My apologies the code would be better read as:

<?php

//No constructor
class Foo {}
$a = null;
$f = new Foo($a = 123);
var_dump($a);
// NULL as expected

//PHP 5 style constructor
class Foo2 { function __construct(){} }
$a = null;
$f = new Foo2($a = 123);
var_dump($a);
// int(123) what????

// PHP 4 style constructor
class Foo3 { function Foo3(){} }
$a = null;
$f = new Foo3($a = 123);
var_dump($a);
// int(123) what????

> specifying a default value when calling the constructor
There is no such thing. What you are doing is exactly what the summary says: variable assignment ($a = 123) and then constructing the objects (with the value 123). It is - should be - identical to writing

$a = null;
$a = 123;
$f = new Foo(123);

The bug I see here is that the first example does not do the assignment.
http://3v4l.org/IFS07

Ok fair enough, maybe more bad practice to assign a value to a variable in object initialization.

But yes there should be consistency regardless upon the existence of a constructor.

This isn't a bug, it's extremely beneficial behaviour you are just noticing:

https://gist.github.com/krakjoe/c6bcbb4e6d019c9297db

Thanks for taking the time to make PHP better :)

I am sorry Krakjoe but I cannot see why the code fragment you have provided is relevant to this report.

In the code I mention a variable in the main scope only changes value when a constructor is present. Your code is simply comparing values.

The behavior specified in the bug report causes inconsistent behavior and has the potential to introduce bugs.

Why should it matter if a constructor is present for a variable to change value in this manner?