You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@myfaces.apache.org by "melloware (via GitHub)" <gi...@apache.org> on 2026/06/19 18:45:17 UTC

[PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

melloware opened a new pull request, #1022:
URL: https://github.com/apache/myfaces/pull/1022

   @tandraschko smallest possible fix I could make is that that the problem is that `CopyOnWriteArrayList.size()` and `get()` are not atomic so by using the `for each` the collection is only grabbed once.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "melloware (via GitHub)" <gi...@apache.org>.
melloware commented on PR #1022:
URL: https://github.com/apache/myfaces/pull/1022#issuecomment-4758600474

   No I will backport but wanted you to approve it first. 


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "melloware (via GitHub)" <gi...@apache.org>.
melloware merged PR #1022:
URL: https://github.com/apache/myfaces/pull/1022


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "tandraschko (via GitHub)" <gi...@apache.org>.
tandraschko commented on PR #1022:
URL: https://github.com/apache/myfaces/pull/1022#issuecomment-4758644321

   and please add a comment like: Use for-each loop because the list can be changed during calling event listeners
   
   ohterwise i will auto-refactor next time


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "tandraschko (via GitHub)" <gi...@apache.org>.
tandraschko commented on code in PR #1022:
URL: https://github.com/apache/myfaces/pull/1022#discussion_r3447259382


##########
impl/src/main/java/org/apache/myfaces/application/ApplicationImplEventManager.java:
##########
@@ -122,11 +122,11 @@ public void subscribeToEvent(Class<? extends SystemEvent> systemEventClass, Clas
         Assert.notNull(systemEventClass, "systemEventClass");
         Assert.notNull(listener, "listener");
 
-        List<EventInfo> eventInfos = globalListeners.get(systemEventClass);
+        // MYFACES-4757 use computeIfAbsent to prevent multiple threads from creating different lists
+        List eventInfos = globalListeners.get(systemEventClass);

Review Comment:
   please readd generics



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "tandraschko (via GitHub)" <gi...@apache.org>.
tandraschko commented on PR #1022:
URL: https://github.com/apache/myfaces/pull/1022#issuecomment-4758540536

   is it only broken on master?


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "melloware (via GitHub)" <gi...@apache.org>.
melloware commented on PR #1022:
URL: https://github.com/apache/myfaces/pull/1022#issuecomment-4758658078

   here was the note from AI on computeIfAbstent
   
   Yes, this is a race condition.
   
   The bug is exactly here:
   
   ```java
   for (int i = 0, size = eventInfos.size(); i < size; i++)
   {
       EventInfo eventInfo = eventInfos.get(i);
       ...
   }
   ```
   
   `eventInfos` is actually a `CopyOnWriteArrayList<EventInfo>`:
   
   ```java
   eventInfos = new CopyOnWriteArrayList<>();
   ```
   
   The problem is that `CopyOnWriteArrayList.size()` and `get()` are **not atomic**. Consider:
   
   ```
   Thread A                         Thread B
   --------                         --------
   
   size = eventInfos.size();   // 3
   
                                    unsubscribeFromEvent()
                                    removeIf(...)
                                    backing array now size = 2
   
   eventInfos.get(2);          // ArrayIndexOutOfBoundsException
   ```
   
   `CopyOnWriteArrayList` intentionally replaces its internal array on every modification. The `size()` came from one snapshot while `get()` reads from a newer snapshot.
   
   This is exactly the exception you're seeing:
   
   ```
   Index 2 out of bounds for length 2
   ```
   
   ---
   
   ## The correct fix
   
   The JDK documentation recommends iterating using the snapshot iterator.
   
   Instead of
   
   ```java
   for (int i = 0, size = eventInfos.size(); i < size; i++)
   {
       EventInfo eventInfo = eventInfos.get(i);
       ...
   }
   ```
   
   use
   
   ```java
   for (EventInfo eventInfo : eventInfos)
   {
       if (eventInfo.sourceClass != null &&
           !eventInfo.sourceClass.isAssignableFrom(sourceBaseType))
       {
           continue;
       }
   
       if (eventInfo.listener.isListenerForSource(source))
       {
           if (event == null)
           {
               event = createEvent(systemEventClass, facesContext, source);
           }
   
           if (event.isAppropriateListener(eventInfo.listener))
           {
               event.processListener(eventInfo.listener);
           }
       }
   }
   ```
   
   or equivalently
   
   ```java
   for (Iterator<EventInfo> it = eventInfos.iterator(); it.hasNext();)
   {
       EventInfo eventInfo = it.next();
       ...
   }
   ```
   
   The iterator captures the backing array once, so it is immune to concurrent additions/removals.
   
   ---
   
   ## There are other occurrences in this class
   
   I also noticed `processComponentAttachedListeners()` contains the same pattern:
   
   ```java
   for (int i = 0, size = listeners.size(); i < size; i++)
   {
       SystemEventListener listener = listeners.get(i);
       ...
   }
   ```
   
   If `holder.getListenersForEventClass()` ever returns a `CopyOnWriteArrayList` (which it commonly does in JSF implementations), this method has the identical race.
   
   `processViewAttachedListeners()` is already written defensively by making a copy because of MYFACES-2935, so the author was clearly aware of concurrent modification there but overlooked the other two methods.
   
   ---
   
   ## I would also fix `subscribeToEvent()`
   
   There is another concurrency issue:
   
   ```java
   List<EventInfo> eventInfos = globalListeners.get(systemEventClass);
   if (eventInfos == null)
   {
       eventInfos = new CopyOnWriteArrayList<>();
       globalListeners.put(systemEventClass, eventInfos);
   }
   ```
   
   Two threads can both create different lists:
   
   ```
   T1: get() -> null
   T2: get() -> null
   
   T1: put(listA)
   T2: put(listB)
   
   T1: listA.add(listener1)
   T2: listB.add(listener2)
   ```
   
   One listener is effectively lost because `listA` is no longer referenced by the map.
   
   This should instead use:
   
   ```java
   List<EventInfo> eventInfos =
       globalListeners.computeIfAbsent(systemEventClass,
           k -> new CopyOnWriteArrayList<>());
   ```
   
   which atomically creates or retrieves the shared list.
   
   So I think there are **two real concurrency bugs** in `ApplicationImplEventManager`:
   
   1. **`processGlobalListeners()`** (the one causing your exception) iterates a `CopyOnWriteArrayList` incorrectly using `size()`/`get()`.
   2. **`subscribeToEvent()`** can lose listener registrations under concurrent subscriptions because it uses `get()`/`put()` instead of `computeIfAbsent()`.
   
   Both are worth fixing in a MyFaces pull request.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "tandraschko (via GitHub)" <gi...@apache.org>.
tandraschko commented on PR #1022:
URL: https://github.com/apache/myfaces/pull/1022#issuecomment-4758667418

   then you can still use:
   
   List<EventInfo> eventInfos = globalListeners.get(systemEventClass);
   if (eventInfos == null)
   {
       eventInfos = globalListeners.computeIfAbsent(systemEventClass, k -> new CopyOnWriteArrayList<>());
   }


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "tandraschko (via GitHub)" <gi...@apache.org>.
tandraschko commented on code in PR #1022:
URL: https://github.com/apache/myfaces/pull/1022#discussion_r3446735321


##########
impl/src/main/java/org/apache/myfaces/application/ApplicationImplEventManager.java:
##########
@@ -122,12 +122,8 @@ public void subscribeToEvent(Class<? extends SystemEvent> systemEventClass, Clas
         Assert.notNull(systemEventClass, "systemEventClass");
         Assert.notNull(listener, "listener");
 
-        List<EventInfo> eventInfos = globalListeners.get(systemEventClass);
-        if (eventInfos == null)
-        {
-            eventInfos = new CopyOnWriteArrayList<>();
-            globalListeners.put(systemEventClass, eventInfos);
-        }
+       List<EventInfo> eventInfos =

Review Comment:
   the old way is faster



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "tandraschko (via GitHub)" <gi...@apache.org>.
tandraschko commented on PR #1022:
URL: https://github.com/apache/myfaces/pull/1022#issuecomment-4766356525

   +1 now


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "melloware (via GitHub)" <gi...@apache.org>.
melloware commented on PR #1022:
URL: https://github.com/apache/myfaces/pull/1022#issuecomment-4758675101

   pushed up some fixes


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] MYFACES-4757: 5.0 Fix race conditions of CopyOnWriteList [myfaces]

Posted by "melloware (via GitHub)" <gi...@apache.org>.
melloware commented on PR #1022:
URL: https://github.com/apache/myfaces/pull/1022#issuecomment-4761812831

   Fixed


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscribe@myfaces.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org