[php] Add <tmp> assignment chain for call chains / property accesses

[AndreiDreyer]

Attempting this issue revealed a number of things that were incorrectly implemented and that had to be fixed as a prerequisite:

• Fixed shadowing in the test suite of builtin_functions.txt in the resources
• Changed how calls are represented - these include a base and receiver in a similar manner to other interpreted languages in Joern
• Due to the nature of PHP, unless a dynamic call is dispatched within the scope of a class, there will not be a $this receiver.
• Built-in functions are statically dispatched
• Field accesses were found to use an identifier in place of a field identifier; this was rectified
• Chained calls use aliasing with temporary variables to ensure that field accesses and calls only occur once, as per the source code.

TODO:

• This broke some type recovery tests and may impact downstream users.
• The way receivers may often be an identifier referring to the call name, the local pass created a local for these - investigate if this is canonical
• Classes need to be split into an instance class and a singleton "meta" class, due to how object instances carry references to static functions. This needs to be referred to when static calls are made on object instances.
• Call bindings need to be created where possible.

[ml86]

There must have been a misunderstanding in our last talk.
The base is the "this" argument and this is just an IDENTIFIER a in this case.
You can reference the same IDENTIFIER node as receiver.
I just red through the PHP docu and played around with it a little bit and i think we can model the PHP classes and the call dispatch in a classic vtable style.

[ml86]

Please check explicitly for the existence of the receiver.

[ml86]

The hole tmp variable creation can be removed because in PHP the receiver and the this argument are always the same and you can just reference the same AST subtree.

[DavidBakerEffendi]

So, should we remove the base altogether? From what I understand, if the base and receiver are the same AST subtree the tmp stuff is there to reduce duplication between the two.

[ml86]

Please add one tree for the this argument and point to one argument and one receiver edge to it. No extra base tree.