Is there a PHPCS or stan rule for the __clone() change maybe?

Is there a PHPCS or stan rule for the __clone() change maybe?

Couldn't find anything after a quick search. psalm does report it though :)

psalm does report it though :)

gotta re-add psalm then :D

Slevomat's ReturnTypeHintSniff at line 213:

$methodsWithoutVoidSupport = ['__construct' => true, '__destruct' => true, '__clone' => true];

This means the sniff will not require : void on __clone(), __construct(), or __destruct().

But not sure if it would also remove or alert the additional one.

Seems like a simple sniff addition: php-collective/code-sniffer#48
I Just made it a feature flag for BC for now, but it can sure be added as a real new sniff as well.

gotta re-add psalm then :D

Maybe we should. It's pretty fast and does seem to be quite good at catching generics/templates related issues currently. For e.g. it reported the docblock type mismatch for PaginatedResultSet::__construct() which phpstan didn't.

it reported the docblock type mismatch for PaginatedResultSet::__construct() which phpstan didn't.

Is this an actual bug in PHPStan or is this something that only gets reported by PHPStan by higher levels or extensions?

If its a bug it would be awesome if it could be made minimally reproducable and report it to phpstan.

Is this an actual bug in PHPStan or is this something that only gets reported by PHPStan by higher levels or extensions?

We have disabled/ignored the missingType.generics identifier. So it's possible that's why it doesn't check the validity of related annotations.