Claude (review-harness setup) 4dc53c5798 Phase D — validator cross-checks LLM findings + 2 close-out fixes
Implements PROMPT.md / docs/REVIEW_PIPELINE.md Phase 3:
"AI may suggest. Code validates."

internal/validators/validate.go — 3 hard checks per the
"Reject A Finding If" list:
- file does not exist (with path-traversal guard against the LLM
  hallucinating ../../../etc/passwd)
- cited evidence does not appear in the file (verbatim or
  trim-line-by-line — models often re-indent quotes when quoting code)
- line hint exceeds file line count

3 soft checks documented as open (claim semantics, suggested-fix
relevance, invented tests/commands — all need another LLM pass).

internal/validators/validate_test.go — 9 tests including:
- TestValidate_RejectsNonexistentFile (gate D1)
- TestValidate_RejectsEvidenceNotInFile
- TestValidate_RejectsLineHintBeyondFile
- TestValidate_AcceptsRealFinding
- TestValidate_AcceptsEvidenceWithDifferentLeadingWhitespace
- TestValidate_RejectsEmptyEvidence
- TestValidate_PassesThroughStaticFindings
- TestValidate_RejectsPathEscapingRepo (path-traversal protection)
- TestValidate_AcceptsRelativeRepoPath (the regression — see below)

Pipeline phase 3 wired between LLM review (Phase C) and report gen
(Phase 4). validated-findings.json contains the confirmed set;
rejected-findings.json contains rejects with per-finding reason +
detail. Receipt phase entry honest about output files + status.

=== Bug J caught ===

First Phase D run rejected EVERY real LLM finding as file_not_found
because the path-traversal check compared a relative joined path
(`tests/fixtures/insecure-repo/src/handler.go`) against an absolute
repoAbs (`/home/profit/share/.../insecure-repo`), so HasPrefix
always returned false. Both sides now resolved via filepath.Abs
before comparison. Regression test
TestValidate_AcceptsRelativeRepoPath locks this in — runs the
validator against a relative repo path AND a relative chdir, the
exact shape that hit the bug.

J's framing was honest: "I don't know what the problem is, but you
know what we're trying to accomplish." The fix-it-yourself signal
let me trace through the rejection details + see the smoking gun
in the detail string ("escapes repo root"). Without that prompt the
9 false rejections might have looked like real LLM bugs.

=== 2 close-out fixes ===

1. .gitignore: changed `/reports/latest/` → `**/reports/latest/`
   (and same for `run-*`). Phase C committed 22 generated files
   from `tests/fixtures/*/reports/latest/` because the original
   pattern was anchored at the harness root only. Existing tracked
   files removed via git rm --cached; new pattern keeps fixture
   reports out of version control going forward.

2. pipeline.cleanOutputDir: pipeline now deletes the bounded list
   of known per-run files at the start of each run. Before this,
   a prior run's rejected-findings.json could linger when the
   current run had no rejections — confused J during the bug hunt
   above. cleanOutputDir is bounded (deletes only files we emit)
   so operator-owned adjacent files stay.

Verified end-to-end: insecure-repo + --enable-llm → 25 confirmed
findings (16 static + 9 LLM), 0 rejected.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-30 01:24:02 -05:00

226 lines
6.9 KiB
Go

package validators
import (
"os"
"path/filepath"
"testing"
"local-review-harness/internal/analyzers"
)
// makeFile is a tiny helper for test setup: write content to a path
// under tmp + return the abs path.
func makeFile(t *testing.T, dir, name, content string) string {
t.Helper()
p := filepath.Join(dir, name)
if err := os.MkdirAll(filepath.Dir(p), 0o755); err != nil {
t.Fatal(err)
}
if err := os.WriteFile(p, []byte(content), 0o644); err != nil {
t.Fatal(err)
}
return p
}
// llmFinding builds a minimal Source=LLM finding.
func llmFinding(file, lineHint, evidence string) analyzers.Finding {
return analyzers.Finding{
Title: "test",
Severity: analyzers.SeverityHigh,
Status: analyzers.StatusSuspected,
File: file,
LineHint: lineHint,
Evidence: evidence,
Reason: "test reason",
Source: analyzers.SourceLLM,
}
}
// === GATE D1 — synthetic LLM finding citing nonexistent file → must reject ===
func TestValidate_RejectsNonexistentFile(t *testing.T) {
repo := t.TempDir()
out := Validate(repo, []analyzers.Finding{
llmFinding("does/not/exist.go", "42", "fmt.Println"),
})
if len(out.Rejected) != 1 {
t.Fatalf("expected 1 rejected finding, got %d", len(out.Rejected))
}
if out.Rejected[0].RejectionReason != ReasonFileNotFound {
t.Errorf("expected ReasonFileNotFound, got %q", out.Rejected[0].RejectionReason)
}
}
func TestValidate_RejectsEvidenceNotInFile(t *testing.T) {
repo := t.TempDir()
makeFile(t, repo, "real.go", "package main\nfunc main() {}\n")
out := Validate(repo, []analyzers.Finding{
llmFinding("real.go", "1", "this string does not exist in the file"),
})
if len(out.Rejected) != 1 {
t.Fatalf("expected 1 rejected, got %d", len(out.Rejected))
}
if out.Rejected[0].RejectionReason != ReasonNoEvidence {
t.Errorf("expected ReasonNoEvidence, got %q", out.Rejected[0].RejectionReason)
}
}
func TestValidate_RejectsLineHintBeyondFile(t *testing.T) {
repo := t.TempDir()
makeFile(t, repo, "small.go", "line one\nline two\n") // 2 lines
out := Validate(repo, []analyzers.Finding{
llmFinding("small.go", "100", "line one"),
})
if len(out.Rejected) != 1 {
t.Fatalf("expected 1 rejected, got %d", len(out.Rejected))
}
if out.Rejected[0].RejectionReason != ReasonLineHintTooHigh {
t.Errorf("expected ReasonLineHintTooHigh, got %q", out.Rejected[0].RejectionReason)
}
}
func TestValidate_AcceptsRealFinding(t *testing.T) {
repo := t.TempDir()
makeFile(t, repo, "good.go", "package main\nfunc badPattern() {}\n")
out := Validate(repo, []analyzers.Finding{
llmFinding("good.go", "2", "func badPattern()"),
})
if len(out.Validated) != 1 {
t.Fatalf("expected 1 validated, got %d (rejected=%d)", len(out.Validated), len(out.Rejected))
}
if out.Validated[0].Status != analyzers.StatusConfirmed {
t.Errorf("expected status=confirmed, got %q", out.Validated[0].Status)
}
}
func TestValidate_AcceptsEvidenceWithDifferentLeadingWhitespace(t *testing.T) {
// Models often re-indent code when quoting; the validator's
// trim-line-by-line fallback should accept it.
repo := t.TempDir()
makeFile(t, repo, "indented.go", "package main\n\n\tfunc indented() {\n\t\treturn\n\t}\n")
out := Validate(repo, []analyzers.Finding{
llmFinding("indented.go", "3", "func indented() {"), // model dropped leading tab
})
if len(out.Validated) != 1 {
t.Fatalf("expected 1 validated; got rejected=%d (reason=%q)",
len(out.Rejected),
func() Reason {
if len(out.Rejected) > 0 {
return out.Rejected[0].RejectionReason
}
return ""
}(),
)
}
}
func TestValidate_RejectsEmptyEvidence(t *testing.T) {
repo := t.TempDir()
makeFile(t, repo, "any.go", "package main\n")
out := Validate(repo, []analyzers.Finding{
llmFinding("any.go", "1", ""),
})
if len(out.Rejected) != 1 || out.Rejected[0].RejectionReason != ReasonEmptyEvidence {
t.Errorf("empty-evidence finding should be rejected with ReasonEmptyEvidence; got %+v", out.Rejected)
}
}
func TestValidate_PassesThroughStaticFindings(t *testing.T) {
// Static findings already have grep-able evidence by construction.
// Validator promotes them to confirmed without re-checking.
repo := t.TempDir()
staticF := analyzers.Finding{
Title: "static finding",
Severity: analyzers.SeverityMedium,
Status: analyzers.StatusSuspected,
File: "anywhere.go",
Evidence: "anything",
Source: analyzers.SourceStatic,
}
out := Validate(repo, []analyzers.Finding{staticF})
if len(out.Validated) != 1 {
t.Fatalf("static finding should pass through validated; got %d", len(out.Validated))
}
if out.Validated[0].Status != analyzers.StatusConfirmed {
t.Errorf("static finding should be promoted to confirmed; got %q", out.Validated[0].Status)
}
}
// TestValidate_AcceptsRelativeRepoPath locks in the fix for the
// 2026-04-30 bug where every real finding was rejected as
// file_not_found because the path-traversal check compared a
// relative joined path against an absolute repoAbs (HasPrefix
// always false). Caught by J running ./review-harness with a
// relative target path; gate D1 now exercises this path.
func TestValidate_AcceptsRelativeRepoPath(t *testing.T) {
repo := t.TempDir()
makeFile(t, repo, "src/handler.go", "package main\nfunc bad() {}\n")
// Pass the repo as a RELATIVE path. Pre-fix this triggered the bug.
cwd, _ := os.Getwd()
defer os.Chdir(cwd)
parent := filepath.Dir(repo)
relRepo, err := filepath.Rel(parent, repo)
if err != nil {
t.Skip("can't compute relative path: " + err.Error())
}
if err := os.Chdir(parent); err != nil {
t.Fatal(err)
}
out := Validate(relRepo, []analyzers.Finding{
llmFinding("src/handler.go", "2", "func bad()"),
})
if len(out.Validated) != 1 {
t.Errorf("relative repo path should still validate; got rejected=%d (reasons: %v)",
len(out.Rejected),
func() []Reason {
rs := []Reason{}
for _, r := range out.Rejected {
rs = append(rs, r.RejectionReason)
}
return rs
}())
}
}
// === Path-traversal protection: hallucinated "../../../etc/passwd" must reject ===
func TestValidate_RejectsPathEscapingRepo(t *testing.T) {
repo := t.TempDir()
out := Validate(repo, []analyzers.Finding{
llmFinding("../../../etc/passwd", "1", "root:x:0:0"),
})
if len(out.Rejected) != 1 {
t.Fatalf("expected path-traversal finding rejected; got %d rejected", len(out.Rejected))
}
if out.Rejected[0].RejectionReason != ReasonFileNotFound {
t.Errorf("expected ReasonFileNotFound for path escape, got %q", out.Rejected[0].RejectionReason)
}
}
// === highestLine extractor ===
func TestHighestLine(t *testing.T) {
cases := []struct {
hint string
want int
ok bool
}{
{"42", 42, true},
{"10-20", 20, true},
{"line 100", 100, true},
{"L42", 42, true},
{"42:5", 42, true},
{"", 0, false},
{"none", 0, false},
{"15-30-50", 50, true}, // pick the largest
}
for _, c := range cases {
got, ok := highestLine(c.hint)
if got != c.want || ok != c.ok {
t.Errorf("highestLine(%q) = (%d, %v); want (%d, %v)", c.hint, got, ok, c.want, c.ok)
}
}
}