Invalid Bidirectional Association Mapping with Mapped Superclass in Class Hierarchy

[bobdercole]

Bug Report

Q	A
BC Break	no
Version	2.11.1

Summary

I receive a mapping exception after defining an inverse-side association on an abstract entity. A mapped superclass extends the abstract entity. A concrete entity extends the mapped superclass.

Current behavior

ClassMetadataFactory is currently throwing an exception if there is an inverse-side association on a mapped superclass.

orm/lib/Doctrine/ORM/Mapping/ClassMetadataFactory.php

Line 381 in e8e61cb

if ($mapping['type'] & ClassMetadata::TO_MANY && ! $mapping['isOwningSide']) {

This is correct, as stated in the documentation. However, I believe this limitation should be imposed only if the association is on the mapped superclass itself.

How to reproduce

• Define an abstract entity (A).
• Define a mapped superclass (B) extends A.
• Define a concrete entity (C) extends B.
• Define another entity (D) that has a many-to-one relationship with A.

I will follow-up with a failing test case.

Expected behavior

I think the above mapping should be valid since the association is not defined directly on the mapped superclass.

[bobdercole]

Here is my failing test case: #9517.

My proposed fix is to add a condition to ClassMetadataFactory to only throw the exception if the association is directly on the mapped superclass. Maybe something like this?

if ($mapping['type'] & ClassMetadata::TO_MANY && ! $mapping['isOwningSide'] && $parentClass->name === $mapping['sourceEntity']) {
    throw MappingException::illegalToManyAssociationOnMappedSuperclass($parentClass->name, $field);
}

[mpdude]

Related to #8415 (as per #9517).

[mpdude]

Please check if this is already fixed in the current 2.14.x branch

[mpdude]

You might also want to subscribe to #10398

[bobdercole]

Please check if this is already fixed in the current 2.14.x branch

Yes! It works now. Thank you for all the work you've done to resolve it!

I suppose we should keep this open until the regression test PR is merged.

[mpdude]

@greg0ire maybe a low hanging fruit:

Merge #10453 and then close the issues and PRs linked from it