Make deep backfill's Phase C (orphan-contact discovery) opt-in

README.md

@@ -168,6 +168,7 @@ The macOS app target lives under `OpenMessage/`.
 | `OPENMESSAGES_HOST` | `127.0.0.1` | Host/interface to bind the local web server to |
 | `OPENMESSAGES_MY_NAME` | system user name | Display name for outgoing imported iMessage/WhatsApp messages |
 | `OPENMESSAGES_STARTUP_BACKFILL` | `auto` | Startup history sync mode: `auto`, `shallow`, `deep`, or `off` |
+| `OPENMESSAGES_BACKFILL_DISCOVER_ORPHANS` | `0` | Opt in to deep backfill's Phase C (contact-based orphan discovery). **Off by default** because it creates an empty SMS thread on your phone for each contact without prior message history. Enable with `1`/`true`/`yes`/`on` only if you understand the side effect. |
 | `OPENMESSAGES_MACOS_NOTIFICATIONS` | interactive macOS `serve` sessions only | Enable/disable native macOS notifications for fresh inbound live messages (`1`/`0`). Click-through opens the matching thread when `terminal-notifier` is available. |
 | `OPENMESSAGE_TELEMETRY` | unset (off) | Set to `1` to send one anonymous heartbeat per launch (max one per 24h). Reports only: random install ID, version, OS/arch, and which platforms are paired (Google Messages / WhatsApp / Signal). No message content, no contact info, no IP-based identity. See `internal/telemetry/`. |
 

internal/app/backfill.go

@@ -4,6 +4,7 @@ import (
 	"encoding/hex"
 	"encoding/json"
 	"fmt"
+	"os"
 	"strings"
 	"time"
 
@@ -19,6 +20,26 @@ const (
 	recentReconcileMaxPages          = 4
 )
 
+// orphanContactDiscoveryEnabled reports whether deep backfill should run
+// Phase C (contact-based orphan discovery).
+//
+// Phase C calls GetOrCreateConversation for every contact without prior
+// message history. Google Messages treats GetOrCreateConversation as a
+// thread-creation call: for each contact that has no existing thread, an
+// empty SMS thread is created on the user's phone. For users who only want
+// a deep history sync, that is an unwanted side effect.
+//
+// Phase C is therefore opt-in. Set OPENMESSAGES_BACKFILL_DISCOVER_ORPHANS=1
+// (or "true"/"yes"/"on") to enable. Default is disabled.
+func orphanContactDiscoveryEnabled() bool {
+	switch strings.ToLower(strings.TrimSpace(os.Getenv("OPENMESSAGES_BACKFILL_DISCOVER_ORPHANS"))) {
+	case "1", "true", "yes", "on":
+		return true
+	default:
+		return false
+	}
+}
+
 // Backfill fetches existing conversations and recent messages from
 // Google Messages and stores them in the local database.
 func (a *App) Backfill() error {
@@ -125,12 +146,20 @@ func (a *App) deepBackfill() {
 		}
 	}
 
-	// Phase C: Contact-based discovery for orphan phone numbers
-	a.BackfillProgress.setPhase(BackfillPhaseContacts)
-	if a.discoverFromContacts(gm, seen, clientToken) {
-		a.emitConversationsChange()
-		a.emitMessagesChange("")
-		return
+	// Phase C: Contact-based discovery for orphan phone numbers.
+	// Off by default because GetOrCreateConversation creates an empty SMS
+	// thread on the user's phone for each contact lacking one. Opt in via
+	// OPENMESSAGES_BACKFILL_DISCOVER_ORPHANS=1.
+	if orphanContactDiscoveryEnabled() {
+		a.BackfillProgress.setPhase(BackfillPhaseContacts)
+		if a.discoverFromContacts(gm, seen, clientToken) {
+			a.emitConversationsChange()
+			a.emitMessagesChange("")
+			return
+		}
+	} else {
+		a.Logger.Info().
+			Msg("Skipping Phase C (orphan-contact discovery); set OPENMESSAGES_BACKFILL_DISCOVER_ORPHANS=1 to enable. Note: enabling creates empty SMS threads for contacts without prior message history.")
 	}
 
 	progress := a.BackfillProgress.snapshot()

internal/app/backfill_test.go

@@ -335,6 +335,7 @@ func TestDeepBackfillMessagePagination(t *testing.T) {
 }
 
 func TestDeepBackfillContactDiscovery(t *testing.T) {
+	t.Setenv("OPENMESSAGES_BACKFILL_DISCOVER_ORPHANS", "1")
 	mock := &mockGMClient{
 		conversations: map[gmproto.ListConversationsRequest_Folder][][]*gmproto.Conversation{
 			gmproto.ListConversationsRequest_INBOX: {
@@ -371,6 +372,7 @@ func TestDeepBackfillContactDiscovery(t *testing.T) {
 }
 
 func TestDeepBackfillContactDiscoverySkipsAlreadySeen(t *testing.T) {
+	t.Setenv("OPENMESSAGES_BACKFILL_DISCOVER_ORPHANS", "1")
 	// Contact's phone maps to a conversation already found in INBOX
 	mock := &mockGMClient{
 		conversations: map[gmproto.ListConversationsRequest_Folder][][]*gmproto.Conversation{
@@ -401,6 +403,48 @@ func TestDeepBackfillContactDiscoverySkipsAlreadySeen(t *testing.T) {
 	}
 }
 
+// TestDeepBackfillSkipsPhaseCWhenOptOut confirms Phase C does NOT run when
+// the env var is unset. The mock has a contact whose phone would map to a
+// brand-new conversation; without orphan discovery enabled, that contact
+// must not be queried via GetOrCreateConversation, so the conversation
+// must not appear in the store.
+func TestDeepBackfillSkipsPhaseCWhenOptOut(t *testing.T) {
+	// Explicitly empty (default behavior).
+	t.Setenv("OPENMESSAGES_BACKFILL_DISCOVER_ORPHANS", "")
+	mock := &mockGMClient{
+		conversations: map[gmproto.ListConversationsRequest_Folder][][]*gmproto.Conversation{
+			gmproto.ListConversationsRequest_INBOX: {
+				{makeConv("c1", "Alice")},
+			},
+		},
+		messages: map[string][][]*gmproto.Message{
+			"c1":     {{makeMsg("m1", "c1", "hi", 100)}},
+			"c-mary": {{makeMsg("m2", "c-mary", "old msg", 50)}},
+		},
+		contacts: []*gmproto.Contact{
+			{
+				Name:   "Mary",
+				Number: &gmproto.ContactNumber{Number: "+15555555555"},
+			},
+		},
+		getOrCreateResults: map[string]*gmproto.Conversation{
+			"+15555555555": makeConv("c-mary", "Mary"),
+		},
+	}
+
+	a := newTestApp(t, mock)
+	a.DeepBackfill()
+
+	convos, _ := a.Store.ListConversations(50)
+	if len(convos) != 1 {
+		t.Fatalf("got %d conversations, want 1 (Phase C should be skipped)", len(convos))
+	}
+	progress := a.BackfillProgress.snapshot()
+	if progress.ContactsChecked != 0 {
+		t.Fatalf("got %d contacts checked, want 0 (Phase C should be skipped)", progress.ContactsChecked)
+	}
+}
+
 func TestDeepBackfillFolderListError(t *testing.T) {
 	mock := &mockGMClient{
 		conversations: map[gmproto.ListConversationsRequest_Folder][][]*gmproto.Conversation{
@@ -463,6 +507,7 @@ func TestDeepBackfillMessageFetchError(t *testing.T) {
 }
 
 func TestDeepBackfillGetOrCreateError(t *testing.T) {
+	t.Setenv("OPENMESSAGES_BACKFILL_DISCOVER_ORPHANS", "1")
 	mock := &mockGMClient{
 		contacts: []*gmproto.Contact{
 			{
@@ -1008,3 +1053,34 @@ func TestBackfillPopulatesDB(t *testing.T) {
 		t.Fatalf("got body %q", msgs[0].Body)
 	}
 }
+
+
+func TestOrphanContactDiscoveryEnabled(t *testing.T) {
+	cases := []struct {
+		name string
+		env  string
+		want bool
+	}{
+		{"unset", "", false},
+		{"empty", "", false},
+		{"explicit zero", "0", false},
+		{"explicit false", "false", false},
+		{"no", "no", false},
+		{"off", "off", false},
+		{"one", "1", true},
+		{"true", "true", true},
+		{"True mixed case", "True", true},
+		{"YES upper", "YES", true},
+		{"on padded", "  on  ", true},
+		{"unrecognized", "maybe", false},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Setenv("OPENMESSAGES_BACKFILL_DISCOVER_ORPHANS", tc.env)
+			if got := orphanContactDiscoveryEnabled(); got != tc.want {
+				t.Fatalf("orphanContactDiscoveryEnabled() = %v with env=%q, want %v", got, tc.env, tc.want)
+			}
+		})
+	}
+}