felhom-controller/controller/internal/web/recovery_test.go

package web

import (
	"context"
	"encoding/json"
	"io"
	"log"
	"net/http"
	"net/http/httptest"
	"net/url"
	"os"
	"path/filepath"
	"strings"
	"sync"
	"testing"
	"time"

	"gitea.dooplex.hu/admin/felhom-controller/internal/agentapi"
	"gitea.dooplex.hu/admin/felhom-controller/internal/backup"
	"gitea.dooplex.hu/admin/felhom-controller/internal/config"
	"gitea.dooplex.hu/admin/felhom-controller/internal/settings"
	"gitea.dooplex.hu/admin/felhom-controller/internal/stacks"
)

// R-193 — the recovery screen. HANDLER-LEVEL tests throughout: a test that reaches a helper while the
// mutation lives in the handler cannot observe it, which is how a red-proof passed three sessions ago.
// Everything below drives the real handler (or the real mux) and asserts the rendered page, the
// on-disk effect, or the absence of the code.

const testRecoveryCode = "helyre-allitasi-kod-tiz-szo-pontosan-igy-ni-most"
const testRepoPW = "b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1b1"

// fakeRecoverer is the agent seam. It records every code it was handed so a test can prove the
// handler passed the RIGHT one, and can fail on demand for the wrong-code path.
type fakeRecoverer struct {
	mu    sync.Mutex
	pw    string
	sha   string
	fail  bool
	// failWith overrides `fail` with an EXACT error value, so a test can drive one specific
	// R-224 class (a 502 fetch failure, a transport error, an unrecognised status).
	failWith error
	codes    []string
}

func (f *fakeRecoverer) RecoverOffsiteRepoPassword(_ context.Context, code string) (string, string, error) {
	f.mu.Lock()
	f.codes = append(f.codes, code)
	f.mu.Unlock()
	if f.failWith != nil {
		return "", "", f.failWith
	}
	if f.fail {
		// R-224: the shape agent >= v0.126.0 returns for a code that was TRIED AND REFUSED — HTTP 400,
		// meaning the bundle was fetched and `age -d` rejected it. It used to be a bare error here,
		// which is the shape of a failure we could NOT classify; under the R-224 rule that now renders
		// the neutral message, and rightly so. Saying "wrong code" in a test requires saying it the way
		// the agent says it.
		return "", "", &agentapi.RecoveryRefusal{Status: 400, Reason: "the recovery code did not open the sealed bundle — nothing was written"}
	}
	return f.pw, f.sha, nil
}

// recoveryRunner is the restic seam for the post-unlock inventory.
type recoveryRunner struct {
	snapshots []map[string]any
	statsSize int64
}

func (rr *recoveryRunner) run(_ context.Context, _ []string, args ...string) ([]byte, error) {
	joined := strings.Join(args, " ")
	switch {
	case strings.Contains(joined, " snapshots"):
		b, _ := json.Marshal(rr.snapshots)
		return b, nil
	case strings.Contains(joined, " stats "):
		b, _ := json.Marshal(map[string]any{"total_size": rr.statsSize})
		return b, nil
	}
	return []byte(""), nil
}

type recoveryFixture struct {
	s       *Server
	mgr     *backup.Manager
	sett    *settings.Settings
	rec     *fakeRecoverer
	runner  *recoveryRunner
	dataDir string
}

// newRecoveryFixture builds a Server in the REBUILT-BOX shape by default: a claimed box (password
// set), no repository password on disk, and the hub holding a sealed package.
func newRecoveryFixture(t *testing.T) *recoveryFixture {
	t.Helper()
	lg := log.New(io.Discard, "", 0)
	dir := t.TempDir()
	cfg := &config.Config{}
	cfg.Paths.DataDir = filepath.Join(dir, "data")
	cfg.Paths.SystemDataPath = filepath.Join(dir, "sys")
	cfg.Paths.StacksDir = filepath.Join(dir, "stacks")
	cfg.Web.SessionSecret = "test-session-secret-abcdef"
	cfg.Web.PasswordHash = "$2a$10$abcdefghijklmnopqrstuv" // claimed: auth is enabled

	sett, err := settings.Load(filepath.Join(dir, "settings.json"), lg)
	if err != nil {
		t.Fatal(err)
	}
	if err := sett.SetHubEscrowIdentityPresent(true); err != nil { // the hub holds a package
		t.Fatal(err)
	}
	// The realistic shape once the credential self-heal has re-applied the tier: coordinates exist,
	// but this box holds no repository password for the history they point at.
	if err := sett.SetOffboxTarget(&settings.OffboxTarget{
		Enabled: true, Host: "nas.local", Port: 22, User: "felhom", RepoPath: "/srv/repo",
		Schedule: "daily", EscrowState: "escrowed",
	}); err != nil {
		t.Fatal(err)
	}
	mgr := backup.NewManager(cfg, sett, lg)
	if err := mgr.WriteOffboxSecrets("PRIVATE-KEY-MATERIAL", "nas.local ssh-ed25519 AAAAhostkey"); err != nil {
		t.Fatal(err)
	}
	// WriteOffboxSecrets auto-generates a repository password — remove it, because "this box cannot
	// open the inherited history" is the whole precondition of the screen.
	if err := os.Remove(filepath.Join(cfg.Paths.DataDir, "offbox", "repo_password")); err != nil {
		t.Fatal(err)
	}
	rr := &recoveryRunner{statsSize: 4 << 20}
	mgr.SetOffboxRunner(rr.run)

	rec := &fakeRecoverer{pw: testRepoPW, sha: backup.HashResticPassword(testRepoPW)}
	stackMgr, serr := stacks.NewManager(cfg, lg)
	if serr != nil {
		t.Fatal(serr)
	}
	s := &Server{cfg: cfg, settings: sett, backupMgr: mgr, stackMgr: stackMgr, logger: lg, version: "test"}
	s.loadTemplates()
	s.SetRecoveryRecoverer(func() (backup.OffsiteKeyRecoverer, error) { return rec, nil })
	// R-216: the capability gate FAILS CLOSED, so the default fixture states the supported case
	// explicitly. Without this every unlock test would exercise the refusal instead — which is exactly
	// the protection working, and exactly not what those tests are about. The refusal has its own
	// tests in recovery_gate_test.go, each overriding this.
	s.SetRecoverySupport(func(context.Context) agentapi.SupportState { return agentapi.SupportYes })
	// R-224: the fixture's agent is a current one, so a 400 may be read as a genuine refusal. Tests
	// that need the OLD-agent behaviour override this explicitly.
	s.SetRecoveryRefusalTrusted(func(context.Context) bool { return true })
	return &recoveryFixture{s: s, mgr: mgr, sett: sett, rec: rec, runner: rr, dataDir: cfg.Paths.DataDir}
}

// placeRepoPassword makes the box look HEALTHY (it holds its own repository password).
func (f *recoveryFixture) placeRepoPassword(t *testing.T) {
	t.Helper()
	if err := f.mgr.InjectOffboxPassword(testRepoPW, true); err != nil {
		t.Fatal(err)
	}
}

func getRecoveryPage(t *testing.T, s *Server) *httptest.ResponseRecorder {
	t.Helper()
	rr := httptest.NewRecorder()
	s.recoveryPageHandler(rr, httptest.NewRequest(http.MethodGet, "/recovery", nil))
	return rr
}

// SCENARIO A — the page appears for the fresh + package box, and interrupts the landing pages.
func TestRecovery_A_PageAppearsForARebuiltBox(t *testing.T) {
	f := newRecoveryFixture(t)

	if !f.s.recoveryOffer() {
		t.Fatal("a rebuilt box (fresh data area + a hub-held package) is not offered the recovery screen")
	}
	if !f.s.recoveryInterrupts() {
		t.Fatal("the full page must interrupt the landing pages before any postpone")
	}
	rr := getRecoveryPage(t, f.s)
	if rr.Code != http.StatusOK {
		t.Fatalf("GET /recovery = %d", rr.Code)
	}
	body := rr.Body.String()
	// The two MANDATORY sentences of §8.2 (ASCII-safe fragments — accented patterns get mangled
	// through the ssh→pct chain and a false 0 reads exactly like the sentence being gone).
	if !strings.Contains(body, "helyre") || !strings.Contains(body, "llít") {
		t.Error("the page does not mention the recovery code at all")
	}
	if !strings.Contains(body, "senki nem tudja p") {
		t.Error("MANDATORY: the page must say that nobody can replace a lost recovery code")
	}
	if !strings.Contains(body, "semmi nem v") {
		t.Error("MANDATORY: the page must say that nothing is restored or changed in this step")
	}
	// It takes the code in a POST body, and the field does not autofill.
	if !strings.Contains(body, `action="/recovery/unlock"`) || !strings.Contains(body, `method="POST"`) {
		t.Error("the code form must POST to /recovery/unlock")
	}
	if !strings.Contains(body, `autocomplete="off"`) {
		t.Error("the recovery-code field must not autofill")
	}
	// And it is not cached.
	if cc := rr.Header().Get("Cache-Control"); !strings.Contains(cc, "no-store") {
		t.Errorf("the recovery page must be no-store, got %q", cc)
	}
}

// SCENARIO B — it does NOT appear for anyone else. THE GUARD ON THE CONJUNCTION.
//
// RED-PROOF: drop the hub-package condition from backup.OffsiteRecoveryOffer → the
// "never had off-site backups" case below FAILS, i.e. a brand-new customer is greeted on day one by
// a recovery screen for data they never had. That is the plausible wrong fix.
func TestRecovery_B_DoesNotAppearForAnyoneElse(t *testing.T) {
	t.Run("healthy box (holds its own repository password)", func(t *testing.T) {
		f := newRecoveryFixture(t)
		f.placeRepoPassword(t) // healthy, and not orphaned
		if f.s.recoveryOffer() {
			t.Fatal("a HEALTHY box was offered the recovery screen")
		}
	})
	t.Run("never had off-site backups (no hub package)", func(t *testing.T) {
		f := newRecoveryFixture(t)
		if err := f.sett.SetHubEscrowIdentityPresent(false); err != nil {
			t.Fatal(err)
		}
		if f.s.recoveryOffer() {
			t.Fatal("a box that never had off-site backups was offered a recovery screen for data it never had")
		}
	})
	t.Run("not claimed — the page is behind the household password", func(t *testing.T) {
		f := newRecoveryFixture(t)
		f.s.cfg.Web.PasswordHash = "" // unclaimed: no password anywhere
		if f.s.authEnabled() {
			t.Fatal("fixture error: the box still reads as claimed")
		}
		// The interception is inside the authenticated surface: RequireAuth gates /launcher and
		// /dashboard before ServeHTTP ever runs. Assert that through the REAL middleware chain.
		mux := http.NewServeMux()
		mux.Handle("/", f.s.RequireAuth(f.s.CsrfProtect(http.HandlerFunc(f.s.ServeHTTP))))
		rr := httptest.NewRecorder()
		mux.ServeHTTP(rr, httptest.NewRequest(http.MethodGet, "/launcher", nil))
		if rr.Code == http.StatusFound && rr.Header().Get("Location") == "/recovery" {
			t.Fatal("an UNCLAIMED box redirected to the recovery screen — it shows metadata that belongs behind the household password")
		}
	})
}

// SCENARIO C — the correct code unlocks, places the key, and the page then shows what is in there.
func TestRecovery_C_UnlockOpensAndLists(t *testing.T) {
	f := newRecoveryFixture(t)
	now := time.Now().UTC()
	f.runner.snapshots = []map[string]any{
		{"short_id": "aaa1111", "time": now.Add(-24 * time.Hour).Format(time.RFC3339), "tags": []string{"immich"}},
		{"short_id": "bbb2222", "time": now.Format(time.RFC3339), "tags": []string{"immich"}},
		{"short_id": "ccc3333", "time": now.Add(-48 * time.Hour).Format(time.RFC3339), "tags": []string{"calibre-web"}},
	}

	rr := postUnlock(t, f.s, testRecoveryCode)
	if rr.Code != http.StatusOK {
		t.Fatalf("unlock = %d", rr.Code)
	}
	// EFFECT 1: the recovered key is on disk.
	got, present := f.mgr.OffboxRepoPasswordHash()
	if !present || got != backup.HashResticPassword(testRepoPW) {
		t.Fatalf("the recovered repository password was not placed (present=%v)", present)
	}
	// EFFECT 2: the handler passed the code it was given, unmodified.
	if len(f.rec.codes) != 1 || f.rec.codes[0] != testRecoveryCode {
		t.Fatalf("the handler did not hand the agent the typed code: %+v", f.rec.codes)
	}
	// EFFECT 3: the page lists what is in there — apps and dates. A success message with nothing
	// shown is indistinguishable from having unlocked an EMPTY store.
	body := rr.Body.String()
	for _, want := range []string{"immich", "calibre-web"} {
		if !strings.Contains(body, want) {
			t.Errorf("the listing does not name %q: the customer cannot tell whether this is their data", want)
		}
	}
	if !strings.Contains(body, "4.0 MB") && !strings.Contains(body, "MB") {
		t.Errorf("the listing shows no size")
	}
	// EFFECT 4: it did NOT restore anything — the page points at the restore page rather than doing it.
	if !strings.Contains(body, "/backups/restore") {
		t.Error("the page must point at the per-app restore rather than restoring")
	}
	if strings.Contains(body, "/backup/offbox/reconstitute") || strings.Contains(body, "/backup/offbox/place") {
		t.Fatal("the recovery page offers a RESTORE action — unlocking and restoring are separate")
	}
}

// The EMPTY store is stated plainly rather than shown as a bare list.
func TestRecovery_C_EmptyStoreSaysSo(t *testing.T) {
	f := newRecoveryFixture(t)
	f.runner.snapshots = nil // opened cleanly, holds nothing

	body := postUnlock(t, f.s, testRecoveryCode).Body.String()
	if !strings.Contains(body, "nincs benne egyetlen ment") {
		t.Fatalf("an empty store must say so plainly — silence there reads as a broken page. body=%.400q", body)
	}
}

// SCENARIO D — a wrong code fails closed, writes nothing, says what to check, and does NOT lock out.
func TestRecovery_D_WrongCodeFailsClosedAndIsKind(t *testing.T) {
	f := newRecoveryFixture(t)
	f.rec.fail = true

	for i := 0; i < 6; i++ { // well past any plausible lockout threshold
		rr := postUnlock(t, f.s, "rossz kod")
		if rr.Code != http.StatusOK {
			t.Fatalf("attempt %d: got %d, want a re-rendered page", i, rr.Code)
		}
		body := rr.Body.String()
		if !strings.Contains(body, "nem fogadtuk el") {
			t.Fatalf("attempt %d: the page does not say the code was not accepted: %.300q", i, body)
		}
		if !strings.Contains(body, "z szót") && !strings.Contains(body, "t sz") {
			t.Errorf("attempt %d: the message does not say what to check", i)
		}
		// The raw agent error must NOT be shown to the customer.
		if strings.Contains(body, "age:") || strings.Contains(body, "incorrect passphrase") {
			t.Errorf("attempt %d: the raw technical error was rendered to the customer", i)
		}
		// NOTHING was written.
		if _, present := f.mgr.OffboxRepoPasswordHash(); present {
			t.Fatalf("attempt %d: a repository password was written on a FAILED unlock", i)
		}
		// And the form is still there — no lockout.
		if !strings.Contains(body, `name="recovery_code"`) {
			t.Fatalf("attempt %d: the customer was locked out of their own data after a mistyped code", i)
		}
	}
}

// SCENARIO E — "most nem" stops the interruption and NOTHING else. The entry point survives.
//
// RED-PROOF: bind the backups-page entry point to recoveryInterrupts instead of recoveryOffer → the
// route to the data disappears after one click.
func TestRecovery_E_PostponeKeepsTheEntryPoint(t *testing.T) {
	f := newRecoveryFixture(t)

	rr := httptest.NewRecorder()
	f.s.recoveryPostponeHandler(rr, httptest.NewRequest(http.MethodPost, "/recovery/postpone", nil))
	if rr.Code != http.StatusFound {
		t.Fatalf("postpone = %d, want a redirect", rr.Code)
	}
	if !f.sett.GetRecoveryNoticePostponed() {
		t.Fatal("the postpone was not recorded")
	}
	// The full page no longer interrupts…
	if f.s.recoveryInterrupts() {
		t.Fatal("the full page still interrupts after 'most nem'")
	}
	// …and it must NOT interrupt through the real mux either.
	mux := http.NewServeMux()
	mux.Handle("/", http.HandlerFunc(f.s.ServeHTTP))
	rr2 := httptest.NewRecorder()
	mux.ServeHTTP(rr2, httptest.NewRequest(http.MethodGet, "/launcher", nil))
	if rr2.Code == http.StatusFound && rr2.Header().Get("Location") == "/recovery" {
		t.Fatal("the landing page still redirects to the recovery screen after 'most nem'")
	}
	// …but the ENTRY POINT is untouched: the offer stands, so the backups page still renders it.
	if !f.s.recoveryOffer() {
		t.Fatal("'most nem' removed the OFFER — the customer has lost the route to their own data")
	}
	// And the page itself is still reachable directly.
	if getRecoveryPage(t, f.s).Code != http.StatusOK {
		t.Fatal("the recovery page is unreachable after 'most nem'")
	}
	// THE ENTRY POINT ITSELF, rendered: the backups page must still carry the route. Asserted on the
	// rendered flag rather than on recoveryOffer alone, because the defect this guards against is the
	// TEMPLATE being fed the wrong predicate.
	rr3 := httptest.NewRecorder()
	f.s.backupsRemoteHandler(rr3, httptest.NewRequest(http.MethodGet, "/backups/remote", nil))
	if !strings.Contains(rr3.Body.String(), `href="/recovery"`) {
		t.Fatal("the backups page no longer offers the route to the recovery screen after 'most nem' — a customer who clicked past it once has lost the way to their own data")
	}
}

// SCENARIO F — "I do not want the old data" is confirmed TWICE and reaches the SHIPPED move-aside.
//
// RED-PROOF: render the final button on the first view (drop the ConfirmSetAside gate) → one click
// suffices, and this test fails on the first assertion.
func TestRecovery_F_SetAsideNeedsTwoConfirmations(t *testing.T) {
	f := newRecoveryFixture(t)
	// The move-aside only exists once the tier is orphaned — that is the shipped handler's own
	// precondition, and the page only offers the choice when it can actually run.
	if err := f.sett.SetOffboxTarget(&settings.OffboxTarget{
		Enabled: true, Host: "nas.local", Port: 22, User: "felhom", RepoPath: "/srv/repo",
		Schedule: "daily", EscrowState: "escrowed", RepoState: "orphaned",
	}); err != nil {
		t.Fatal(err)
	}
	if !f.mgr.OffboxOrphaned() {
		t.Fatal("fixture: the tier is not orphaned, so the set-aside cannot be offered")
	}

	// FIRST VIEW — the destructive-looking button must NOT be present yet.
	first := getRecoveryPage(t, f.s).Body.String()
	if strings.Contains(first, `action="/backup/offbox/reset"`) {
		t.Fatal("the set-aside form is on the FIRST view — one click would set the customer's history aside")
	}
	if !strings.Contains(first, "setaside=1") {
		t.Fatal("the first view offers no route to the set-aside choice at all")
	}
	// And it must read as the exceptional path, not an equal third button.
	if strings.Count(first, "btn-primary") > 1 {
		t.Error("the set-aside is styled as an equal primary action")
	}

	// SECOND VIEW — now the confirmation, naming exactly what happens.
	rr := httptest.NewRecorder()
	f.s.recoveryPageHandler(rr, httptest.NewRequest(http.MethodGet, "/recovery?setaside=1", nil))
	second := rr.Body.String()
	if !strings.Contains(second, `action="/backup/offbox/reset"`) {
		t.Fatal("the second view does not reach the SHIPPED move-aside handler")
	}
	if !strings.Contains(second, `name="confirm" value="1"`) {
		t.Fatal("the second view does not carry the shipped handler's confirm gate")
	}
	// The copy must say SET ASIDE, not delete — the whole point of the ruling.
	if !strings.Contains(second, "lretessz") {
		t.Error("the confirmation does not say the backups are SET ASIDE")
	}
	if !strings.Contains(second, "nem t") {
		t.Error("the confirmation does not say they are NOT deleted")
	}
}

// SCENARIO G — the command line and the page drive ONE core.
//
// RED-PROOF: give the handler its own copy of fetch→compare→install instead of calling
// RecoverInstallCore → this test still passes on the happy path, so it asserts the SHARED SYMBOL
// from source (below) as well as the behaviour here.
func TestRecovery_G_PageAndCLIShareOneCore(t *testing.T) {
	// Behavioural half: the same fake, the same outcome, through both callers.
	f := newRecoveryFixture(t)
	res, err := backup.RecoverInstallCore(context.Background(), f.mgr, f.rec, testRecoveryCode, true)
	if err != nil || res.Outcome != backup.RecoverInstalled {
		t.Fatalf("core install: outcome=%q err=%v", res.Outcome, err)
	}
	// Re-running is UNCHANGED, not a second install — the same three outcomes the CLI documents.
	res2, err2 := backup.RecoverInstallCore(context.Background(), f.mgr, f.rec, testRecoveryCode, true)
	if err2 != nil || res2.Outcome != backup.RecoverUnchanged {
		t.Fatalf("core re-run: outcome=%q err=%v", res2.Outcome, err2)
	}
}

// SCENARIO H (§8.3) — the recovery code persists NOWHERE, with the planted-copy positive control.
func TestRecovery_H_CodeLeavesNoTrace(t *testing.T) {
	f := newRecoveryFixture(t)
	var logBuf strings.Builder
	f.s.logger = log.New(&logBuf, "", 0)

	rr := postUnlock(t, f.s, testRecoveryCode)

	// 1) not echoed in the response
	if strings.Contains(rr.Body.String(), testRecoveryCode) {
		t.Fatal("the recovery code was echoed back in the rendered page")
	}
	// 2) not in any log line
	if strings.Contains(logBuf.String(), testRecoveryCode) {
		t.Fatal("the recovery code reached the log")
	}
	// 3) not in any file under the data dir
	found := grepTree(t, f.dataDir, testRecoveryCode)
	if found != "" {
		t.Fatalf("the recovery code was persisted to %s", found)
	}

	// THE POSITIVE CONTROL — a sweep whose sensitivity was never shown is not evidence. Plant a copy
	// where the sweep looks and require it to be found; then remove it.
	planted := filepath.Join(f.dataDir, "planted-control.txt")
	if err := os.WriteFile(planted, []byte("x "+testRecoveryCode+" x"), 0o600); err != nil {
		t.Fatal(err)
	}
	if got := grepTree(t, f.dataDir, testRecoveryCode); got == "" {
		t.Fatal("POSITIVE CONTROL FAILED: the sweep could not find a planted copy, so its earlier silence proves nothing")
	}
	if err := os.Remove(planted); err != nil {
		t.Fatal(err)
	}
	if got := grepTree(t, f.dataDir, testRecoveryCode); got != "" {
		t.Fatalf("the control was not cleaned up: %s", got)
	}
}

// grepTree returns the first file under root whose contents contain needle ("" when none).
func grepTree(t *testing.T, root, needle string) string {
	t.Helper()
	var hit string
	_ = filepath.Walk(root, func(p string, info os.FileInfo, err error) error {
		if err != nil || info == nil || info.IsDir() || hit != "" {
			return nil
		}
		b, rerr := os.ReadFile(p)
		if rerr == nil && strings.Contains(string(b), needle) {
			hit = p
		}
		return nil
	})
	return hit
}

func postUnlock(t *testing.T, s *Server, code string) *httptest.ResponseRecorder {
	t.Helper()
	form := url.Values{"recovery_code": {code}}
	req := httptest.NewRequest(http.MethodPost, "/recovery/unlock", strings.NewReader(form.Encode()))
	req.Header.Set("Content-Type", "application/x-www-form-urlencoded")
	rr := httptest.NewRecorder()
	s.recoveryUnlockHandler(rr, req)
	return rr
}