This does a compile time transformation of ``iterable`` into ``Traversable|array`` which simplifies some of the LSP variance handling.
The arginfo generation script from stubs is updated to produce a union type when it encounters the type ``iterable``
Extension functions which do not regenerate the arginfo, or write them manually are still supported by mimicking the compile time transformation while registering the function.
Type Reflection is preserved for single ``iterable`` (and ``?iterable``) to produce a ReflectionNamedType with name ``iterable``, however usage of ``iterable`` in union types will be converted to ``array|Traversable``
From now on, we always display the given object's type instead of just reporting "object".
Additionally, make the format of return type errors match the format of argument errors.
Closes GH-5625
PHP requires integer typehints to be written "int" and does not
allow "integer" as an alias. This changes type error messages to
match the actual type name and avoids confusing messages like
"must be of the type integer, integer given".
We currently don't show the argument at which the error actually
occured in the trace - should probably either add it or don't
display args on incomplete frames altogether, otherwise this'll
probably be confusing.
Also fix a bug with return types where "self" was rejected inside
a class, but not on a method.
Fallout: A couple of tests changed to more generic error messages.
* The class names false, true and null are now reserved.
* The code dealing with reserved class names is now decoupled
from scalar type hint handling. It also includes self, parent,
and static, which are class names which were already reserved
previously.
* Reuse existing messages for reserved class names.
Fallout: class_alias() can no longer alias self, parent and static.
However this never really worked in the first place, as the test
which was testing this shows.
a) The SAME_FAKE_TYPE check should always occur on the dereferenced
value, otherwise we'll end up separating large referenced arrays.
b) For by-val return with an rc=1 reference, the SEPARATE_ZVAL
wouldn't do anything. Add a ZVAL_UNREF for this case.
I don't like this fix - there must be some better way to express
this.
This is allowed as per the return types RFC. The test for this
behavior happened to use an internal class without arginfo, which
is why this was not properly implemented.
