Re: NSRecursiveLock.lock() causing deadlock??
Re: NSRecursiveLock.lock() causing deadlock??
- Subject: Re: NSRecursiveLock.lock() causing deadlock??
- From: Dov Rosenberg <email@hidden>
- Date: Thu, 28 Aug 2008 08:12:01 -0400
- Thread-topic: NSRecursiveLock.lock() causing deadlock??
Title: Re: NSRecursiveLock.lock() causing deadlock??
We only have a single database and a single model in our application. The only other NSRecursiveLocks that are being held are from the EOObjectStoreCoordinator.lock().
I have noticed that we have not been very consistent in using EOEditingContexts. Most of the newer code creates a subclass of EOEditingContext (LockScreamingEditingContext – thanks Chuck!!), there are some references to session.defaultEditingContext(), and some from the WOComponent.defaultEditingContext (I think the WOComponent and session EOEditingContext are the same). We replace the connection dictionary at startup. Could the difference in EOEditingContexts cause a second EOCooperatingObjectStore to be added?
Once again – thanks for your help
Dov Rosenberg
On 8/27/08 10:57 PM, "Chuck Hill" <email@hidden> wrote:
On Aug 27, 2008, at 7:42 PM, Dov Rosenberg wrote:
Would this cause a lock to be held that would bring down the entire app after awhile?
Yes.
Could EODatabaseContext.lock be causing the NSRecursiveLock to be held?
I think what is happening is that some other thread has this lock and EODatabaseContext.lock() is blocking.
Are you connecting to more than one database? If not, I suspect that you have a small (insidious perhaps) difference in the connection dictionaries of your models. That is the only thing that I can think of that would cause a second EOCooperatingObjectStore (EODatabaseContext) to be added.
Is the WO 5.4.2 the same one that came with the iPhone SDK and Xcode? I couldn’t find a download for 5.4.2
It is part of the latest Xcode download, 1 GB more or less.
The offending line of code in our app that was being executed was
siteList = EOUtilities.objectsWithQualifierFormat(ec, "Site", "(referenceKey = %@) AND (parent = null) ", args);
We make extensive use of EOUtilities thru out our app.
That is not causing the problem that is just the place where it surfaces. EOUtilities is safe.
The full thread dump was
at com.webobjects.foundation.NSRecursiveLock.lock(NSRecursiveLock.java:72)
at com.webobjects.eoaccess.EODatabaseContext.lock(EODatabaseContext.java:1973)
at com.webobjects.eocontrol.EOObjectStoreCoordinator.addCooperatingObjectStore(EOObjectStoreCoordinator.java:130)
at com.webobjects.eoaccess.EODatabaseChannel.setCurrentEditingContext(EODatabaseChannel.java:166)
at com.webobjects.eoaccess.EODatabaseChannel._selectWithFetchSpecificationEditingContext(EODatabaseChannel.java:788)
at com.webobjects.eoaccess.EODatabaseChannel.selectObjectsWithFetchSpecification(EODatabaseChannel.java:215)
at com.webobjects.eoaccess.EODatabaseContext._objectsWithFetchSpecificationEditingContext(EODatabaseContext.java:3205)
at com.webobjects.eoaccess.EODatabaseContext.objectsWithFetchSpecification(EODatabaseContext.java:3346)
at com.webobjects.eocontrol.EOObjectStoreCoordinator.objectsWithFetchSpecification(EOObjectStoreCoordinator.java:539)
at com.webobjects.eocontrol.EOEditingContext.objectsWithFetchSpecification(EOEditingContext.java:4114)
at com.webobjects.eocontrol.EOEditingContext.objectsWithFetchSpecification(EOEditingContext.java:4500)
at com.webobjects.eoaccess.EOUtilities.objectsWithQualifierFormat(EOUtilities.java:145)
Thanks again for your insights
Dov Rosenberg
Chuck
On 8/27/08 10:18 PM, "Art Isbell" <email@hidden> wrote:
On Aug 27, 2008, at 3:47 PM, Dov Rosenberg wrote:
I am not sure why a cooperating object store is being added.
"http-10042-Processor46" nid=59971 state=WAITING
- waiting on <0xcb1792> (a com.webobjects.foundation.NSRecursiveLock)
- locked <0xcb1792> (a com.webobjects.foundation.NSRecursiveLock)
at java.lang.Object.wait(Native Method)
at java.lang.Object.wait(Unknown Source)
at com.webobjects.foundation.NSRecursiveLock.lock(NSRecursiveLock.java:72)
at com.webobjects.eoaccess.EODatabaseContext.lock(EODatabaseContext.java:1973)
at com.webobjects.eocontrol.EOObjectStoreCoordinator.addCooperatingObjectStore(EOObjectStoreCoordinator.java:130)
Hmm, you may be suffering from a bug whose effect can be minimized. I think this bug can occur when a cooperating object store is being added, and that the default implementation of addCooperatingObjectStore() doesn't check whether the cooperating object store being added has already been added. By not invoking the default implementation unnecessarily, exposure to this bug can be minimized considerably. This bug may have been fixed in WO 5.4.2, but the following workaround seems pretty harmless even after this bug has been fixed.
In your Application constructor:
EOObjectStoreCoordinator.setDefaultCoordinator(new ObjectStoreCoordinator());
ObjectStoreCoordinator.java:
public class ObjectStoreCoordinator extends EOObjectStoreCoordinator {
@Override
public void addCooperatingObjectStore(EOCooperatingObjectStore anObjectStore) {
if (cooperatingObjectStores().indexOfIdenticalObject(anObjectStore) == NSArray.NotFound) {
if (anObjectStore.coordinator() != null) {
throw new IllegalStateException("Cannot add " + anObjectStore + " to this EOObjectStoreCoordinator because it already has another.");
}
super.addCooperatingObjectStore(anObjectStore);
}
}
}
I also found this line of code – I don’t think it is being called at this time, but it seems like perhaps this should have a try/catch/finally block and a lock()/unlock(). Thoughts?
EOObjectStoreCoordinator.defaultCoordinator().invalidateAllObjects();
I avoid invoking invalidateAllObjects() at almost all costs. Is there no other way to do what this is doing?
Aloha,
Art
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden