Signature is valid if there are less args

[nikic]

Patch for http://markmail.org/message/ejgmqjnjnbmfnuzv.

[lstrojny]

YES, PLEASE! :)

[dsp]

This needs an RFC.

Also think about the following:

error_reporting(E_STRICT);
Class X
{
    function cleanup($recurse, $unlink=False) {
    }
}

class Z extends X
{
    function cleanup($unlink) {
    }
}

$x = getInstanceOfX(); // returns Z
$x->cleanup(true); // woops..suddenly everything gone

[Trainmaster]

From a userland perspective:

class Parents
{
    public function doSomething($x)
    {
    }
}

class ChildA extends Parents
{
    public function doSomething()
    {
    }
}
// Strict Standards: Declaration of ChildC::doSomething() should be compatible with that of Parents::doSomething()

class ChildB extends Parents
{
    public function doSomething($x, $y)
    {
    }
}
// Strict Standards: Declaration of ChildB::doSomething() should be compatible with that of Parents::doSomething()

class ChildC extends Parents
{
    public function doSomething($x)
    {
    }
}
// Works

No matter if ChildA, ChildB oder ChildC, doSomething() should always be callable with one argument.

Another thing, see the following code:

interface BazInterface
{
    public function doSomething($x);
}

class BazA implements BazInterface
{
    public function doSomething()
    {
    }
}
// Fatal error: Declaration of BazA::doSomething() must be compatible with that of BazInterface::doSomething()

class BazB implements BazInterface
{
    public function doSomething($x, $y)
    {
    }
}
// Fatal error: Declaration of BazB::doSomething() must be compatible with that of BazInterface::doSomething()

Will this affect interfaces, too?

[nikic]

In light of a possible addition of named parameters in the future it's likely better not to allow this. Withdrawing this proposal.