Christophe Coevoet

@VincentLanglet You could do `$project = $this->em->find($id);` again after your try/catch. If the EM was not reset in the meantime, `find()` will not hit the DB (as that id is...

Also, even if `$project` is not in an inconsistent state, `$url` would still be in the unit of work if it were not cleared, and so the next flush would...

the failing tests are because the testsuite currently relies on unmapped public properties to track the usage of postLoad listeners. For doctrine/common proxies, public properties were unsupported and were not...

I wanted to refactor those tests to store the initialization tracking externally, but I haven't done it yet (I kinda forgot this PR)

I haven't had the time to go back at it to update the tests. But I saw that https://github.com/doctrine/orm/pull/11853 had similar needs to update the tests, so I might be...

@ronanguilloux any change to get this merged ?

Note that this would probably also be beneficial for stubs. I'm almost sure that PHP does not consider the `RoundingMode` enum as exhaustive for instance, allowing future RFCs to add...

I cannot really show this in phpstan.org/try as the semantic relies on differentiating project code from vendor code. I'm proposing an code snippet below, defining 2 namespaces. We should imagine...

@ondrejmirtes for your **own** enums, I totally agree with you (and that's why own enums are always treated as exhaustive in that proposal). However, if a library says that it...

@vojtech-dobes for the case where the consumer code is in the final project (responsible for updating dependencies), running phpstan during the upgrade can indeed have you covered in both cases....