valkyrjaio / sindri-java / 30510857961

100%

push

github

[CI] Fail the JUnit build below 100% coverage (#67)

# Description

Coverage was reported here and enforced nowhere. `junit` ran `test`
finalized by `jacocoTestReport`,
so the report was *generated* and then nothing asserted anything about
it —
`jacocoTestCoverageVerification` was never wired in. A build at 55%
passed exactly like one at 100%,
and the only thing that noticed a drop was Coveralls, *after* push, on
an already-open PR.

The architecture guide's definition of done is 100% line **and** branch,
per file, and "must never
drop". This makes the gate say so.

## Two rules, not one

A bundle-wide rule alone is not enough, and there is a concrete case
proving it:
[sindri-java#62](https://github.com/valkyrjaio/sindri-java/pull/62)
landed a new class at **65% line
/ 50% branch** while the repo-wide number only fell from 100% to 96.37%.
A large, well-covered
codebase absorbs one entirely untested new class almost without moving.

So the verification carries a **bundle** rule *and* a per-**CLASS**
rule, both at 100% line and
branch. Confirmed against a deliberately uncovered probe class — note
how differently the two rules
report the same file:

```
Rule violated for bundle junit: lines covered ratio is 0.99, but expected minimum is 1.00
Rule violated for bundle junit: branches covered ratio is 0.99, but expected minimum is 1.00
Rule violated for class io.valkyrja.probe.ThresholdProbe: lines covered ratio is 0.00, ...
Rule violated for class io.valkyrja.probe.ThresholdProbe: branches covered ratio is 0.00, ...
```

The bundle barely notices; the class rule names the file.

## Exclusions are shared, not duplicated

The report's class-directory exclusions are lifted into a single
`coverageExclusions` list that both
`jacocoTestReport` and `jacocoTestCoverageVerification` consume. Scoping
only one of them would let
the gate quietly measure a different set of classes than the report
shows.

## Types of changes

- [x] Improvement _(non-breaking change whic... (continued)

618 of 618 branches covered (100.0%)

Branch coverage included in aggregate %.

1679 of 1679 relevant lines covered (100.0%)

4.39 hits per line