fix: clean-lens findings + trim the dogfood swarm to strong reviewers (#4)
Build & push image / build-and-push (push) Successful in 9s
Build & push image / build-and-push (push) Successful in 9s
emit() now skips findings extraction for a "No material issues found" lens (its path:line refs are verification notes, not problems), fixing the FP inflation that penalized thorough clean-pass reviewers. Also trims the dogfood swarm to the strong reviewers: drops m5/qwen3.6 (last local lane), gemma4, gpt-oss:120b, and kimi-k2.7-code — leaving 6 cloud + claude-code/sonnet. Fittingly, PR #4's own 11-model review produced 43 findings that were ALL clean-verification bullets (zero real) — a live demonstration of the bug this fixes. gofmt clean, go vet quiet, go test -race green. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: Steve Dudenhoeffer <steve@stevedudenhoeffer.com> Co-committed-by: Steve Dudenhoeffer <steve@stevedudenhoeffer.com>
This commit was merged in pull request #4.
This commit is contained in:
@@ -121,6 +121,14 @@ func emit(results []specialistResult, elapsed time.Duration) {
|
||||
if r.errored {
|
||||
continue // a failed lens contributes no findings
|
||||
}
|
||||
// A lens that reports "No material issues found" has nothing to flag —
|
||||
// its path:line references are verification notes ("verified X at
|
||||
// file:line is safe"), not problems. Extracting them pollutes the
|
||||
// findings store with false positives and unfairly penalizes thorough
|
||||
// reviewers that do clean passes, so a clean lens emits no findings.
|
||||
if r.verdict == verdictClean {
|
||||
continue
|
||||
}
|
||||
sev := r.verdict.label()
|
||||
for _, f := range parseFindings(r.spec, r.out) {
|
||||
reports = append(reports, reportPayload{
|
||||
|
||||
@@ -179,6 +179,46 @@ func TestEmit_PostsRunsAndReports(t *testing.T) {
|
||||
}
|
||||
}
|
||||
|
||||
// A clean "No material issues found" lens must NOT emit findings, even though
|
||||
// its markdown contains path:line references (those are verification notes, not
|
||||
// problems). /runs is still posted; /reports is not.
|
||||
func TestEmit_SkipsCleanVerdictLens(t *testing.T) {
|
||||
var runs, reports int32
|
||||
srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
|
||||
switch r.URL.Path {
|
||||
case "/runs":
|
||||
atomic.AddInt32(&runs, 1)
|
||||
w.WriteHeader(http.StatusOK)
|
||||
_, _ = w.Write([]byte(`{"run_id":"x"}`))
|
||||
case "/reports":
|
||||
atomic.AddInt32(&reports, 1)
|
||||
w.WriteHeader(http.StatusOK)
|
||||
default:
|
||||
http.Error(w, "not found", http.StatusNotFound)
|
||||
}
|
||||
}))
|
||||
defer srv.Close()
|
||||
|
||||
t.Setenv("GADFLY_FINDINGS_URL", srv.URL)
|
||||
t.Setenv("GADFLY_REPO", "owner/repo")
|
||||
t.Setenv("GADFLY_PR", "7")
|
||||
t.Setenv("GADFLY_MODEL", "ollama-cloud/qwen3")
|
||||
|
||||
// Clean verdict, but the markdown is full of path:line "verified X" notes.
|
||||
cleanMarkdown := "No material issues found.\n\nVerified `run/executor.go:166` handles the error.\n"
|
||||
results := []specialistResult{
|
||||
{spec: Specialist{Name: "security"}, out: cleanMarkdown, verdict: verdictClean},
|
||||
}
|
||||
emit(results, time.Second)
|
||||
|
||||
if got := atomic.LoadInt32(&runs); got != 1 {
|
||||
t.Fatalf("/runs received %d times, want 1 (the run is still recorded)", got)
|
||||
}
|
||||
if got := atomic.LoadInt32(&reports); got != 0 {
|
||||
t.Fatalf("/reports received %d times, want 0 (clean lens emits no findings)", got)
|
||||
}
|
||||
}
|
||||
|
||||
func TestEmit_DisabledMakesNoRequests(t *testing.T) {
|
||||
var hits int32
|
||||
srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
|
||||
|
||||
Reference in New Issue
Block a user