fix(sync): re-arm a recovery probe that died while offline (closes #207) (#212)

* fix(sync): re-arm a recovery probe that died while offline (closes #207)

* chore(release): bump version to 0.2.24

---------

Co-authored-by: gnacho <hnacho@proton.me>

Author: gnacho

CHANGELOG.md

@@ -2,6 +2,12 @@
 
 Todas las versiones notables de NextSync se documentan aquí. El formato sigue [Keep a Changelog](https://keepachangelog.com/es/1.1.0/) y el versionado es **+0.0.2 por release, reiniciado en 0.1.4** (decisión del usuario, 22-Ago-2026; sustituye al +0.02 anterior).
 
+## [0.2.24] - 2026-09-17
+
+### Fixed
+
+- Folders no longer stay Offline after connectivity returns when the recovery probe fired while the machine was offline: the probe is re-armed on network restore and synchronization resumes on its own, without a manual sync or an app restart (#207).
+
 ## [0.2.22] - 2026-09-17
 
 ### Fixed

Cargo.lock

@@ -1,6 +1,6 @@
 [package]
 name = "nextsync"
-version = "0.2.22"
+version = "0.2.24"
 edition = "2021"
 rust-version = "1.83"
 license = "GPL-3.0-or-later"

Cargo.toml

@@ -1,6 +1,6 @@
 # Maintainer: gnacho <https://github.com/gnacho>
 pkgname=nextsync
-pkgver=0.2.22
+pkgver=0.2.24
 pkgrel=1
 pkgdesc='Nextcloud desktop synchronization client for GNOME (Rust rewrite)'
 arch=('x86_64' 'aarch64')

PKGBUILD

@@ -17,7 +17,7 @@
   </p>
   <p>
     <a href="https://github.com/gnacho/nextsync/actions/workflows/ci.yml"><img src="https://github.com/gnacho/nextsync/actions/workflows/ci.yml/badge.svg" alt="Estado de CI"></a>
-    <img src="https://img.shields.io/badge/version-0.2.22-blue?style=flat-square" alt="Versión 0.2.22">
+    <img src="https://img.shields.io/badge/version-0.2.24-blue?style=flat-square" alt="Versión 0.2.24">
     <img src="https://img.shields.io/badge/license-GPL--3.0%2B-informational?style=flat-square" alt="GNU GPL v3 o posterior">
   </p>
 </div>
@@ -84,7 +84,7 @@ Ambos motores se esconden detrás del mismo trait pequeño, así que un proveedo
 Descarga el `.pkg.tar.zst` de la [última release](https://github.com/gnacho/nextsync/releases/latest) e instálalo:
 
 ```bash
-sudo pacman -U nextsync-0.2.22-1-x86_64.pkg.tar.zst
+sudo pacman -U nextsync-0.2.24-1-x86_64.pkg.tar.zst
 ```
 
 El paquete depende de `gtk4` y `libadwaita`. Para cuentas Nextcloud instala `nextcloud-client` (aporta `nextcloudcmd`); para cuentas OpenCloud, el `opencloudcmd` oficial.

README.es.md

@@ -17,7 +17,7 @@
   </p>
   <p>
     <a href="https://github.com/gnacho/nextsync/actions/workflows/ci.yml"><img src="https://github.com/gnacho/nextsync/actions/workflows/ci.yml/badge.svg" alt="CI status"></a>
-    <img src="https://img.shields.io/badge/version-0.2.22-blue?style=flat-square" alt="Version 0.2.22">
+    <img src="https://img.shields.io/badge/version-0.2.24-blue?style=flat-square" alt="Version 0.2.24">
     <img src="https://img.shields.io/badge/license-GPL--3.0%2B-informational?style=flat-square" alt="GNU GPL v3 or later">
   </p>
 </div>
@@ -84,7 +84,7 @@ Both engines sit behind the same small trait, so a new provider is a command bui
 Download the `.pkg.tar.zst` from the [latest release](https://github.com/gnacho/nextsync/releases/latest) and install it:
 
 ```bash
-sudo pacman -U nextsync-0.2.22-1-x86_64.pkg.tar.zst
+sudo pacman -U nextsync-0.2.24-1-x86_64.pkg.tar.zst
 ```
 
 The package depends on `gtk4` and `libadwaita`. For Nextcloud accounts install `nextcloud-client` (it provides `nextcloudcmd`); for OpenCloud accounts, the official `opencloudcmd`.

README.md

@@ -32,6 +32,7 @@
   <icon type="stock">io.github.gnacho.nextsync</icon>
   <content_rating type="oars-1.1"/>
   <releases>
+    <release version="0.2.24" date="2026-09-17"/>
     <release version="0.2.22" date="2026-09-17"/>
     <release version="0.2.20" date="2026-09-08"/>
     <release version="0.2.18" date="2026-09-08"/>

data/io.github.gnacho.nextsync.metainfo.xml

@@ -210,7 +210,7 @@ <h3 data-i18n="compare.ht">Lo que NextSync no hace (todavía)</h3>
       <h2 class="reveal" data-i18n="install.title">Instalación fácil y rápida</h2>
       <p class="lead reveal" data-i18n="install.lead">En Arch, CachyOS y derivadas hay paquete listo en cada release.</p>
       <div class="card install-card reveal">
-        <pre class="code-block"><code id="installCmd">sudo pacman -U nextsync-0.2.22-1-x86_64.pkg.tar.zst</code></pre>
+        <pre class="code-block"><code id="installCmd">sudo pacman -U nextsync-0.2.24-1-x86_64.pkg.tar.zst</code></pre>
         <button class="btn suggested" id="copyBtn" data-i18n="install.copy">Copiar comando</button>
         <p class="small-print" data-i18n="install.note">Descarga el paquete <code>.pkg.tar.zst</code> más reciente desde <a href="https://github.com/gnacho/nextsync/releases" target="_blank" rel="noopener">GitHub Releases</a> y ajusta el nombre del fichero.</p>
       </div>

landing/index.html

@@ -1192,6 +1192,16 @@ impl SchedulerInner {
             self.state
                 .set(AppState::Offline, t("Waiting for a network connection"));
         } else if !was_online {
+            // Issue #207: the recovery probe is a one-shot timer; if it fired
+            // while the machine was offline, `start` aborted on !online and
+            // the timer died. Re-arm it now so the folder recovers by itself
+            // once the network is back, instead of sitting queued behind the
+            // server_unreachable gate until a manual sync. The probe guard
+            // (start_source) makes the re-arm a no-op if one is still pending.
+            if self.server_unreachable {
+                self.queue.add(Trigger::Retry);
+                self.schedule_server_probe();
+            }
             if !self.queue.is_empty() || !self.manual_only() {
                 self.request(Trigger::NetworkRestored);
             } else {
@@ -1777,6 +1787,49 @@ mod tests {
         assert_eq!(scheduler.state().snapshot().state, AppState::IdleOk);
     }
 
+    /// Issue #207: the recovery probe is a one-shot timer. If it fires while
+    /// the machine is offline, `start` aborts on !online and the timer is
+    /// gone for good. When the network comes back, the folder must still
+    /// recover on its own - the probe is re-armed and the queued retry runs -
+    /// without a manual sync or an app restart.
+    #[test]
+    fn network_restore_rearms_a_probe_that_died_while_offline() {
+        let (scheduler, source, runner) = make_scheduler(None);
+        scheduler.request(Trigger::Startup);
+        run_idle(&source);
+        finish(&runner, SyncOutcome::NetworkError);
+        assert!(scheduler.server_unreachable());
+        let probe_id = source.borrow().only_id();
+
+        // Connectivity drops before the probe fires; the probe fires anyway
+        // and its one-shot start() aborts on !online: the timer is dead.
+        scheduler.set_online(false);
+        fire_timer(&source, probe_id);
+        assert_eq!(
+            runner.0.borrow().start_calls,
+            1,
+            "an offline probe must not run the engine"
+        );
+        assert_eq!(
+            source.borrow().pending(),
+            0,
+            "the one-shot probe timer is gone"
+        );
+
+        // Connectivity returns: the folder recovers without a manual sync.
+        scheduler.set_online(true);
+        assert!(
+            source.borrow().pending() >= 1,
+            "the recovery probe is re-armed on network restore"
+        );
+        let rearmed_id = source.borrow().only_id();
+        fire_timer(&source, rearmed_id);
+        assert_eq!(runner.0.borrow().start_calls, 2);
+        finish(&runner, SyncOutcome::Success);
+        assert!(!scheduler.server_unreachable());
+        assert_eq!(scheduler.state().snapshot().state, AppState::IdleOk);
+    }
+
     /// Issue #179: after the server recovers (Success clears the gate), a
     /// fresh automatic trigger runs the engine again.
     #[test]

src/core/scheduler.rs

@@ -1,12 +1,11 @@
 {
   "schema_version": 1,
-  "version": "0.2.22",
+  "version": "0.2.24",
   "mandatory": false,
-  "summary": "Detect a missing sync engine and stop reporting success when nothing can be synchronized.",
+  "summary": "Folders recover on their own after connectivity returns, even when the recovery probe fired while the machine was offline.",
   "changelog": [
-    "When the sync binary (nextcloudcmd / opencloudcmd) is not installed, folders now show a clear 'sync engine is not installed' state instead of a generic failure - or worse, a misleading 'Synchronization completed' from the unchanged-ETag interval shortcut while nothing was actually synced.",
-    "A configured engine path that no longer exists resolves as missing too, so a stale override surfaces the same actionable state.",
-    "The missing-engine reason is written to the daily log and raises a desktop notification like other problem outcomes."
+    "After a period without connectivity, folders no longer stay Offline until a manual sync or an app restart: when the network comes back, the server recovery probe is re-armed automatically and synchronization resumes on its own.",
+    "Fixes the case where the one-shot recovery probe fired while the machine was offline, died silently, and the NetworkRestored trigger stayed parked behind the server-unreachable gate forever."
   ],
   "released_at": "2026-09-17T00:00:00Z"
 }