Each function contains a list of all instructions inside, thus we should consider only OP_CALL that has already been executed. As the list is browsed from bottom to top, no more checks are necessary. The first match is what we
are looking for.
Only executed bytecode should be registered for debugging. In this way we do not register all emitted bytecode and save some memory. What is more, this change registers the bytecode in correct executing order including jumps.
Previously all bytecode generated by compiler has been saved what made the bytecode to be saved in compile, not execute order.
This commit introduces ne debug feature. The PH7 Engine limits the VM dump to the global scope. Since Aer Script is fully object-oriented language the dump option contains only information about last call of OP_DONE. This change,
forces the VM to store all instructions set in a global container when debugging is enabled, thus providing information the dump of whole script parse.
This is a new memory subsystem implementing heap calculations as well as new builtin functions:
* get_memory_usage()
* get_memory_peak_usage()
* get_memory_limit()
It also allows to set an upper memory limit, ensuring that processed script will not be able to allocate more memory from OS.
New subsystem is based on work done in 'memory_limit' branch. Big thanks to devnexen!
This finally fixes#25.
Aer is object-oriented language and the code it includes is executed in the global scope. Global scope can contain only classes and namespaces and they cannot be re-declared. Thus every file can be included just once. We can
leave this to programmer, or take care about it on interpreter side. With this commit, include() and require() builtin functions allows to load file just once. Thanks to that, there is no need for include_once() and
require_once(). Done in scope of #30.