The original example from the issue still doesn't work: it says

Psalm\Exception\CodeException : InvalidArgument - src/somefile.php:27:38 - 
Argument 1 of takesIntToHolder expects Closure(int):Holder<C>, but impure-Closure(int):Holder<static> provided

I found these snippets:

<?php

/** @template T */
class Holder
{
    /** @param T $value */
    public function __construct(public $value) {}
}

abstract class A
{
    final public function __construct(public int $i) {}
    
    /** @return Holder<static> */
    public static function create(int $i): Holder
    {
        return new Holder(new static($i));
    }
}

class C extends A
{
}

/** @param \Closure(int):Holder<C> $_ */
function takesIntToHolder(\Closure $_): void {}

takesIntToHolder(C::create(...));

Psalm output (using commit 24f7920):

INFO: MixedArgumentTypeCoercion - 28:18 - Argument 1 of takesIntToHolder expects Closure(int):Holder<C>, parent type Closure provided

Okay, so I managed to fix my exact case in the last commit, however it still feels like a hack and duplicated efforts.

Needless to say it will suffer from other static problems which were fixed or maybe will be fixed in future for other features of the language.

Perhaps the proper way to implement the first-class callables is first to desugar them into a closure, or maybe a Closure::fromCallable(<xxx>) expression as per RFC?

It's one of the feature that was implemented after Matt departed from the project, so sadly, it's possible it was added without a "big picture" mindset at the time.
I didn't work very much on first class callable either so I wouldn't be a good adviser on that.

Frankly, if you feel refactoring it would be best and you're up for it, go ahead! Now is the best moment to do so, Psalm 5 was not yet released so implementation changes won't bother anyone depending on it

Ok, however obviously I have even less vision than you do, by a lot.

Each such fix for me is a debugging exercise, and frankly I only fix those problems which I encounter myself, to solve my immediate problem :P

I suggest to merge this one then, but with the proper refactoring kept in mind, WDYT?

Each such fix for me is a debugging exercise

There's a few modules in Psalm I know and I can pinpoint some issues, but aside from that, I'm really the same. For example, I'm pretty sure you know more than me about static issues because I never went check how it works and it would take me a lot of time too :)

I'm up to merge this :)

I'm up to merge this :)

@orklah pingie 👉👈

Thanks!