A related Github Pull-Request [GH-408] was closed
https://github.com/doctrine/doctrine2/pull/408

A related Github Pull-Request [GH-407] was closed
https://github.com/doctrine/doctrine2/pull/407

Merged this PR

Can you try to replace your cache with an `ArrayCache` and see if the problem may come from there?
It would also be interesting to see your bootstrap code and the code you use to interact with the collection.

I'm already using an ArrayCache for my cache.

Also for my bootstrap I'm using this implementation of it : https://github.com/guilhermeblanco/ZendFramework1-Doctrine2.

Here is how i interact with the collection.

$collection->clear();
foreach ($values as $value) {
$collection->add(
$this->_em->getReference(
'Entity\User\Privilege',
$value
));}

... Later on ...
... I have an entity $userAccount with many memberships set to Cascade persist, ....
... and on the membership entity relationship i have my collection ...

$this->_em->persist($userAccount);
$this->_em->flush();

I don't know if it helps but it works if I replace $mapping with $coll on line 181 :

$sourceClass = $this->_em->getClassMetadata(get_class($mapping->getOwner()));

Replaced by :

$sourceClass = $this->_em->getClassMetadata(get_class($coll->getOwner()));

Looks like code coming from the (removed) AssociationMapping class. The fix seems also to be valid. I'm patching this.

Nevermind, I don't think this needs a test. It is just something overlooked during a refactoring. Being handled at DDC-1941

Could you please provide the models anyway? It would be interesting to see why the test suite doesn't cover that part
Thank you!

Ok I attached it. Tell me if you need more informations i didn't upload the full models. But just the parts i thought were relevant.

Fixed and applied to 2.2.3 and 2.3

A related Github Pull-Request [GH-405] was closed
https://github.com/doctrine/doctrine2/pull/405

PHP 5.3.6-13ubuntu3.8 with Suhosin-Patch (cli) (built: Jun 13 2012 17:19:54)
Copyright (c) 1997-2011 The PHP Group
Zend Engine v2.3.0, Copyright (c) 1998-2011 Zend Technologies
with Xdebug v2.2.0, Copyright (c) 2002-2012, by Derick Rethans

Can you try namespacing your caches (metadata/query/results) and see if the problem persists?
Also, it would be interesting to see your bootstrap configuration...

We need the complete bootstrap configuratino and options.

DoctrineCache.php is a controller plugin that runs first and initializes doctrine cache usually memcache servers
options.ini contains all configuration options for doctrine

Bisna initializes Doctrine based on options and integrates it into ZF

as you can see all caches are namespaced, and i'm currently the only one using system locally so no cache conflicts should occur even without namespaces

Can you show one of your controller actions/model services where this error occurs, explicitly the code that generates the ResultCache keys.

You should definitely namespace your cache (with something like Doctrine\Common\Cache\ApcCache#setNamespace() for example), since you're manually defining your cache key here...

Is the explicit usage of the cache ID intended?

Benjamin Eberlei, should the fetched cache items be checked for their type?

What are the contents of $urls and $possible_match? You should namespace them even further in your code:

$q->useResultCache(true, \SCMS\DbCache::TTL_ROUTE, "my_query_type_something_" . $possible_match);

This isn't related to the result cache, but to the query cache though i just realized. However i don't see any way this is possible, the DQL Query Cache key is generated as a hash, the Metadata cache entry is generated using the ClassName as key. I don't udnerstand how the Metadata could end up in a Query Cache key.

@ocramius As a fix, in AbstractQuery and ClassMetadataFactory, we should check if the return value is really instanceof ClassMetadata or ParserResult. And if that happens throw an error. Also we should prefix the metadata and DQL queries ourself.

$urls is an array of URIs array('/tv/samsung/lcd', '/tv/samsung', '/tv');
$possible_match is the value of array ex.: '/tv/samsung/lcd'

i'm setting the namespace in the configuration, and providing a unique id for each result cache. the documentation isn't clear
is namespace prefixed automatically if you set your cache id or is namespace ignored if you set cache id yourself.
so i always assumed that namespace is prefixed automatically, and that you have to set cacheId yourself for same query with different parameters, so that query searching for '/tv/samsung' would not return the same result as searching for '/tv';

can you try setting different namespace for Query and Metadata cache? They seem to use the same prefix. Only the Result cache uses its own prefix.

Justinas, setting different caches for query, results and metadata allows you to have, as an example, 3 caches writing/reading from APC. If you set a namespace for those caches, those won't collide (everything is handled transparently).

Anyway Benjamin Eberlei is correct when he states that this shouldn't happen here. Even if your approach isn't correct (you don't need to explicitly set any cache key here, plus you should be using namespaced caches) the issue seems to be valid...

Does this happen also with a clean cache?

i'm not sure, about the clean cache, it just doesn't happen all the time so it's hard to check. If i remember correctly i flushed the cache, and the error appeared, after a few refreshes it disappeared, without flushing cache. it sorta appears and disappears from time to time.

even if i don't set the namespace for my caches i have a local memcache, so no cache collision from other projects shouldn't occur

Added a guard to avoid this problem. Its not a real fix, but i cannot come up with the a way to reproduce the problem that you have exactly.

Symfony2

Fixed

A related Github Pull-Request [GH-386] was closed
https://github.com/doctrine/doctrine2/pull/386

See DDC-1865

DDC-1895 i mean

A related Github Pull-Request [GH-389] was opened
https://github.com/doctrine/doctrine2/pull/389

A related Github Pull-Request [GH-389] was closed
https://github.com/doctrine/doctrine2/pull/389

Fixed

Mysql 5.5.22-0ubuntu1

Any update or ETA on this one? The work-around is still in my production code and I would like to get it out to get it cleaned-up a bit.

Why does the $this->_em->lock() in EntityRepository#find() don't work for you? It should grab the entity from cache, then do a SELECT 1 FROM table and do the pessemistic write lock on the row.

I don't exactly know why that does not work. I created this bugreport on your request (see: https://groups.google.com/forum/?fromgroups#!topic/doctrine-user/N8Xop2-XbTY) because I couldnt figure out what I was doïng wrong.

The script that updates the entity is a long running console script, allot of times the entity in question is already in the cache and did not get locked properly. There are two seperate processes that work on the same entity (processing jobs from a jobqueu) and when updating the counters at the same time, and both scripts (with the same locking code as described in the bugreport) have to update a counter, the counter is not updated properly (e.g. both scripts do for example 2-1 where the first script should do 2-1, the second script 1-1 since the counter is updated by the first script).

Can you paste the SQL log that gets executed?

Allright, I reverted the code back to my original code. Doctrine is updated to the latest dev version. Code looks like this:

 
$em = $this->getEntityManager();
#$em->clear(); // This fixes the locking problem
$em->getConnection()->beginTransaction();

try {
    // Lock entity in db and load it to update counters
    $entity = $em->find($this->getEntityName(), $this->getEppEntityId(), LockMode::PESSIMISTIC_WRITE);

    printf('Decreasing pendingjobs with 1. Current amount of pending jobs is %d', $entity->getPendingJobs());

    $entity->setPendingJobs($entity->getPendingJobs() - 1);

    $em->persist($entity);
    $em->flush();
    $em->getConnection()->commit();
} catch ( \Exception $ex ) {
    $em->getConnection()->rollback();
}

The output of the debug statement looks like this (e.g. each line is printed by a different process. Instead of the first script decreasing 2 to 1 and the second script decreasing 1 to 0, both scripts decrease 2 to 1):

 
[2012-07-05 11:31:08] Decreasing pendingjobs with 1. Current amount of pending jobs is 2
[2012-07-05 11:31:07] Decreasing pendingjobs with 1. Current amount of pending jobs is 2

I enabled SQL logging but the thing is huge and I dont really know where to look. There is data in it that I dont want publically availible on the internet, is it possible to mail the to you directly?

Fixed

A related Github Pull-Request [GH-356] was closed
https://github.com/doctrine/doctrine2/pull/356

Test and fix: https://github.com/doctrine/doctrine2/pull/356

A related Github Pull-Request [GH-354] was closed
https://github.com/doctrine/doctrine2/pull/354

A related Github Pull-Request [GH-353] was closed
https://github.com/doctrine/doctrine2/pull/353

A related Github Pull-Request [GH-351] was closed
https://github.com/doctrine/doctrine2/pull/351

MAC OS X 10.6.8, Symfony 2.0.12, PHP 5.3.6, mysql server 5.5.9

This problem is encountered only in this case, 1-1 and n-m relationship are handled in the right way, has anyone else faced this problem too?

linuxatico

Hi all,
I wanna give the Doctrine community my full support to help fixing this bug, but I need someone who can give me an answer.....
I couldn't figure out in the source code which is the method executed when given the command

$ ./../../bin/php/php5.3.6/bin/php app/console doctrine:mapping:convert yml ./src/Acme/TaskBundle/Resources/config/doctrine/ --from-database --force

I will keep on looking for it, but some help will be appreciated,

linuxatico

Found in vendor/doctrine/lib/Doctrine/ORM/Tools/Console/Command/ConvertMappingCommand.php
I'll see what I can do.

linuxatico

Even if I keep being ignored, I want to report a very useful discovery about this annoying bug: it's 100% related to the YAML conversion, because if you execute the first command

 $ ./../../bin/php/php5.3.6/bin/php app/console doctrine:mapping:convert xml ./src/Acme/TaskBundle/Resources/config/doctrine/ --from-database --force

Using XML instead of YML it works in the expected way. I wonder if the author of this code have written a unit test before integrating this function in the official release of Doctrine.... (ironic question)

linuxatico

This case was indeed not unit-tested, many-to-one and one-to-one were handled the same in YAML Exporter. No need to get picky about it though, we are investing our free time here.

Fixed

I think the problem is on the line 1037 to 1042 of Class Doctrine \ ORM \ Tools \ EntityGenerator:

if (isset($metadata->sequenceGeneratorDefinition['allocationSize']))

if (isset($metadata->sequenceGeneratorDefinition['initialValue']))

Replace to:

if (isset($metadata->sequenceGeneratorDefinition['allocationSize']))

if (isset($metadata->sequenceGeneratorDefinition['initialValue']))

Fixed by : https://github.com/doctrine/doctrine2/commit/d5d47222c1dc5ea97ebd8f4c68834fbe4abeb238

Merged into 2.2

ubuntu 11.04, php 5.3.6-13ubuntu3.6

code paste above didn't work out so well. Attached is a .tgz of the sandbox used above.

Fixed

Fixed

Pull request with 3 failing tests (That are fixed by modifications to PersistentCollection in the same commit)

https://github.com/doctrine/doctrine2/pull/336

Fixed

Is this backported to 2.1 as well?

This is intentional to me.
According to code and documentation, PropertyChanged is supposed to operate over updates, never over inserts.

I experimented changing the code to also notify about changed during new and the consequences were very drastic. Internally, propertyChanged creates entityChangesets that implies an UPDATE.

Marking ticket as won't fix.

In that case the Notify strategy is partially broken:
Category entity:

..............

  *\Entity\Category
 *
 * @ORM\Table(name="category")
 * @ORM\Entity
 */
class Category implements TimestampableInterface
 
    /**
     * Caption
     *
     * @var string $caption
     * @ORM\Column(name="caption", type="text", nullable=true)
     */
    protected $caption;
	
    /**
     * Added At
     *
     * @var \DateTime $addedAt
     *
     * @ORM\Column(name="added_at", type="datetime", nullable=true)
     */
    protected $addedAt;

    public function setCaption($caption)
    {
        if ($this->caption != $caption)  {
            $this->_onPropertyChanged('caption', $this->caption, $caption);
            $this->caption = $caption;
        }
    }
	
    public function setAddedAt($addedAt)
    {
        $oldValue = $this->addedAt;
		$this->addedAt = $addedAt;
        if ((is_null($oldValue) || ($oldValue->getTimestamp() != $this->addedAt->getTimestamp()))) {
            $this->_onPropertyChanged('addedAt', $oldValue, $this->addedAt);
        }
    }

    protected function _onPropertyChanged($propName, $oldValue, $newValue)
    {
        if ($this->_listeners) {
            /** @var $listener \Doctrine\Common\PropertyChangedListener */
            foreach ($this->_listeners as $listener) {
                $listener->propertyChanged($this, $propName, $oldValue, $newValue);
            }
        }
    }

    public function addPropertyChangedListener(PropertyChangedListener $listener)
    {
        $this->_listeners[] = $listener;
    }
	
..............	

OnFlush event handler:

..............

        foreach ($uow->getScheduledEntityInsertions() as $entity) {
            if ($entity instanceof TimestampableInterface) {
                $entity->setUpdatedAt(new \DateTime('now'));
                $entity->setAddedAt(new \DateTime('now'));
                $uow->recomputeSingleEntityChangeSet($em->getClassMetadata(get_class($entity)), $entity);
            }
        }
		
...............		

Code that uses the entity:

        $cat = new \Entity\Category();
        // @todo this is a workaround until the http://www.doctrine-project.org/jira/browse/DDC-1775 is resolved
        $cat->addPropertyChangedListener($this->_em->getUnitOfWork());
        $cat->setCaption('Please explain');
        $this->_em->persist($cat);
        $this->_em->flush();

If there is no explicit call to the addPropertyChangedListener method, the caption field gets saved but the $addedAt remains null after flush. The entity does not have the attached listener so UnitOfWork does not know anything about the update that happened in the OnFlush event, and the recomputeSingleEntityChangeSet method skips entities with Notify strategy.

Changing computeScheduleInsertsChangeSets() would solve this, but it would also mean that the notifier gets injected more than once.

    private function computeScheduleInsertsChangeSets()
    {
        foreach ($this->entityInsertions as $entity) {
            $class = $this->em->getClassMetadata(get_class($entity));

            $this->computeChangeSet($class, $entity);

            if ($entity instanceof NotifyPropertyChanged) {
                $entity->addPropertyChangedListener($this);
            }
        }
    }

I think injecting in scheduleForInsert() is ok, but we have to look at potential problems also

Fixed

A related Github Pull-Request [GH-322] was synchronize
https://github.com/doctrine/doctrine2/pull/322

A related Github Pull-Request [GH-322] was synchronize
https://github.com/doctrine/doctrine2/pull/322

A related Github Pull-Request [GH-322] was synchronize
https://github.com/doctrine/doctrine2/pull/322

A related Github Pull-Request [GH-316] was synchronize
https://github.com/doctrine/doctrine2/pull/316

A related Github Pull-Request [GH-316] was synchronize
https://github.com/doctrine/doctrine2/pull/316

A related Github Pull-Request [GH-316] was synchronize
https://github.com/doctrine/doctrine2/pull/316

A related Github Pull-Request [GH-316] was synchronize
https://github.com/doctrine/doctrine2/pull/316

Fixed

A related Github Pull-Request [GH-312] was synchronize
https://github.com/doctrine/doctrine2/pull/312

A related Github Pull-Request [GH-312] was synchronize
https://github.com/doctrine/doctrine2/pull/312

A related Github Pull-Request [GH-312] was synchronize
https://github.com/doctrine/doctrine2/pull/312

A related Github Pull-Request [GH-312] was synchronize
https://github.com/doctrine/doctrine2/pull/312

A related Github Pull-Request [GH-312] was synchronize
https://github.com/doctrine/doctrine2/pull/312

A related Github Pull-Request [GH-312] was synchronize
https://github.com/doctrine/doctrine2/pull/312

A related Github Pull-Request [GH-312] was synchronize
https://github.com/doctrine/doctrine2/pull/312

Changed to using temporary filename + rename

Ubuntu 10.10, syfmony 2.0.4

The same problem, described here
http://stackoverflow.com/questions/7766999/doctrine-2-0-vs-2-1-cascade-remove-onetomany

I know, it's me who answered. But the solution i've found is "ugly".

Fixed formatting

If you add orphanRemoval="true" it works.
But I think doctrine should somehow let the user know about this problem (throw exception, etc.)
Maybe it worths making orphanRemoval=true when cascading=remove?

1. orphanRemoval is not the right solution, it has different semantics. If you want this semantics the solution is ok.
2. Has anyone tried updating to anything above 2.1.2?
3. I cannot reproduce with a test-case against the current master or 2.1.5

I need a better testcase. Attached is the one i am working with.

No feedback given.