.sh (both the eco message and --cov-fail-under are updated consistently, so they cannot drift). CI [Test (Backend)] passes at the new 85% floor, so the branch is above the stricter gate. One minor follow-up: docs/architecture.md:689 still states Coverage targets: ... overall 80% - please update that line (or in a follow-up) so the documented target matches the enforced one. No test files removed and no skip/xfail markers added.
modulo-reviewbot
left a comment
modulo-reviewbot
left a comment
There was a problem hiding this comment.
Approved by the PR Reviewer post-decision node.
Low-risk change: .github/scripts/coverage-thresholds.sh raises the overall unit-test coverage floor from 80% to 85%, consistently in both the workflow-group message and the --cov-fail-under argument. No test files deleted, no skip/xfail markers added. CI Test (Backend) passes at the new 85% floor.
Non-blocking follow-up: docs/architecture.md:689 still documents the coverage target as 80% and is now stale vs the enforced 85% floor; update in this PR or a follow-up.
|
2 participants
Ramps the repo's own overall coverage floor from 80% to 85% (part of FAR-835).
Measured first, not assumed. The value enforced by
--cov-fail-underis pytest-cov's own measure, which is a different number from SonarCloud's (Sonar is line+branch over a different file set). Read from the latest main CI run (34854306043):92.85% gives 7.85pp of headroom over the new 85% floor, so this cannot red-line CI today. Per-module floors (auth 90%, pipeline_engine 85%, db/rls 95%) are unchanged.
This is the ratchet companion to the changed-lines coverage gate in #534: the changed-lines gate catches a PR whose own new code is under-tested; this floor stops the whole-repo measure drifting down over time. Next step once coverage sustains it: 90%, then 95%.