- to retreive the zval **, thus allowing it to use the convert_to_*_ex()
- family of functions to do type conversions without effecting the value in
- the engine itself. (Josh Fuhs <fuhs@purdue.edu>)
- Enabling it by default so that it gets tested. We should decide before
beta 2 if we want to revert back to malloc or not.
- Thanks to Sebastian for benchmarking it
#
# Declaring a method private and final would only be an error for an abstract
# class. But at the moment the method is defined and it's modifiers are checked
# we do not know whether or not we have an abstract class. It could already be
# abstract but it also become abstract later.
#
# Since i made the mistake in first place i remove the check now.
#
# Providing the correct test would slow down the compiler becuase we'd have to
# iterate through all methods on all abstract classes and check for this. I
# guess we can live without. Or does anybody wants this to be implemented ?
hidden destructors.
#
# If we set the error level to E_ERROR what we must to to be correct and an
# object needs to be automatically destructed in shutdown process then the
# exit would cause memory corruption and a SEGV.
Heads up - this will break syntactical compatiblity, return($foo) will
not work with functions that return references - return $foo should be used
instead. It never worked well before, and caused all sorts of odd bugs.
It *might* be possible to support this specifically, albeit unlikely
# The default behaviour is to copy all properties with all current values
# from the old object. But if __clone is overwritten then only the default
# properties are cloned with their correct default values. So we keep
# the type system intact and also allow real __clone overwriting now.