Add generic return type for the PHP 8.5 clone() function

[delolmo]

Summary

PHP 8.5 adds a function form of clone, clone(object $object, array $withProperties = []), which PR #11604 registered in CallMap_85 with a plain object return type. That loses the concrete type of the cloned object, so clone($user, [...]) is inferred as object rather than User.

This adds a templated stub to stubs/Php85.phpstub so the function form returns the same type as its first argument:

/**
 * @psalm-template T of object
 * @param T $object
 * @param array<string, mixed> $withProperties
 * @return T
 */
function clone(object $object, array $withProperties = []): object {}

This mirrors what CloneAnalyzer already does for the clone $x operator form (it preserves the operand's type).

Tests

Adds a case to tests/CloneTest.php (pinned to php_version: 8.5) asserting clone($a, [...]) is inferred as the argument's class.

Notes

I split this out from the array_first/array_last/array_any/array_all generics (separate PR) because clone is a reserved keyword and this needs your eyes.

Routing — CloneAnalyzer handles the Clone_ AST node (the clone $x operator). I'm assuming the two-argument function form clone($x, [...]) is parsed as a regular call and resolves against the CallMap/stub (consistent with #11604 adding clone to CallMap_85). If it is instead routed through Clone_/CloneAnalyzer, this stub won't apply and the fix belongs there or in a return-type provider instead.

Placement — the stub lives in stubs/Php85.phpstub, which is only registered when analysis_php_version_id >= 8_05_00. This gates the function to PHP 8.5+ so Psalm doesn't treat clone(...) as a valid call when analyzing code targeting earlier versions (an earlier revision placed it in CoreGenericFunctions.phpstub, which is loaded unconditionally — fixed per Cursor Bugbot's review).

[delolmo]

Heads up for reviewers: the failing ci/circleci: test-with-real-projects check is not related to this change. It fails at the Analyse PHPUnit step on a pre-existing baseline drift (psalm dev reporting ImpureFunctionCall/ImpureMethodCall on get_loaded_extensions(), Duration::fromSecondsAndNanoseconds, TestMethod::hasDataFromDataProvider, etc., plus unused-baseline entries) — none of which touch clone. The same step is currently red on unrelated PRs (e.g. #11879, #11878). The macOS build job is also an unrelated shivammathur/setup-php infra flake ("Could not setup PHP 8.5"); the Linux build (Psalm self-analysis) and the full unit-test matrix including PHP 8.5 pass.

Two open questions for maintainers are noted in the PR description: (1) whether the two-argument clone($x, [...]) form resolves against the CallMap/stub rather than being routed through Clone_/CloneAnalyzer, and (2) preferred placement (CoreGenericFunctions.phpstub vs version-gated stubs/Php85.phpstub).

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit e9d13ed. Configure here.}