Correct hydration of resultsets coming from DQL queries where the root of the selection is aliased and also the root of an inheritance

[SirWaddles]

Resolves Issue #6362

Added a test case for reproducability.

[SirWaddles]

The main reason for test is to check that the Undefined Property error does not occur, and if you run the test without the fix, it will error out. It does however also make some assertions about the format of the result which may be questionable.

The resultset only has one row, yet the hydrated result has two. My assumption was that this is intentional as this change did not break any tests, but it does seem odd to me.

[SirWaddles]

Might be worth running the tests again, only one job failed and it had just stalled before even getting to my code. I have no idea why my change would cause only PHP7.1/MariaDB to halt.

[Ocramius]

@SirWaddles restarted

[lcobucci]

The rest LGTM, what do you think @Ocramius?

[lcobucci]

@group GH6362 missing

[lcobucci]

@SirWaddles I've rebased your branch to sync with the latest updates and group the commits. Patch LGTM but I'd love to have @guilhermeblanco's and/or @Ocramius' opinion just to double check.

Thanks for the contribution 😉

[lcobucci]

Quite easy to backport for v2.5.7 too...

[Ocramius]

Good to go! Will also backport 👍

Thanks @SirWaddles, pristine work.

[Ocramius]

Hah, I backported to 2.5 via 04fc7a9, but it will obviously fail due to lack of PHP 5.4 compliance =_=

[Ocramius]

Should be corrected in d2c805b (see https://travis-ci.org/doctrine/doctrine2/builds/227851219)