Add type check user cpt array

[cbravobernal]

What?

Fixes #45240

Why?

If $user_cpt is not an array. array_key_exists( 'post_content', $user_cpt ) function can throw a Fatal error or a warning.

How?

Add the is_array() check to the conditional.

Testing Instructions

Screenshots or screencast

[scruffian]

If $user_cpt was null I think this would still create a notice?

[ockham]

If $user_cpt was null I think this would still create a notice?

TBH I don't quite see why? is_array( $user_cpt ) seems perfectly fine with $user_cpt being null, and that check should short-circuit the whole if conditional and skip the array_key_exists one. Or am I missing something? 🤔

[ockham]

FWIW, I don't mind adding the in_array check, since it seems like it can't have any negative side effects.

Ideally, I'd like to be able to reproduce the issue (see) and verify the fix.

Finally, I'd like to understand which code path leads to $user_cpt being null here? Seems like get_user_data_from_wp_global_styles would fall back to an empty array pretty much always 🤔

[ockham]

Okay, LGTM. See #45275 (comment) for some more thoughts.

[desrosj]

I'm a bit hesitant to add a type check here without understanding what's actually going on at a deeper level. I worry this could be silencing an error that would alert someone of a problem in their code.

I'm a little confused as to the steps I should be using to try and reproduce this. Is the contributing factor using unregister_post_type( 'wp_global_styles' ) (which in my opinion is not expected and can't be guaranteed to result in a bug-free experience), or deleting the wp_global_styles post that holds the site's global styles (as @ockham's testing on WordPress/wordpress-develop#3525 reflects)?

[spacedmonkey]

I think a unit test would be useful here. But the change looks good to me.

[ockham]

Update: Per this discussion, we've decided not to include this in 6.1. Tentatively bumping to 6.1.1.

[Mamaduka]

This is no longer an issue in the plugin or in the core. See: #45240 (comment).

[Mamaduka]

@c4rl0sbr4v0, I think we can close this and the core backport PR.