Fix race condition in alloc_ongoing()

Hold the same write lock while checking if ongoing
process is already allocated and while allocating it.
Otherwise it is possible for two parallel processes
running alloc_ongoing() to decide that no ongoing
process is allocated and allocate two ongoing processes.

CHANGELOG.md

@@ -24,8 +24,12 @@
 - node: throw error when getting context with an invalid account id
 - node: throw error when instanciating a wrapper class on `null` (Context, Message, Chat, ChatList and so on)
 - use same contact-color if email address differ only in upper-/lowercase #3327
+- fix race condition in ongoing process (import/export, configuration) allocation
 - repair encrypted mails "mixed up" by Google Workspace "Append footer" function #3315
 
+### Removed
+- node: remove unmaintained coverage scripts
+
 
 ## 1.80.0
 

src/context.rs

@@ -278,13 +278,11 @@ impl Context {
     // Ongoing process allocation/free/check
 
     pub(crate) async fn alloc_ongoing(&self) -> Result<Receiver<()>> {
-        if self.has_ongoing().await {
+        let mut s = self.running_state.write().await;
+        if s.ongoing_running || !s.shall_stop_ongoing {
             bail!("There is already another ongoing process running.");
         }
 
-        let s_a = &self.running_state;
-        let mut s = s_a.write().await;
-
         s.ongoing_running = true;
         s.shall_stop_ongoing = false;
         let (sender, receiver) = channel::bounded(1);
@@ -294,25 +292,16 @@ impl Context {
     }
 
     pub(crate) async fn free_ongoing(&self) {
-        let s_a = &self.running_state;
-        let mut s = s_a.write().await;
+        let mut s = self.running_state.write().await;
 
         s.ongoing_running = false;
         s.shall_stop_ongoing = true;
         s.cancel_sender.take();
     }
 
-    pub(crate) async fn has_ongoing(&self) -> bool {
-        let s_a = &self.running_state;
-        let s = s_a.read().await;
-
-        s.ongoing_running || !s.shall_stop_ongoing
-    }
-
     /// Signal an ongoing process to stop.
     pub async fn stop_ongoing(&self) {
-        let s_a = &self.running_state;
-        let mut s = s_a.write().await;
+        let mut s = self.running_state.write().await;
         if let Some(cancel) = s.cancel_sender.take() {
             if let Err(err) = cancel.send(()).await {
                 warn!(self, "could not cancel ongoing: {:?}", err);
@@ -324,7 +313,7 @@ impl Context {
             s.shall_stop_ongoing = true;
         } else {
             info!(self, "No ongoing process to stop.",);
-        };
+        }
     }
 
     pub(crate) async fn shall_stop_ongoing(&self) -> bool {

src/securejoin.rs

@@ -741,7 +741,6 @@ mod tests {
         );
 
         let sent = bob.pop_sent_msg().await;
-        assert!(!bob.ctx.has_ongoing().await);
         assert_eq!(sent.recipient(), "alice@example.org".parse().unwrap());
         let msg = alice.parse_msg(&sent).await;
         assert!(!msg.was_encrypted());
@@ -1291,7 +1290,6 @@ mod tests {
         let bob_chat = Chat::load_from_db(&bob.ctx, bob_chatid).await?;
         assert!(bob_chat.is_protected());
         assert!(bob_chat.typ == Chattype::Group);
-        assert!(!bob.ctx.has_ongoing().await);
 
         // On this "happy path", Alice and Bob get only a group-chat where all information are added to.
         // The one-to-one chats are used internally for the hidden handshake messages,