fix(fips): harden companion mesh joins

Android/COMPANION_RELEASE.md

@@ -23,6 +23,10 @@ used by the `.githooks/pre-push` hook), which:
    **aborts** if any is missing.
 5. Stages the signed APK at `neode-ui/public/packages/archipelago-companion.apk`,
    commits, and pushes with `SHIP_COMPANION=1` (the sanctioned pre-push bypass).
+6. The first-launch companion modal and Android "Share this app" QR point at
+   `http://146.59.87.168:2100/packages/archipelago-companion.apk`. After the
+   repo artifact is built, mirror that exact APK to the VPS2-served path before
+   calling the release done.
 
 **Never** hand-roll `gradlew assembleDebug` + `cp` to the served path. That path
 skips the clean build and the signature enforcement and is exactly how a broken
@@ -82,13 +86,16 @@ home-screen app layouts wiped by an over-broad action.
 
 ## Verify the published download after shipping
 
-The download served to nodes is Gitea raw-on-main. Confirm the live bytes match
-what you built and signed:
+The checked-in artifact is Gitea raw-on-main. The QR/App Store download served
+to users is the VPS2 `:2100` URL. Confirm both live byte streams match what you
+built and signed:
 
 ```bash
 SERVED=neode-ui/public/packages/archipelago-companion.apk
-URL=http://146.59.87.168:3000/lfg2025/archy/raw/branch/main/$SERVED
-curl -sS -o /tmp/live.apk "$URL"
-shasum -a 256 "$SERVED" /tmp/live.apk          # must match
-apksigner verify -v --min-sdk-version 21 /tmp/live.apk | grep -i "scheme"  # v1/v2/v3 = true
+GITEA_URL=http://146.59.87.168:3000/lfg2025/archy/raw/branch/main/$SERVED
+QR_URL=http://146.59.87.168:2100/packages/archipelago-companion.apk
+curl -sS -o /tmp/live-gitea.apk "$GITEA_URL"
+curl -sS -o /tmp/live-qr.apk "$QR_URL"
+shasum -a 256 "$SERVED" /tmp/live-gitea.apk /tmp/live-qr.apk  # all must match
+apksigner verify -v --min-sdk-version 21 /tmp/live-qr.apk | grep -i "scheme"  # v1/v2/v3 = true
 ```

Android/app/build.gradle.kts

@@ -11,8 +11,8 @@ android {
         applicationId = "com.archipelago.app"
         minSdk = 26
         targetSdk = 35
-        versionCode = 35
-        versionName = "0.5.15"
+        versionCode = 38
+        versionName = "0.5.18"
 
         vectorDrawables {
             useSupportLibrary = true

Android/app/src/main/java/com/archipelago/app/fips/FipsPreferences.kt

@@ -25,6 +25,26 @@ internal const val ARCHY_ANCHOR_NPUB =
 internal const val ARCHY_ANCHOR_ADDR = "146.59.87.168:8444"
 internal const val ARCHY_ANCHOR_TRANSPORT = "tcp"
 
+/**
+ * Public FIPS network anchors (join.fips.network — the dual-transport TCP
+ * pair; keep in lockstep with core/archipelago/src/fips/anchors.rs
+ * fips_network_anchors()). Baked into every pairing at trailing priority so
+ * a degraded/unreachable vps2 anchor can never strand the phone: the mesh
+ * still joins the public tree and routes to the node through it.
+ */
+internal val PUBLIC_FIPS_ANCHORS = listOf(
+    Triple(
+        "npub10yffd020a4ag8zcy75f9pruq3rnghvvhd5hphl9s62zgp35s560qrksp9u",
+        "23.182.128.74:443",
+        "tcp",
+    ),
+    Triple(
+        "npub1qmc3cvfz0yu2hx96nq3gp55zdan2qclealn7xshgr448d3nh6lks7zel98",
+        "217.77.8.91:443",
+        "tcp",
+    ),
+)
+
 /**
  * Mesh identity + known node peers. Follows the same plaintext-DataStore
  * storage model as ServerPreferences (the server password lives there the
@@ -153,9 +173,40 @@ class FipsPreferences(private val context: Context) {
                 }))
             })
         }
+        // Backfill anchor peers for entries paired before newer QR/app
+        // releases added them. `upsertNodePeer` persists these on re-scan, but
+        // startup must also self-heal old DataStore state so updating the APK is
+        // enough to get off-LAN redundancy.
+        if (merged.length() > 0) {
+            addAnchorIfMissing(merged, ARCHY_ANCHOR_NPUB, "archipelago-anchor", ARCHY_ANCHOR_ADDR, ARCHY_ANCHOR_TRANSPORT, 40)
+            for ((i, anchor) in PUBLIC_FIPS_ANCHORS.withIndex()) {
+                val (npub, addr, transport) = anchor
+                addAnchorIfMissing(merged, npub, "fips-network-anchor-${i + 1}", addr, transport, 50 + i * 10)
+            }
+        }
         return merged.toString()
     }
 
+    private fun addAnchorIfMissing(
+        peers: JSONArray,
+        npub: String,
+        alias: String,
+        addr: String,
+        transport: String,
+        priority: Int,
+    ) {
+        if ((0 until peers.length()).any { peers.optJSONObject(it)?.optString("npub") == npub }) return
+        peers.put(JSONObject().apply {
+            put("npub", npub)
+            put("alias", alias)
+            put("addresses", JSONArray().put(JSONObject().apply {
+                put("transport", transport)
+                put("addr", addr)
+                put("priority", priority)
+            }))
+        })
+    }
+
     private fun parsePartyPeers(json: String): List<PartyPeer> = try {
         val arr = JSONArray(json)
         (0 until arr.length()).mapNotNull { i ->
@@ -249,6 +300,21 @@ class FipsPreferences(private val context: Context) {
                 }))
             }
         }
+        // And the public FIPS network anchors at trailing priority, so one
+        // degraded rendezvous (vps2, 2026-07-24) can never strand the phone.
+        for ((i, anchor) in PUBLIC_FIPS_ANCHORS.withIndex()) {
+            val (npub, addr, transport) = anchor
+            if (info.npub == npub || incoming.any { it.optString("npub") == npub }) continue
+            incoming += JSONObject().apply {
+                put("npub", npub)
+                put("alias", "fips-network-anchor-${i + 1}")
+                put("addresses", JSONArray().put(JSONObject().apply {
+                    put("transport", transport)
+                    put("addr", addr)
+                    put("priority", 50 + i * 10)
+                }))
+            }
+        }
         val incomingNpubs = incoming.map { it.optString("npub") }.toSet()
         context.fipsDataStore.edit { prefs ->
             val current = JSONArray(prefs[peersKey] ?: "[]")

core/archipelago/src/fips/anchors.rs

@@ -156,7 +156,24 @@ pub async fn load(data_dir: &Path) -> Result<Vec<SeedAnchor>> {
         .with_context(|| format!("read {}", path.display()))?;
     let anchors: Vec<SeedAnchor> =
         serde_json::from_slice(&bytes).with_context(|| format!("parse {}", path.display()))?;
-    Ok(anchors)
+    Ok(repair_legacy_anchor_set(anchors))
+}
+
+/// Add the newer redundant public TCP anchors to legacy files that only carry
+/// the Archipelago-operated vps2 anchor. A deliberately custom/private anchor
+/// list remains authoritative; this only repairs the exact stale shape shipped
+/// before the join.fips.network anchors became defaults.
+fn repair_legacy_anchor_set(mut anchors: Vec<SeedAnchor>) -> Vec<SeedAnchor> {
+    let has_archy = anchors.iter().any(|a| a.npub == ARCHY_ANCHOR_NPUB);
+    let has_fips_network = fips_network_anchors()
+        .iter()
+        .any(|default| anchors.iter().any(|a| a.npub == default.npub));
+    if has_archy && !has_fips_network {
+        for anchor in fips_network_anchors() {
+            anchors.push(anchor);
+        }
+    }
+    anchors
 }
 
 /// Persist the list. Overwrites atomically via write-then-rename so a
@@ -338,6 +355,30 @@ mod tests {
         assert!(got.iter().all(|a| a.transport == "tcp"));
     }
 
+    #[tokio::test]
+    async fn load_repairs_legacy_archy_only_anchor_file() {
+        let dir = tempfile::tempdir().unwrap();
+        save(dir.path(), &[archy_anchor()]).await.unwrap();
+
+        let got = load(dir.path()).await.unwrap();
+        assert!(got.iter().any(|a| a.npub == ARCHY_ANCHOR_NPUB));
+        for anchor in fips_network_anchors() {
+            assert!(got.iter().any(|a| a.npub == anchor.npub));
+        }
+    }
+
+    #[tokio::test]
+    async fn load_keeps_private_anchor_file_authoritative() {
+        let dir = tempfile::tempdir().unwrap();
+        let private = mk("npub1private");
+        save(dir.path(), std::slice::from_ref(&private))
+            .await
+            .unwrap();
+
+        let got = load(dir.path()).await.unwrap();
+        assert_eq!(got, vec![private]);
+    }
+
     #[tokio::test]
     async fn removing_one_default_persists_and_keeps_the_other() {
         // Editing the anchor list (here removing one default) makes the file

core/archipelago/src/fips/service.rs

@@ -81,6 +81,7 @@ async fn sudo_systemctl(verb: &str, unit: &str) -> Result<()> {
 /// Unmask + start + enable the FIPS service. Idempotent — safe to call
 /// on every backend startup once the key is on disk.
 pub async fn activate(unit: &str) -> Result<()> {
+    kill_stale_daemons().await?;
     // Order matters: unmask before enable/start, otherwise enable fails
     // on a masked unit.
     sudo_systemctl("unmask", unit).await?;
@@ -94,9 +95,42 @@ pub async fn stop(unit: &str) -> Result<()> {
 }
 
 pub async fn restart(unit: &str) -> Result<()> {
+    kill_stale_daemons().await?;
     sudo_systemctl("restart", unit).await
 }
 
+/// Kill orphaned `fips` processes not owned by either known systemd unit.
+///
+/// Field failure, 2026-07-24: a stale daemon survived outside systemd and kept
+/// `0.0.0.0:8443` bound. The supervised daemon then started UDP-only, so every
+/// TCP seed-anchor connect failed with "no operational transport" and phones
+/// on 5G could not discover the node. We keep the cleanup narrow: preserve the
+/// MainPID of both units and terminate only extra `pgrep -x fips` matches.
+pub async fn kill_stale_daemons() -> Result<()> {
+    let script = format!(
+        r#"keep="$(systemctl show -p MainPID --value {managed} 2>/dev/null; systemctl show -p MainPID --value {upstream} 2>/dev/null)"
+for pid in $(pgrep -x fips 2>/dev/null || true); do
+  case " $keep " in
+    *" $pid "*) ;;
+    *) kill "$pid" 2>/dev/null || true ;;
+  esac
+done
+"#,
+        managed = super::SERVICE_UNIT,
+        upstream = super::UPSTREAM_SERVICE_UNIT,
+    );
+    let out = Command::new("sudo")
+        .args(["sh", "-c", &script])
+        .output()
+        .await
+        .context("sudo stale fips cleanup failed to launch")?;
+    if !out.status.success() {
+        let stderr = String::from_utf8_lossy(&out.stderr).trim().to_string();
+        anyhow::bail!("stale fips cleanup failed: {}", stderr);
+    }
+    Ok(())
+}
+
 /// Resolve which systemd unit is actually supervising the fips daemon
 /// on this host. Nodes installed from the archipelago ISO run
 /// `archipelago-fips.service`; nodes that were apt-installed (or had

neode-ui/src/components/CompanionIntroOverlay.vue

@@ -150,7 +150,7 @@ const STORAGE_KEY = 'neode_companion_intro_seen'
 // exposes the release-server address.
 const DEFAULT_DOWNLOAD_URL = IS_DEMO
   ? `${window.location.origin}/packages/archipelago-companion.apk`
-  : 'http://146.59.87.168:3000/lfg2025/archy/raw/branch/main/neode-ui/public/packages/archipelago-companion.apk'
+  : 'http://146.59.87.168:2100/packages/archipelago-companion.apk'
 
 // Deep-link scheme the companion app registers; carries the server entry the
 // app should create (see docs/companion-pairing-qr.md for the contract).