Fix narrowing of superglobals #3901

I did not take over the changes where the expressions are passed over through classes, functions and namespaces in scopes. Those would deal with some edge cases, but not sure if it's worth.

fyi @staabm

herndlm

herndlm commented

Mar 24, 2025

View reviewed changes

tests/PHPStan/Analyser/nsrt/superglobals.php Outdated

Comment on lines 47 to 48

assertType('array', $GLOBALS);

assertNativeType('array<mixed>', $GLOBALS);

Copy link

Contributor Author

@herndlm herndlm Mar 24, 2025 •

edited

Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

these 2 fail without the adaptions here

@herndlm

Copy link

Contributor Author

herndlm commented Mar 24, 2025

phpstan-doctrine failures are unrelated, those seemed to have appeared because of recent phpstan improvements, opened phpstan/phpstan-doctrine#650 for it

@herndlm herndlm mentioned this pull request

Mar 24, 2025

Bad assumption about array key phpstan/phpstan#12779

Closed

mvorisek

mvorisek reviewed

Mar 24, 2025

View reviewed changes

tests/PHPStan/Rules/Comparison/StrictComparisonOfDifferentTypesRuleTest.php

]);

}

public function testBug12772(): void

Copy link

Contributor

@mvorisek mvorisek Mar 24, 2025 •

edited

Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please add also $_ENV['CI'] test explicitly, as I want to make sure GitHub CI environment variable is never assumed to exist on runtime - https://github.com/atk4/ui/blob/235416b31066827226e462f1916d9fc99c0d14b4/demos/init-db.php#L556. And thank you very much for looking into this issue!

Our reproducible environment is the same as here, GitHub CI, so as long as the tests passes here, our should pass then too.

Copy link

Contributor Author

@herndlm herndlm Mar 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I get your point, but adaptions here are not related to any environment variables. If Ondřej is fine with this, I expect it to be merged very soon and then we should also see the effects

staabm

staabm approved these changes

Mar 24, 2025

View reviewed changes

Copy link

Contributor

@staabm staabm left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

thanks for looking into it

ondrejmirtes

ondrejmirtes requested changes

Mar 24, 2025

View reviewed changes

src/Analyser/MutatingScope.php Outdated

$superglobalType = new ArrayType(new BenevolentUnionType([new IntegerType(), new StringType()]), new MixedType(true));

$this->expressionTypes[$varExprString] = ExpressionTypeHolder::createYes(new Variable($variableName), $superglobalType);

$this->nativeExpressionTypes[$varExprString] = ExpressionTypeHolder::createYes(new Variable($variableName), $superglobalType);

Copy link

Member

@ondrejmirtes ondrejmirtes Mar 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't get why the fix looks like this.

This is polluting $this->expressionTypes with any superglobal in the analysed code.
It's mutating Scope in a getter method which is a big red-flag.

Copy link

Contributor Author

@herndlm herndlm Mar 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

good point, I don't understand how this even worked before without adding an expression tbh :P the idea of setting the expression for proper narrowing came from #3762 (comment) basically.

not sure if relevant for the bug or fix, but after the expression is created it is dealing with late resolveable types still in the same method in

phpstan-src/src/Analyser/MutatingScope.php

Line 566 in a11741d

return TypeUtils::resolveLateResolvableTypes($this->expressionTypes[$varExprString]->getType());

Copy link

Contributor Author

@herndlm herndlm Mar 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

looks like I can't get around step-debugging this issue :D I assume the expressions are just overriding something that is incorrectly set and retained or so? let's see..