I am adding a obvious fix, but I am not sure about it, as we see tests failling now - these tests obviously expected a trustfully phpdoc. do we agree these other tests expectations are wrong as they shouldn't trust the phpdoc in the same way?

I wonder whether we should/want to trust a array-shape when its produced by a php-src builtin function/method, but not when its based on a userland phpdoc (and whether we are able to differentiate the 2)? created a PHPStan upstream discussion.

In this case, can we detect the type is being compared from an array shale?
If so, we should skip it.

I'd keep it for other cases, as so far the rule has beem reliable.

thanks for the suggestion. adapted accordingly

Looking good, thank you @staabm