from __future__ import unicode_literals import collections import io import struct import zlib from .utils import ( compat_str, ExtractorError, ) def _extract_tags(file_contents): if file_contents[1:3] != b'WS': raise ExtractorError( 'Not an SWF file; header is %r' % file_contents[:3]) if file_contents[:1] == b'C': content = zlib.decompress(file_contents[8:]) else: raise NotImplementedError( 'Unsupported compression format %r' % file_contents[:1]) # Determine number of bits in framesize rectangle framesize_nbits = struct.unpack('!B', content[:1])[0] >> 3 framesize_len = (5 + 4 * framesize_nbits + 7) // 8 pos = framesize_len + 2 + 2 while pos < len(content): header16 = struct.unpack('> 6 tag_len = header16 & 0x3f if tag_len == 0x3f: tag_len = struct.unpack('= 0x80) else b'\x00' return struct.unpack('= 0 resb = reader.read(count) assert len(resb) == count return resb def _read_byte(reader): resb = _read_bytes(1, reader=reader) res = struct.unpack('> 4 methods = {} if kind in [0x00, 0x06]: # Slot or Const u30() # Slot id u30() # type_name_idx vindex = u30() if vindex != 0: read_byte() # vkind elif kind in [0x01, 0x02, 0x03]: # Method / Getter / Setter u30() # disp_id method_idx = u30() methods[self.multinames[trait_name_idx]] = method_idx elif kind == 0x04: # Class u30() # slot_id u30() # classi elif kind == 0x05: # Function u30() # slot_id function_idx = u30() methods[function_idx] = self.multinames[trait_name_idx] else: raise ExtractorError('Unsupported trait kind %d' % kind) if attrs & 0x4 != 0: # Metadata present metadata_count = u30() for _c3 in range(metadata_count): u30() # metadata index return methods # Classes class_count = u30() classes = [] for class_id in range(class_count): name_idx = u30() cname = self.multinames[name_idx] avm_class = _AVMClass(name_idx, cname) classes.append(avm_class) u30() # super_name idx flags = read_byte() if flags & 0x08 != 0: # Protected namespace is present u30() # protected_ns_idx intrf_count = u30() for _c2 in range(intrf_count): u30() u30() # iinit trait_count = u30() for _c2 in range(trait_count): trait_methods = parse_traits_info() avm_class.register_methods(trait_methods) assert len(classes) == class_count self._classes_by_name = dict((c.name, c) for c in classes) for avm_class in classes: u30() # cinit trait_count = u30() for _c2 in range(trait_count): trait_methods = parse_traits_info() avm_class.register_methods(trait_methods) # Scripts script_count = u30() for _c in range(script_count): u30() # init trait_count = u30() for _c2 in range(trait_count): parse_traits_info() # Method bodies method_body_count = u30() Method = collections.namedtuple('Method', ['code', 'local_count']) for _c in range(method_body_count): method_idx = u30() u30() # max_stack local_count = u30() u30() # init_scope_depth u30() # max_scope_depth code_length = u30() code = read_bytes(code_length) for avm_class in classes: if method_idx in avm_class.method_idxs: m = Method(code, local_count) avm_class.methods[avm_class.method_idxs[method_idx]] = m exception_count = u30() for _c2 in range(exception_count): u30() # from u30() # to u30() # target u30() # exc_type u30() # var_name trait_count = u30() for _c2 in range(trait_count): parse_traits_info() assert p + code_reader.tell() == len(code_tag) def extract_class(self, class_name): try: return self._classes_by_name[class_name] except KeyError: raise ExtractorError('Class %r not found' % class_name) def extract_function(self, avm_class, func_name): print('Extracting %s.%s' % (avm_class.name, func_name)) if func_name in avm_class.method_pyfunctions: return avm_class.method_pyfunctions[func_name] if func_name in self._classes_by_name: return self._classes_by_name[func_name].make_object() if func_name not in avm_class.methods: raise ExtractorError('Cannot find function %s.%s' % ( avm_class.name, func_name)) m = avm_class.methods[func_name] def resfunc(args): # Helper functions coder = io.BytesIO(m.code) s24 = lambda: _s24(coder) u30 = lambda: _u30(coder) print('Invoking %s.%s(%r)' % (avm_class.name, func_name, tuple(args))) registers = [avm_class.variables] + list(args) + [None] * m.local_count stack = [] scopes = collections.deque([ self._classes_by_name, avm_class.variables]) while True: opcode = _read_byte(coder) print('opcode: %r, stack(%d): %r' % (opcode, len(stack), stack)) if opcode == 17: # iftrue offset = s24() value = stack.pop() if value: coder.seek(coder.tell() + offset) elif opcode == 18: # iffalse offset = s24() value = stack.pop() if not value: coder.seek(coder.tell() + offset) elif opcode == 36: # pushbyte v = _read_byte(coder) stack.append(v) elif opcode == 42: # dup value = stack[-1] stack.append(value) elif opcode == 44: # pushstring idx = u30() stack.append(self.constant_strings[idx]) elif opcode == 48: # pushscope new_scope = stack.pop() scopes.append(new_scope) elif opcode == 70: # callproperty index = u30() mname = self.multinames[index] arg_count = u30() args = list(reversed( [stack.pop() for _ in range(arg_count)])) obj = stack.pop() if isinstance(obj, _AVMClass_Object): func = self.extract_function(obj.avm_class, mname) res = func(args) stack.append(res) continue elif isinstance(obj, _ScopeDict): if mname in obj.avm_class.method_names: func = self.extract_function(obj.avm_class, mname) res = func(args) else: res = obj[mname] stack.append(res) continue elif isinstance(obj, compat_str): if mname == 'split': assert len(args) == 1 assert isinstance(args[0], compat_str) if args[0] == '': res = list(obj) else: res = obj.split(args[0]) stack.append(res) continue elif isinstance(obj, list): if mname == 'slice': assert len(args) == 1 assert isinstance(args[0], int) res = obj[args[0]:] stack.append(res) continue elif mname == 'join': assert len(args) == 1 assert isinstance(args[0], compat_str) res = args[0].join(obj) stack.append(res) continue raise NotImplementedError( 'Unsupported property %r on %r' % (mname, obj)) elif opcode == 72: # returnvalue res = stack.pop() return res elif opcode == 74: # constructproperty index = u30() arg_count = u30() args = list(reversed( [stack.pop() for _ in range(arg_count)])) obj = stack.pop() mname = self.multinames[index] assert isinstance(obj, _AVMClass) construct_method = self.extract_function( obj, mname) # We do not actually call the constructor for now; # we just pretend it does nothing stack.append(obj.make_object()) elif opcode == 79: # callpropvoid index = u30() mname = self.multinames[index] arg_count = u30() args = list(reversed( [stack.pop() for _ in range(arg_count)])) obj = stack.pop() if mname == 'reverse': assert isinstance(obj, list) obj.reverse() else: raise NotImplementedError( 'Unsupported (void) property %r on %r' % (mname, obj)) elif opcode == 86: # newarray arg_count = u30() arr = [] for i in range(arg_count): arr.append(stack.pop()) arr = arr[::-1] stack.append(arr) elif opcode == 93: # findpropstrict index = u30() mname = self.multinames[index] for s in reversed(scopes): if mname in s: res = s break else: res = scopes[0] stack.append(res[mname]) elif opcode == 94: # findproperty index = u30() mname = self.multinames[index] for s in reversed(scopes): if mname in s: res = s break else: res = avm_class.variables stack.append(res) elif opcode == 96: # getlex index = u30() mname = self.multinames[index] for s in reversed(scopes): if mname in s: scope = s break else: scope = avm_class.variables # I cannot find where static variables are initialized # so let's just return None res = scope.get(mname) stack.append(res) elif opcode == 97: # setproperty index = u30() value = stack.pop() idx = self.multinames[index] obj = stack.pop() obj[idx] = value elif opcode == 98: # getlocal index = u30() stack.append(registers[index]) elif opcode == 99: # setlocal index = u30() value = stack.pop() registers[index] = value elif opcode == 102: # getproperty index = u30() pname = self.multinames[index] if pname == 'length': obj = stack.pop() assert isinstance(obj, list) stack.append(len(obj)) else: # Assume attribute access idx = stack.pop() assert isinstance(idx, int) obj = stack.pop() assert isinstance(obj, list) stack.append(obj[idx]) elif opcode == 115: # convert_ value = stack.pop() intvalue = int(value) stack.append(intvalue) elif opcode == 128: # coerce u30() elif opcode == 133: # coerce_s assert isinstance(stack[-1], (type(None), compat_str)) elif opcode == 160: # add value2 = stack.pop() value1 = stack.pop() res = value1 + value2 stack.append(res) elif opcode == 161: # subtract value2 = stack.pop() value1 = stack.pop() res = value1 - value2 stack.append(res) elif opcode == 164: # modulo value2 = stack.pop() value1 = stack.pop() res = value1 % value2 stack.append(res) elif opcode == 175: # greaterequals value2 = stack.pop() value1 = stack.pop() result = value1 >= value2 stack.append(result) elif opcode == 208: # getlocal_0 stack.append(registers[0]) elif opcode == 209: # getlocal_1 stack.append(registers[1]) elif opcode == 210: # getlocal_2 stack.append(registers[2]) elif opcode == 211: # getlocal_3 stack.append(registers[3]) elif opcode == 212: # setlocal_0 registers[0] = stack.pop() elif opcode == 213: # setlocal_1 registers[1] = stack.pop() elif opcode == 214: # setlocal_2 registers[2] = stack.pop() elif opcode == 215: # setlocal_3 registers[3] = stack.pop() else: raise NotImplementedError( 'Unsupported opcode %d' % opcode) avm_class.method_pyfunctions[func_name] = resfunc return resfunc