Fix/desktop startup logs and update

desktop/app/loading.html

@@ -34,7 +34,8 @@
     .foot .warn { color:var(--warn); }
     .diag { margin-top:10px; font-size:.72rem; color:var(--muted); font-family: ui-monospace, Menlo, monospace;
       background: rgba(255,255,255,.04); border:1px solid rgba(255,255,255,.08); border-radius:8px;
-      padding:8px 10px; user-select:text; display:none; white-space:pre-wrap; text-align:left; }
+      padding:8px 10px; user-select:text; display:none; white-space:pre-wrap; text-align:left;
+      max-height:260px; overflow:auto; }
     .actions { margin-top:18px; display:flex; gap:12px; align-items:center; justify-content:center; min-height:34px; }
     .btn { font: inherit; font-size:.85rem; font-weight:600; padding:7px 16px; border-radius:9px; cursor:pointer;
       border:1px solid rgba(255,255,255,.16); background:rgba(255,255,255,.05); color:var(--fg); }
@@ -56,7 +57,6 @@
     <div class="diag" id="diag"></div>
     <div class="actions">
       <button class="btn hidden" id="retryBtn">Retry</button>
-      <a class="link hidden" id="openAnyway">Open anyway →</a>
       <button class="btn hidden" id="quitBtn">Quit</button>
     </div>
   </div>
@@ -100,6 +100,20 @@
     function stepUrl(s){ return 'http://127.0.0.1:' + PORTS[s.portKey] + s.path; }
 
     var t0 = Date.now(), quipIdx = 0, quipStep = null, doneRedirect = false, failed = false, curStartedAt = {}, cur = null;
+    var STARTUP_LOG = '';
+    // Pull the teed backend logs (<app_data>/startup.log) from the Tauri host so the
+    // failure screen can show WHY the services didn't come up.
+    function fetchStartupLog(){
+      try {
+        if (window.__TAURI__ && window.__TAURI__.core) {
+          window.__TAURI__.core.invoke('read_startup_log')
+            .then(function(t){ STARTUP_LOG = (t && t.trim()) ? t : '(startup log is empty)'; render(); })
+            .catch(function(){ STARTUP_LOG = '(could not read startup log)'; render(); });
+          return;
+        }
+      } catch (e) {}
+      STARTUP_LOG = '(startup log unavailable)';
+    }
     var $ = function(id){ return document.getElementById(id); };
     STEPS.forEach(function(s){
       var li = document.createElement('li'); li.id = 'li-'+s.id;
@@ -136,10 +150,15 @@
         $('stage').textContent = "Orcabot couldn't finish starting up";
         $('quip').textContent = 'The background services didn’t come up. A previous instance may still be running, or a port is in use.';
         $('hint').textContent = '';
-        $('diag').style.display = 'block'; $('diag').textContent = diagText();
+        $('diag').style.display = 'block';
+        // Show the port/health summary plus the actual backend logs (workerd/D1),
+        // which explain WHY startup failed (e.g. a port bind error) — otherwise
+        // invisible in a Finder-launched .app.
+        $('diag').textContent = diagText()
+          + '\n\n─── background server logs ───\n'
+          + (STARTUP_LOG || '(fetching…)');
         $('retryBtn').classList.remove('hidden');
         $('quitBtn').classList.remove('hidden');
-        $('openAnyway').classList.remove('hidden');
         return;
       }
       cur = current();
@@ -149,7 +168,6 @@
       $('quip').textContent = cur.quips[quipIdx % cur.quips.length];
       var waited = (Date.now() - (curStartedAt[cur.id]||t0))/1000;
       $('hint').innerHTML = waited > (cur.slow||12) ? ' · <span class="warn">'+cur.label+' is taking longer than usual…</span>' : '';
-      if (elapsed() > MANUAL_AFTER){ $('openAnyway').classList.remove('hidden'); }
     }
 
     function tick(){
@@ -158,12 +176,11 @@
         .then(function(){
           var ready = STEPS.every(function(s){ return !s.required || s.up; });
           if (ready && !doneRedirect){ doneRedirect = true; render(); setTimeout(function(){ window.location.replace(redirectUrl()); }, 500); return; }
-          if (!ready && elapsed() >= FAIL_AFTER){ failed = true; }
+          if (!ready && elapsed() >= FAIL_AFTER){ failed = true; fetchStartupLog(); }
           render();
         });
     }
 
-    $('openAnyway').addEventListener('click', function(){ window.location.replace(redirectUrl()); });
     $('retryBtn').addEventListener('click', function(){ window.location.reload(); });
     $('quitBtn').addEventListener('click', function(){
       try { window.__TAURI__.core.invoke('quit_app'); } catch(e) { try { window.close(); } catch(_){} }

desktop/app/src-tauri/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "orcabot-desktop"
-version = "0.6.0"
+version = "0.8.0"
 edition = "2021"
 description = "Orcabot Desktop shell"
 license = "UNLICENSED"

desktop/app/src-tauri/src/commands.rs

@@ -616,6 +616,20 @@ pub fn get_app_version(app: tauri::AppHandle) -> String {
     app.package_info().version.to_string()
 }
 
+/// Read this boot's startup log (`<app_data>/startup.log`) — the teed workerd / D1
+/// output plus the chosen ports — so the loading screen can show WHY the backend
+/// failed to start (a Finder-launched .app has no console). Empty string if none.
+#[tauri::command]
+pub fn read_startup_log(app: tauri::AppHandle) -> String {
+    use tauri::Manager;
+    app.path()
+        .app_data_dir()
+        .ok()
+        .map(|d| d.join("startup.log"))
+        .and_then(|p| std::fs::read_to_string(p).ok())
+        .unwrap_or_default()
+}
+
 #[derive(Serialize, Clone)]
 pub struct OrcabotAccount {
     pub email: String,

desktop/app/src-tauri/src/main.rs

@@ -26,6 +26,16 @@ fn pid_file_path(data_dir: &Path) -> PathBuf {
   data_dir.join("desktop-services.pid")
 }
 
+/// Set once the user accepts an auto-update ("Update & restart"). The whole stack is
+/// about to be torn down and relaunched, so we must NOT keep spinning up heavy
+/// processes (especially the sandbox VM boot / image download) during the ~minutes
+/// download — `start_sandbox_vm` checks this and bails.
+static UPDATING: std::sync::atomic::AtomicBool = std::sync::atomic::AtomicBool::new(false);
+
+fn is_updating() -> bool {
+  UPDATING.load(std::sync::atomic::Ordering::SeqCst)
+}
+
 /// Progress of an in-flight auto-update, emitted to the GUI as `update-progress`
 /// so the frontend can show a download bar (the native "Update available" dialog
 /// otherwise gives no feedback between "Update & restart" and the relaunch).
@@ -201,6 +211,34 @@ fn ensure_port_env(var: &str, preferred: u16, used: &[u16]) -> u16 {
   port
 }
 
+/// Tee a child process stream line-by-line to the console AND (if available) the
+/// per-boot startup log, each line prefixed with the service label. Runs on its own
+/// thread so it drains the pipe continuously (never blocking the child on a full buffer).
+fn tee_child_stream<R: std::io::Read + Send + 'static>(
+  stream: R,
+  label: String,
+  log_path: Option<PathBuf>,
+  is_err: bool,
+) {
+  use std::io::{BufRead, BufReader, Write};
+  std::thread::spawn(move || {
+    let mut logf =
+      log_path.and_then(|p| std::fs::OpenOptions::new().create(true).append(true).open(p).ok());
+    for line in BufReader::new(stream).lines() {
+      let Ok(line) = line else { break };
+      let out = format!("[{}] {}", label, line);
+      if is_err {
+        eprintln!("{}", out);
+      } else {
+        println!("{}", out);
+      }
+      if let Some(ref mut f) = logf {
+        let _ = writeln!(f, "{}", out);
+      }
+    }
+  });
+}
+
 /// Per-boot token that gates dev-auth to the trusted host frontend. Generated
 /// once at startup, passed to the control-plane worker (`SURFACE_TOKEN`) and
 /// handed to the GUI webview via the `get_surface_token` command. The sandbox VM
@@ -461,6 +499,13 @@ impl DesktopServices {
       );
     }
 
+    // Start the per-boot startup log with the chosen ports — the first thing to check
+    // when startup fails (was a default busy? did allocation move a service?).
+    self.reset_startup_log(&format!(
+      "ports: control-plane={} frontend={} d1-shim={} sandbox={}",
+      cp_port, fe_port, d1_port, sandbox_host_port
+    ));
+
     if cp_port != 8787 || fe_port != 8788 || d1_port != 9001 || sandbox_host_port != 8080 {
       eprintln!(
         "[ports] a default port was busy — using control-plane={} frontend={} d1-shim={} sandbox={}",
@@ -665,6 +710,13 @@ impl DesktopServices {
     data_dir: &Path,
     resource_root: &Path,
   ) -> Result<(), vm::VMError> {
+    // The user accepted an app update → don't spin the VM up (or download its image)
+    // just to tear it all down on the imminent relaunch.
+    if is_updating() {
+      eprintln!("[vm] app update accepted — skipping sandbox VM startup");
+      return Ok(());
+    }
+
     // Check if VM resources exist
     let vm_resource_paths = vm::image::VMResourcePaths::from_resource_root(resource_root);
 
@@ -775,9 +827,20 @@ impl DesktopServices {
     };
     config = config.with_cmdline(cmdline);
 
-    // Create and start VM
+    // Create and start VM — unless an update was accepted while we were staging.
+    if is_updating() {
+      eprintln!("[vm] app update accepted — not booting sandbox VM");
+      return Ok(());
+    }
     let mut vm = create_platform_vm();
     vm.start(&config)?;
+    // If the update landed during boot, stop the VM we just started rather than
+    // waiting 120s for health only to tear it down on relaunch.
+    if is_updating() {
+      eprintln!("[vm] app update accepted mid-boot — stopping sandbox VM");
+      let _ = vm.stop();
+      return Ok(());
+    }
 
     // Wait for sandbox to be healthy
     eprintln!("Waiting for sandbox VM to become healthy...");
@@ -818,24 +881,76 @@ impl DesktopServices {
 
     let mut command = Command::new(binary_path);
     command.args(args);
-    command.stdout(Stdio::inherit());
-    command.stderr(Stdio::inherit());
+    // Tee stdout+stderr to the console AND <data_dir>/startup.log. A Finder-launched
+    // .app has no attached terminal, so inherited output vanishes — this keeps a
+    // per-boot record of WHY a service failed (surfaced in the loading screen and
+    // recoverable as a file).
+    command.stdout(Stdio::piped());
+    command.stderr(Stdio::piped());
     for (key, value) in envs {
       command.env(key, value);
     }
 
     match command.spawn() {
-      Ok(child) => {
+      Ok(mut child) => {
+        let log_path = self.startup_log_path();
+        if let Some(out) = child.stdout.take() {
+          tee_child_stream(out, label.to_string(), log_path.clone(), false);
+        }
+        if let Some(err) = child.stderr.take() {
+          tee_child_stream(err, label.to_string(), log_path, true);
+        }
         if let Ok(mut children) = self.children.lock() {
           children.push(child);
         }
       }
       Err(err) => {
         eprintln!("Failed to start {}: {}", label, err);
+        self.append_startup_log(&format!("[{}] FAILED TO START: {}", label, err));
       }
     }
   }
 
+  /// `<data_dir>/startup.log` — where service output is teed for post-mortem.
+  fn startup_log_path(&self) -> Option<PathBuf> {
+    self
+      .data_dir
+      .lock()
+      .ok()
+      .and_then(|dd| dd.clone())
+      .map(|d| d.join("startup.log"))
+  }
+
+  fn append_startup_log(&self, line: &str) {
+    if let Some(p) = self.startup_log_path() {
+      use std::io::Write;
+      if let Ok(mut f) = std::fs::OpenOptions::new().create(true).append(true).open(p) {
+        let _ = writeln!(f, "{}", line);
+      }
+    }
+  }
+
+  /// Truncate the startup log at the start of a boot and write a header, so it only
+  /// ever reflects the CURRENT startup attempt (not accumulated across launches).
+  fn reset_startup_log(&self, header: &str) {
+    if let Some(p) = self.startup_log_path() {
+      let _ = std::fs::write(p, format!("=== Orcabot desktop startup ===\n{}\n", header));
+    }
+  }
+
+  /// Stop ONLY the sandbox VM (leave workerd/frontend running). Used when the user
+  /// accepts an update: the heavy VM shouldn't keep running/booting during the
+  /// download, but the frontend must stay up so the update-progress bar keeps working.
+  fn stop_sandbox_vm(&self) {
+    if let Ok(mut vm_lock) = self.sandbox_vm.lock() {
+      if let Some(ref mut vm) = *vm_lock {
+        eprintln!("Stopping sandbox VM (app update in progress)...");
+        let _ = vm.stop();
+      }
+      *vm_lock = None;
+    }
+  }
+
   fn shutdown(&self) {
     // Stop sandbox VM first
     if let Ok(mut vm_lock) = self.sandbox_vm.lock() {
@@ -1012,6 +1127,7 @@ fn main() {
       commands::reveal_workspace,
       commands::get_ports,
       commands::get_app_version,
+      commands::read_startup_log,
       commands::verify_orcabot_account,
       commands::set_cloud_credential,
       commands::sign_in_google_loopback,
@@ -1130,6 +1246,7 @@ fn main() {
         use tauri_plugin_dialog::{DialogExt, MessageDialogButtons};
         use tauri_plugin_updater::UpdaterExt;
         let handle = app.handle().clone();
+        let upd_services = Arc::clone(&services);
         tauri::async_runtime::spawn(async move {
           let updater = match handle.updater() {
             Ok(u) => u,
@@ -1158,6 +1275,13 @@ fn main() {
             return;
           }
 
+          // The user chose update+restart. Everything is about to be torn down and
+          // relaunched, so stop spinning up the heavy sandbox VM during the ~minutes
+          // download: flag it so start_sandbox_vm won't (re)boot, and stop any VM that
+          // already came up. Leave workerd/frontend running so this progress UI works.
+          UPDATING.store(true, std::sync::atomic::Ordering::SeqCst);
+          upd_services.stop_sandbox_vm();
+
           // Tell the GUI a download is starting so it can show a progress bar,
           // then stream progress from the download closure. Emits are throttled
           // to once per MB so a ~190MB download doesn't flood IPC.

desktop/app/src-tauri/tauri.conf.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://schema.tauri.app/config/2",
   "productName": "Orcabot",
-  "version": "0.6.0",
+  "version": "0.8.0",
   "identifier": "com.orcabot.desktop",
   "build": {
     "beforeDevCommand": "sh -c \"cd ../../frontend && NEXT_PUBLIC_API_URL=http://localhost:8787 NEXT_PUBLIC_SITE_URL=http://localhost:8788 NEXT_PUBLIC_DEV_MODE_ENABLED=true NEXT_PUBLIC_DESKTOP_MODE=true npx wrangler dev -c wrangler.toml --port 8788\"",
@@ -48,7 +48,7 @@
     "macOS": {
       "entitlements": "entitlements.plist",
       "signingIdentity": "Developer ID Application: Robert Macrae (3927MKQNPA)",
-      "minimumSystemVersion": "13.0"
+      "minimumSystemVersion": "13.5"
     }
   },
   "plugins": {

desktop/app/src-tauri/vm-image.json

@@ -1,5 +1,5 @@
 {
-  "version": "v2",
-  "sha256": "eba0a9e5c17778f5de08371dd9bcb46aa3697a827d8c0ed7c801a465c907f633",
-  "url": "https://github.com/Hyper-Int/OrcaBot/releases/download/vm-image-v2/sandbox.img.gz"
+  "version": "v3",
+  "sha256": "a040638617597b660ea8baef915d4de70267fa48632f910fc892d9115313daa4",
+  "url": "https://github.com/Hyper-Int/OrcaBot/releases/download/vm-image-v3/sandbox.img.gz"
 }

frontend/src/app/download/download.css

@@ -47,6 +47,11 @@
   margin-bottom: 1rem;
 }
 
+.dl-req {
+  margin: -0.5rem 0 1rem;
+  font-size: 0.85rem;
+}
+
 .dl-btn {
   display: inline-flex;
   flex-direction: column;