def uncompyle(version, co, out=None, showasm=False, showast=False,
timestamp=None, showgrammar=False, code_objects={}):
"""
disassembles and deparses a given code block 'co'
"""
assert iscode(co)
# store final output stream for case of error
real_out = out or sys.stdout
print('# Python %s (decompiled from Python %s)' % (version, PYTHON_VERSION),
file=real_out)
if co.co_filename:
print('# Embedded file name: %s' % co.co_filename,
file=real_out)
if timestamp:
print('# Compiled at: %s' % datetime.datetime.fromtimestamp(timestamp),
file=real_out)
try:
pysource.deparse_code(version, co, out, showasm, showast, showgrammar,
code_objects=code_objects)
except pysource.SourceWalkerError as e:
# deparsing failed
if real_out != out:
print(e, file=real_out)
raise
def python_parser(version, co, out=sys.stdout, showasm=False,
parser_debug=PARSER_DEFAULT_DEBUG):
assert iscode(co)
from uncompyle6.scanner import get_scanner
scanner = get_scanner(version)
tokens, customize = scanner.disassemble(co)
if showasm:
for t in tokens:
print(t)
p = get_python_parser(version, parser_debug)
return parse(p, tokens, customize)
def deparse_code(version,
co,
out=StringIO(),
showasm=False,
showast=False,
showgrammar=False):
assert iscode(co)
# store final output stream for case of error
scanner = get_scanner(version)
tokens, customize = scanner.disassemble(co)
tokens, customize = scanner.disassemble(co)
if showasm:
for t in tokens:
print(t)
debug_parser = dict(PARSER_DEFAULT_DEBUG)
debug_parser['reduce'] = showgrammar
# Build AST from disassembly.
# deparsed = pysource.FragmentsWalker(out, scanner, showast=showast)
deparsed = FragmentsWalker(version,
scanner,
showast=showast,
debug_parser=debug_parser)
deparsed.ast = deparsed.build_ast(tokens, customize)
assert deparsed.ast == 'stmts', 'Should have parsed grammar start'
del tokens # save memory
# convert leading '__doc__ = "..." into doc string
assert deparsed.ast == 'stmts'
deparsed.mod_globs = pysource.find_globals(deparsed.ast, set())
# Just when you think we've forgotten about what we
# were supposed to to: Generate source from AST!
deparsed.gen_source(deparsed.ast, co.co_name, customize)
deparsed.set_pos_info(deparsed.ast, 0, len(deparsed.text))
deparsed.fixup_parents(deparsed.ast, None)
for g in deparsed.mod_globs:
deparsed.write('# global %s ## Warning: Unused global' % g)
if deparsed.ERROR:
raise deparsed.ERROR
return deparsed
def python_parser(version, co, out=sys.stdout, showasm=False,
parser_debug=PARSER_DEFAULT_DEBUG):
assert iscode(co)
from uncompyle6.scanner import get_scanner
scanner = get_scanner(version)
tokens, customize = scanner.disassemble(co)
if showasm:
for t in tokens:
print(t)
# For heavy grammar debugging
parser_debug = {'rules': True, 'transition': True, 'reduce' : True}
p = get_python_parser(version, parser_debug)
return parse(p, tokens, customize)
def listcomprehension_walk3(self, node, iter_index, code_index=-5):
"""List comprehensions the way they are done in Python3.
They're more other comprehensions, e.g. set comprehensions
See if we can combine code.
"""
p = self.prec
self.prec = 27
code = node[code_index].attr
assert iscode(code)
# Or Code3
code = Code(code, self.scanner, self.currentclass)
# assert isinstance(code, Code)
ast = self.build_ast(code._tokens, code._customize)
self.customize(code._customize)
ast = ast[0][0][0][0][0]
n = ast[iter_index]
assert n == 'list_iter'
# find innermost node
while n == 'list_iter': # list_iter
n = n[0] # recurse one step
if n == 'list_for':
designator = n[2]
n = n[3]
elif n == 'list_if':
# FIXME: just a guess
designator = n[1]
n = n[2]
elif n == 'list_ifnot':
# FIXME: just a guess
designator = n[1]
n = n[2]
assert n == 'lc_body', ast
self.preorder(n[0])
self.write(' for ')
start = len(self.f.getvalue())
self.preorder(designator)
self.set_pos_info(designator, start, len(self.f.getvalue()))
self.write(' in ')
start = len(self.f.getvalue())
node[-3].parent = node
self.preorder(node[-3])
self.set_pos_info(node[-3], start, len(self.f.getvalue()))
# self.preorder(ast[iter_index])
self.prec = p
def listcomprehension_walk3(self, node, iter_index, code_index=-5):
"""List comprehensions the way they are done in Python3.
They're more other comprehensions, e.g. set comprehensions
See if we can combine code.
"""
p = self.prec
self.prec = 27
code = node[code_index].attr
assert iscode(code)
# Or Code3
code = Code(code, self.scanner, self.currentclass)
ast = self.build_ast(code._tokens, code._customize)
self.customize(code._customize)
ast = ast[0][0][0][0][0]
n = ast[iter_index]
assert n == 'list_iter'
# find innermost node
while n == 'list_iter':
n = n[0] # recurse one step
if n == 'list_for':
designator = n[2]
n = n[3]
elif n == 'list_if':
# FIXME: just a guess
designator = n[1]
n = n[2]
elif n == 'list_if_not':
# FIXME: just a guess
designator = n[1]
n = n[2]
assert n == 'lc_body', ast
self.preorder(n[0])
self.write(' for ')
start = len(self.f.getvalue())
self.preorder(designator)
self.set_pos_info(designator, start, len(self.f.getvalue()))
self.write(' in ')
start = len(self.f.getvalue())
node[-3].parent = node
self.preorder(node[-3])
self.set_pos_info(node[-3], start, len(self.f.getvalue()))
# self.preorder(ast[iter_index])
self.prec = p
def python_parser(version,
co,
out=sys.stdout,
showasm=False,
parser_debug=PARSER_DEFAULT_DEBUG):
assert iscode(co)
from uncompyle6.scanner import get_scanner
scanner = get_scanner(version)
tokens, customize = scanner.disassemble(co)
if showasm:
for t in tokens:
print(t)
p = get_python_parser(version, parser_debug)
return parse(p, tokens, customize)
def deparse_code(version, co, out=StringIO(), showasm=False, showast=False,
showgrammar=False):
assert iscode(co)
# store final output stream for case of error
scanner = get_scanner(version)
tokens, customize = scanner.disassemble(co)
tokens, customize = scanner.disassemble(co)
if showasm:
for t in tokens:
print(t)
debug_parser = dict(PARSER_DEFAULT_DEBUG)
if showgrammar:
debug_parser['reduce'] = showgrammar
debug_parser['errorstack'] = True
# Build AST from disassembly.
# deparsed = pysource.FragmentsWalker(out, scanner, showast=showast)
deparsed = FragmentsWalker(version, scanner, showast=showast, debug_parser=debug_parser)
deparsed.ast = deparsed.build_ast(tokens, customize)
assert deparsed.ast == 'stmts', 'Should have parsed grammar start'
del tokens # save memory
# convert leading '__doc__ = "..." into doc string
assert deparsed.ast == 'stmts'
deparsed.mod_globs = pysource.find_globals(deparsed.ast, set())
# Just when you think we've forgotten about what we
# were supposed to to: Generate source from AST!
deparsed.gen_source(deparsed.ast, co.co_name, customize)
deparsed.set_pos_info(deparsed.ast, 0, len(deparsed.text))
deparsed.fixup_parents(deparsed.ast, None)
for g in deparsed.mod_globs:
deparsed.write('# global %s ## Warning: Unused global' % g)
if deparsed.ERROR:
raise deparsed.ERROR
return deparsed
def disco(version, co, out=None):
"""
diassembles and deparses a given code block 'co'
"""
assert iscode(co)
# store final output stream for case of error
real_out = out or sys.stdout
print('# Python %s' % version, file=real_out)
if co.co_filename:
print('# Embedded file name: %s' % co.co_filename, file=real_out)
scanner = get_scanner(version)
tokens, customize = scanner.disassemble(co)
for t in tokens:
print(t, file=real_out)
print(file=out)
def comprehension_walk(self, node, iter_index, code_index=-5):
p = self.prec
self.prec = 27
if hasattr(node[code_index], 'attr'):
code = node[code_index].attr
elif hasattr(node[1][1], 'attr'):
code = node[1][1].attr
else:
assert False
assert iscode(code)
code = Code(code, self.scanner, self.currentclass)
ast = self.build_ast(code._tokens, code._customize)
self.customize(code._customize)
ast = ast[0][0][0]
n = ast[iter_index]
assert n == 'comp_iter'
# find innermost node
while n == 'comp_iter':
n = n[0] # recurse one step
if n == 'comp_for': n = n[3]
elif n == 'comp_if': n = n[2]
elif n == 'comp_ifnot': n = n[2]
assert n == 'comp_body', ast
self.preorder(n[0])
self.write(' for ')
start = len(self.f.getvalue())
designator = ast[iter_index-1]
self.preorder(designator)
self.set_pos_info(ast[iter_index-1], start, len(self.f.getvalue()))
self.write(' in ')
start = len(self.f.getvalue())
node[-3].parent = node
self.preorder(node[-3])
self.set_pos_info(node[-3], start, len(self.f.getvalue()))
start = len(self.f.getvalue())
self.preorder(ast[iter_index])
self.set_pos_info(iter_index, start, len(self.f.getvalue()))
self.prec = p
def disco(version, co, out=None):
"""
diassembles and deparses a given code block 'co'
"""
assert iscode(co)
# store final output stream for case of error
real_out = out or sys.stdout
print('# Python %s' % version, file=real_out)
if co.co_filename:
print('# Embedded file name: %s' % co.co_filename,
file=real_out)
scanner = get_scanner(version)
tokens, customize = scanner.disassemble(co)
for t in tokens:
print(t, file=real_out)
print(file=out)
def uncompyle(version,
co,
out=None,
showasm=False,
showast=False,
timestamp=None,
showgrammar=False,
code_objects={}):
"""
disassembles and deparses a given code block 'co'
"""
assert iscode(co)
# store final output stream for case of error
real_out = out or sys.stdout
print('# Python %s (decompiled from Python %s)' %
(version, PYTHON_VERSION),
file=real_out)
if co.co_filename:
print('# Embedded file name: %s' % co.co_filename, file=real_out)
if timestamp:
print('# Compiled at: %s' % datetime.datetime.fromtimestamp(timestamp),
file=real_out)
try:
pysource.deparse_code(version,
co,
out,
showasm,
showast,
showgrammar,
code_objects=code_objects)
except pysource.SourceWalkerError as e:
# deparsing failed
if real_out != out:
print(e, file=real_out)
raise
def comprehension_walk(self, node, iter_index):
p = self.prec
self.prec = 27
code = node[-5].attr
assert iscode(co)
code = Code(code, self.scanner, self.currentclass)
# assert isinstance(code, Code)
ast = self.build_ast(code._tokens, code._customize)
self.customize(code._customize)
ast = ast[0][0][0]
n = ast[iter_index]
assert n == 'comp_iter'
# find innermost node
while n == 'comp_iter':
n = n[0] # recurse one step
if n == 'comp_for': n = n[3]
elif n == 'comp_if': n = n[2]
elif n == 'comp_ifnot': n = n[2]
assert n == 'comp_body', ast
self.preorder(n[0])
self.write(' for ')
start = len(self.f.getvalue())
designator = ast[iter_index - 1]
self.preorder(designator)
self.set_pos_info(ast[iter_index - 1], start, len(self.f.getvalue()))
self.write(' in ')
start = len(self.f.getvalue())
node[-3].parent = node
self.preorder(node[-3])
self.set_pos_info(node[-3], start, len(self.f.getvalue()))
start = len(self.f.getvalue())
self.preorder(ast[iter_index])
self.set_pos_info(iter_index, start, len(self.f.getvalue()))
self.prec = p
def disassemble(self, co, classname=None, code_objects={}):
"""
Disassemble a code object, returning a list of 'Token'.
The main part of this procedure is modelled after
dis.disassemble().
"""
# import dis; dis.disassemble(co) # DEBUG
rv = []
customize = {}
Token = self.Token # shortcut
self.code = array('B', co.co_code)
for i in self.op_range(0, len(self.code)):
if self.code[i] in (RETURN_VALUE, END_FINALLY):
n = i + 1
self.code = array('B', co.co_code[:n])
self.prev = [0]
# mapping addresses of instruction & argument
for i in self.op_range(0, n):
op = self.code[i]
self.prev.append(i)
if op >= HAVE_ARGUMENT:
self.prev.append(i)
self.prev.append(i)
self.lines = []
linetuple = namedtuple('linetuple', ['l_no', 'next'])
j = 0
# linestarts is a tuple of (offset, line number).
# Turn that in a has that we can index
linestarts = list(dis.findlinestarts(co))
linestartoffsets = {}
for offset, lineno in linestarts:
linestartoffsets[offset] = lineno
(prev_start_byte, prev_line_no) = linestarts[0]
for (start_byte, line_no) in linestarts[1:]:
while j < start_byte:
self.lines.append(linetuple(prev_line_no, start_byte))
j += 1
prev_line_no = start_byte
while j < n:
self.lines.append(linetuple(prev_line_no, n))
j+=1
# self.lines contains (block,addrLastInstr)
if classname:
classname = '_' + classname.lstrip('_') + '__'
def unmangle(name):
if name.startswith(classname) and name[-2:] != '__':
return name[len(classname) - 2:]
return name
free = [ unmangle(name) for name in (co.co_cellvars + co.co_freevars) ]
names = [ unmangle(name) for name in co.co_names ]
varnames = [ unmangle(name) for name in co.co_varnames ]
else:
free = co.co_cellvars + co.co_freevars
names = co.co_names
varnames = co.co_varnames
self.load_asserts = set()
for i in self.op_range(0, n):
if self.code[i] == PJIT and self.code[i+3] == LOAD_GLOBAL:
if names[self.get_argument(i+3)] == 'AssertionError':
self.load_asserts.add(i+3)
cf = self.find_jump_targets(self.code)
# contains (code, [addrRefToCode])
last_stmt = self.next_stmt[0]
i = self.next_stmt[last_stmt]
replace = {}
while i < n-1:
if self.lines[last_stmt].next > i:
if self.code[last_stmt] == PRINT_ITEM:
if self.code[i] == PRINT_ITEM:
replace[i] = 'PRINT_ITEM_CONT'
elif self.code[i] == PRINT_NEWLINE:
replace[i] = 'PRINT_NEWLINE_CONT'
last_stmt = i
i = self.next_stmt[i]
imports = self.all_instr(0, n, (IMPORT_NAME, IMPORT_FROM, IMPORT_STAR))
if len(imports) > 1:
last_import = imports[0]
for i in imports[1:]:
if self.lines[last_import].next > i:
if self.code[last_import] == IMPORT_NAME == self.code[i]:
replace[i] = 'IMPORT_NAME_CONT'
last_import = i
extended_arg = 0
for offset in self.op_range(0, n):
if offset in cf:
k = 0
for j in cf[offset]:
rv.append(Token('COME_FROM', None, repr(j),
offset="%s_%d" % (offset, k)))
k += 1
op = self.code[offset]
op_name = opname[op]
oparg = None; pattr = None
if op >= HAVE_ARGUMENT:
oparg = self.get_argument(offset) + extended_arg
extended_arg = 0
if op == EXTENDED_ARG:
extended_arg = oparg * scan.L65536
continue
if op in hasconst:
const = co.co_consts[oparg]
if iscode(const):
oparg = const
if const.co_name == '<lambda>':
assert op_name == 'LOAD_CONST'
op_name = 'LOAD_LAMBDA'
elif const.co_name == '<genexpr>':
op_name = 'LOAD_GENEXPR'
elif const.co_name == '<dictcomp>':
op_name = 'LOAD_DICTCOMP'
elif const.co_name == '<setcomp>':
op_name = 'LOAD_SETCOMP'
# verify() uses 'pattr' for comparison, since 'attr'
# now holds Code(const) and thus can not be used
# for comparison (todo: think about changing this)
# pattr = 'code_object @ 0x%x %s->%s' %\
# (id(const), const.co_filename, const.co_name)
pattr = '<code_object ' + const.co_name + '>'
else:
pattr = const
elif op in hasname:
pattr = names[oparg]
elif op in hasjrel:
pattr = repr(offset + 3 + oparg)
elif op in hasjabs:
pattr = repr(oparg)
elif op in haslocal:
pattr = varnames[oparg]
elif op in hascompare:
pattr = cmp_op[oparg]
elif op in hasfree:
pattr = free[oparg]
if op in (BUILD_LIST, BUILD_TUPLE, BUILD_SET, BUILD_SLICE,
UNPACK_SEQUENCE,
MAKE_FUNCTION, CALL_FUNCTION, MAKE_CLOSURE,
CALL_FUNCTION_VAR, CALL_FUNCTION_KW,
CALL_FUNCTION_VAR_KW, DUP_TOPX, RAISE_VARARGS
):
# CE - Hack for >= 2.5
# Now all values loaded via LOAD_CLOSURE are packed into
# a tuple before calling MAKE_CLOSURE.
if op == BUILD_TUPLE and \
self.code[self.prev[offset]] == LOAD_CLOSURE:
continue
else:
op_name = '%s_%d' % (op_name, oparg)
if op != BUILD_SLICE:
customize[op_name] = oparg
elif op == JA:
target = self.get_target(offset)
if target < offset:
if offset in self.stmts and self.code[offset+3] not in (END_FINALLY, POP_BLOCK) \
and offset not in self.not_continue:
op_name = 'CONTINUE'
else:
op_name = 'JUMP_BACK'
elif op == LOAD_GLOBAL:
if offset in self.load_asserts:
op_name = 'LOAD_ASSERT'
elif op == RETURN_VALUE:
if offset in self.return_end_ifs:
op_name = 'RETURN_END_IF'
if offset in linestartoffsets:
linestart = linestartoffsets[offset]
else:
linestart = None
if offset not in replace:
rv.append(Token(op_name, oparg, pattr, offset, linestart))
else:
rv.append(Token(replace[offset], oparg, pattr, offset, linestart))
return rv, customize
def make_function(self, node, isLambda, nested=1, code_index=-2):
"""Dump function defintion, doc string, and function body."""
def build_param(ast, name, default):
"""build parameters:
- handle defaults
- handle format tuple parameters
"""
if self.version < 3.0:
# if formal parameter is a tuple, the paramater name
# starts with a dot (eg. '.1', '.2')
if name.startswith('.'):
# replace the name with the tuple-string
name = self.get_tuple_parameter(ast, name)
pass
pass
if default:
if self.showast:
print()
print('--', name)
print(default)
print('--')
pass
result = '%s=' % name
old_last_finish = self.last_finish
self.last_finish = len(result)
value = self.traverse(default, indent='')
self.last_finish = old_last_finish
result += value
if result[-2:] == '= ': # default was 'LOAD_CONST None'
result += 'None'
return result
else:
return name
# node[-1] == MAKE_FUNCTION_n
args_node = node[-1]
if isinstance(args_node.attr, tuple):
defparams = node[:args_node.attr[0]]
pos_args, kw_args, annotate_args = args_node.attr
else:
defparams = node[:args_node.attr]
kw_args, annotate_args = (0, 0)
pos_args = args_node.attr
pass
code = node[code_index].attr
assert iscode(code)
code = Code(code, self.scanner, self.currentclass)
# add defaults values to parameter names
argc = code.co_argcount
paramnames = list(code.co_varnames[:argc])
# defaults are for last n parameters, thus reverse
paramnames.reverse(); defparams.reverse()
try:
ast = self.build_ast(code._tokens,
code._customize,
isLambda = isLambda,
noneInNames = ('None' in code.co_names))
except ParserError as p:
self.write(str(p))
self.ERROR = p
return
# build parameters
params = [build_param(ast, name, default) for
name, default in zip_longest(paramnames, defparams, fillvalue=None)]
params.reverse() # back to correct order
if 4 & code.co_flags: # flag 2 -> variable number of args
params.append('*%s' % code.co_varnames[argc])
argc += 1
if 8 & code.co_flags: # flag 3 -> keyword args
params.append('**%s' % code.co_varnames[argc])
argc += 1
# dump parameter list (with default values)
indent = self.indent
if isLambda:
self.write("lambda ", ", ".join(params))
else:
self.write("(", ", ".join(params))
# self.println(indent, '#flags:\t', int(code.co_flags))
if kw_args > 0:
if argc > 0:
self.write(", *, ")
else:
self.write("*, ")
for n in node:
if n == 'pos_arg':
continue
self.preorder(n)
break
pass
if isLambda:
self.write(": ")
else:
self.println("):")
if len(code.co_consts)>0 and code.co_consts[0] is not None and not isLambda: # ugly
# docstring exists, dump it
self.print_docstring(indent, code.co_consts[0])
code._tokens = None # save memory
assert ast == 'stmts'
all_globals = find_all_globals(ast, set())
for g in ((all_globals & self.mod_globs) | find_globals(ast, set())):
self.println(self.indent, 'global ', g)
self.mod_globs -= all_globals
rn = ('None' in code.co_names) and not find_none(ast)
self.gen_source(ast, code.co_name, code._customize, isLambda=isLambda,
returnNone=rn)
code._tokens = None; code._customize = None # save memory
def disassemble_generic(self, co, classname=None, code_objects={}):
"""
Convert code object <co> into a sequence of tokens.
The below is based on (an older version?) of Python dis.disassemble_bytes().
"""
# Container for tokens
tokens = []
customize = {}
self.code = code = array('B', co.co_code)
codelen = len(code)
self.build_lines_data(co)
self.build_prev_op()
self.code_objects = code_objects
# self.lines contains (block,addrLastInstr)
if classname:
classname = '_' + classname.lstrip('_') + '__'
def unmangle(name):
if name.startswith(classname) and name[-2:] != '__':
return name[len(classname) - 2:]
return name
free = [ unmangle(name) for name in (co.co_cellvars + co.co_freevars) ]
names = [ unmangle(name) for name in co.co_names ]
varnames = [ unmangle(name) for name in co.co_varnames ]
else:
free = co.co_cellvars + co.co_freevars
names = co.co_names
varnames = co.co_varnames
pass
# Scan for assertions. Later we will
# turn 'LOAD_GLOBAL' to 'LOAD_ASSERT' for those
# assertions
self.load_asserts = set()
for i in self.op_range(0, codelen):
if self.code[i] == POP_JUMP_IF_TRUE and self.code[i+3] == LOAD_GLOBAL:
if names[self.get_argument(i+3)] == 'AssertionError':
self.load_asserts.add(i+3)
# Get jump targets
# Format: {target offset: [jump offsets]}
jump_targets = self.find_jump_targets()
# contains (code, [addrRefToCode])
last_stmt = self.next_stmt[0]
i = self.next_stmt[last_stmt]
replace = {}
imports = self.all_instr(0, codelen, (IMPORT_NAME, IMPORT_FROM, IMPORT_STAR))
if len(imports) > 1:
last_import = imports[0]
for i in imports[1:]:
if self.lines[last_import].next > i:
if self.code[last_import] == IMPORT_NAME == self.code[i]:
replace[i] = 'IMPORT_NAME_CONT'
last_import = i
# Initialize extended arg at 0. When extended arg op is encountered,
# variable preserved for next cycle and added as arg for next op
extended_arg = 0
for offset in self.op_range(0, codelen):
# Add jump target tokens
if offset in jump_targets:
jump_idx = 0
for jump_offset in jump_targets[offset]:
tokens.append(Token('COME_FROM', None, repr(jump_offset),
offset='%s_%s' % (offset, jump_idx)))
jump_idx += 1
pass
pass
op = code[offset]
op_name = op3.opname[op]
oparg = None; pattr = None
if op >= op3.HAVE_ARGUMENT:
oparg = self.get_argument(offset) + extended_arg
extended_arg = 0
if op == op3.EXTENDED_ARG:
extended_arg = oparg * scan.L65536
continue
if op in op3.hasconst:
const = co.co_consts[oparg]
if not PYTHON3 and isinstance(const, str):
if const in code_objects:
const = code_objects[const]
if iscode(const):
oparg = const
if const.co_name == '<lambda>':
assert op_name == 'LOAD_CONST'
op_name = 'LOAD_LAMBDA'
elif const.co_name == '<genexpr>':
op_name = 'LOAD_GENEXPR'
elif const.co_name == '<dictcomp>':
op_name = 'LOAD_DICTCOMP'
elif const.co_name == '<setcomp>':
op_name = 'LOAD_SETCOMP'
elif const.co_name == '<listcomp>':
op_name = 'LOAD_LISTCOMP'
# verify() uses 'pattr' for comparison, since 'attr'
# now holds Code(const) and thus can not be used
# for comparison (todo: think about changing this)
# pattr = 'code_object @ 0x%x %s->%s' %\
# (id(const), const.co_filename, const.co_name)
pattr = '<code_object ' + const.co_name + '>'
else:
pattr = const
elif op in op3.hasname:
pattr = names[oparg]
elif op in op3.hasjrel:
pattr = repr(offset + 3 + oparg)
elif op in op3.hasjabs:
pattr = repr(oparg)
elif op in op3.haslocal:
pattr = varnames[oparg]
elif op in op3.hascompare:
pattr = op3.cmp_op[oparg]
elif op in op3.hasfree:
pattr = free[oparg]
if op_name in ('BUILD_LIST', 'BUILD_TUPLE', 'BUILD_SET', 'BUILD_SLICE',
'UNPACK_SEQUENCE',
'MAKE_FUNCTION', 'CALL_FUNCTION', 'MAKE_CLOSURE',
'CALL_FUNCTION_VAR', 'CALL_FUNCTION_KW',
'CALL_FUNCTION_VAR_KW', 'RAISE_VARARGS'
):
# CALL_FUNCTION OP renaming is done as a custom rule in parse3
if op_name not in ('CALL_FUNCTION', 'CALL_FUNCTION_VAR',
'CALL_FUNCTION_VAR_KW', 'CALL_FUNCTION_KW',
):
op_name = '%s_%d' % (op_name, oparg)
if op_name != 'BUILD_SLICE':
customize[op_name] = oparg
elif op_name == 'JUMP_ABSOLUTE':
target = self.get_target(offset)
if target < offset:
if (offset in self.stmts
and self.code[offset+3] not in (END_FINALLY, POP_BLOCK)
and offset not in self.not_continue):
op_name = 'CONTINUE'
else:
op_name = 'JUMP_BACK'
elif op_name == 'LOAD_GLOBAL':
if offset in self.load_asserts:
op_name = 'LOAD_ASSERT'
elif op_name == 'RETURN_VALUE':
if offset in self.return_end_ifs:
op_name = 'RETURN_END_IF'
if offset in self.linestarts:
linestart = self.linestarts[offset]
else:
linestart = None
if offset not in replace:
tokens.append(Token(op_name, oparg, pattr, offset, linestart))
else:
tokens.append(Token(replace[offset], oparg, pattr, offset, linestart))
pass
return tokens, customize
def disassemble_built_in(self, co, classname=None, code_objects={}):
# Container for tokens
tokens = []
customize = {}
self.code = array('B', co.co_code)
self.build_lines_data(co)
self.build_prev_op()
# Get jump targets
# Format: {target offset: [jump offsets]}
jump_targets = self.find_jump_targets()
bytecode = dis.Bytecode(co)
# self.lines contains (block,addrLastInstr)
if classname:
classname = '_' + classname.lstrip('_') + '__'
def unmangle(name):
if name.startswith(classname) and name[-2:] != '__':
return name[len(classname) - 2:]
return name
# free = [ unmangle(name) for name in (co.co_cellvars + co.co_freevars) ]
# names = [ unmangle(name) for name in co.co_names ]
# varnames = [ unmangle(name) for name in co.co_varnames ]
else:
# free = co.co_cellvars + co.co_freevars
# names = co.co_names
# varnames = co.co_varnames
pass
# Scan for assertions. Later we will
# turn 'LOAD_GLOBAL' to 'LOAD_ASSERT' for those
# assertions
self.load_asserts = set()
bs = list(bytecode)
n = len(bs)
for i in range(n):
inst = bs[i]
if inst.opname == 'POP_JUMP_IF_TRUE' and i + 1 < n:
next_inst = bs[i + 1]
if (next_inst.opname == 'LOAD_GLOBAL'
and next_inst.argval == 'AssertionError'):
self.load_asserts.add(next_inst.offset)
for inst in bytecode:
if inst.offset in jump_targets:
jump_idx = 0
for jump_offset in jump_targets[inst.offset]:
tokens.append(
Token('COME_FROM',
None,
repr(jump_offset),
offset='%s_%s' % (inst.offset, jump_idx)))
jump_idx += 1
pass
pass
pattr = inst.argrepr
opname = inst.opname
if opname in ['LOAD_CONST']:
const = inst.argval
if iscode(const):
if const.co_name == '<lambda>':
opname = 'LOAD_LAMBDA'
elif const.co_name == '<genexpr>':
opname = 'LOAD_GENEXPR'
elif const.co_name == '<dictcomp>':
opname = 'LOAD_DICTCOMP'
elif const.co_name == '<setcomp>':
opname = 'LOAD_SETCOMP'
elif const.co_name == '<listcomp>':
opname = 'LOAD_LISTCOMP'
# verify() uses 'pattr' for comparison, since 'attr'
# now holds Code(const) and thus can not be used
# for comparison (todo: think about changing this)
# pattr = 'code_object @ 0x%x %s->%s' %\
# (id(const), const.co_filename, const.co_name)
pattr = '<code_object ' + const.co_name + '>'
else:
pattr = const
pass
elif opname in ('BUILD_LIST', 'BUILD_TUPLE', 'BUILD_SET',
'BUILD_SLICE', 'UNPACK_SEQUENCE', 'MAKE_FUNCTION',
'MAKE_CLOSURE', 'DUP_TOPX', 'RAISE_VARARGS'):
# if opname == 'BUILD_TUPLE' and \
# self.code[self.prev[offset]] == LOAD_CLOSURE:
# continue
# else:
# op_name = '%s_%d' % (op_name, oparg)
# if opname != BUILD_SLICE:
# customize[op_name] = oparg
opname = '%s_%d' % (opname, inst.argval)
if inst.opname != 'BUILD_SLICE':
customize[opname] = inst.argval
elif opname == 'JUMP_ABSOLUTE':
pattr = inst.argval
target = self.get_target(inst.offset)
if target < inst.offset:
if (inst.offset in self.stmts
and self.code[inst.offset + 3]
not in (END_FINALLY, POP_BLOCK)
and offset not in self.not_continue):
opname = 'CONTINUE'
else:
opname = 'JUMP_BACK'
elif inst.offset in self.load_asserts:
opname = 'LOAD_ASSERT'
tokens.append(
Token(
type_=opname,
attr=inst.argval,
pattr=pattr,
offset=inst.offset,
linestart=inst.starts_line,
))
pass
return tokens, {}
def disassemble(self, co, classname=None, code_objects={}):
"""
Disassemble a code object, returning a list of 'Token'.
The main part of this procedure is modelled after
dis.disassemble().
"""
rv = []
customize = {}
Token = self.Token # shortcut
self.code = array('B', co.co_code)
for i in self.op_range(0, len(self.code)):
if self.code[i] in (RETURN_VALUE, END_FINALLY):
n = i + 1
self.code = array('B', co.co_code[:n])
self.prev = [0]
# mapping addresses of instruction & argument
for i in self.op_range(0, n):
op = self.code[i]
self.prev.append(i)
if op >= HAVE_ARGUMENT:
self.prev.append(i)
self.prev.append(i)
self.lines = []
linetuple = namedtuple('linetuple', ['l_no', 'next'])
j = 0
# linestarts is a tuple of (offset, line number).
# Turn that in a has that we can index
linestarts = list(dis.findlinestarts(co))
linestartoffsets = {}
for offset, lineno in linestarts:
linestartoffsets[offset] = lineno
(prev_start_byte, prev_line_no) = linestarts[0]
for (start_byte, line_no) in linestarts[1:]:
while j < start_byte:
self.lines.append(linetuple(prev_line_no, start_byte))
j += 1
prev_line_no = start_byte
while j < n:
self.lines.append(linetuple(prev_line_no, n))
j += 1
# self.lines contains (block,addrLastInstr)
if classname:
classname = '_' + classname.lstrip('_') + '__'
def unmangle(name):
if name.startswith(classname) and name[-2:] != '__':
return name[len(classname) - 2:]
return name
free = [
unmangle(name) for name in (co.co_cellvars + co.co_freevars)
]
names = [unmangle(name) for name in co.co_names]
varnames = [unmangle(name) for name in co.co_varnames]
else:
free = co.co_cellvars + co.co_freevars
names = co.co_names
varnames = co.co_varnames
self.load_asserts = set()
for i in self.op_range(0, n):
if self.code[i] == PJIT and self.code[i + 3] == LOAD_GLOBAL:
if names[self.get_argument(i + 3)] == 'AssertionError':
self.load_asserts.add(i + 3)
cf = self.find_jump_targets(self.code)
# contains (code, [addrRefToCode])
last_stmt = self.next_stmt[0]
i = self.next_stmt[last_stmt]
replace = {}
while i < n - 1:
if self.lines[last_stmt].next > i:
if self.code[last_stmt] == PRINT_ITEM:
if self.code[i] == PRINT_ITEM:
replace[i] = 'PRINT_ITEM_CONT'
elif self.code[i] == PRINT_NEWLINE:
replace[i] = 'PRINT_NEWLINE_CONT'
last_stmt = i
i = self.next_stmt[i]
imports = self.all_instr(0, n, (IMPORT_NAME, IMPORT_FROM, IMPORT_STAR))
if len(imports) > 1:
last_import = imports[0]
for i in imports[1:]:
if self.lines[last_import].next > i:
if self.code[last_import] == IMPORT_NAME == self.code[i]:
replace[i] = 'IMPORT_NAME_CONT'
last_import = i
extended_arg = 0
for offset in self.op_range(0, n):
if offset in cf:
k = 0
for j in cf[offset]:
rv.append(
Token('COME_FROM',
None,
repr(j),
offset="%s_%d" % (offset, k)))
k += 1
op = self.code[offset]
op_name = opname[op]
oparg = None
pattr = None
if op >= HAVE_ARGUMENT:
oparg = self.get_argument(offset) + extended_arg
extended_arg = 0
if op == EXTENDED_ARG:
extended_arg = oparg * scan.L65536
continue
if op in hasconst:
const = co.co_consts[oparg]
if iscode(const):
oparg = const
if const.co_name == '<lambda>':
assert op_name == 'LOAD_CONST'
op_name = 'LOAD_LAMBDA'
elif const.co_name == '<genexpr>':
op_name = 'LOAD_GENEXPR'
elif const.co_name == '<dictcomp>':
op_name = 'LOAD_DICTCOMP'
elif const.co_name == '<setcomp>':
op_name = 'LOAD_SETCOMP'
# verify() uses 'pattr' for comparison, since 'attr'
# now holds Code(const) and thus can not be used
# for comparison (todo: think about changing this)
# pattr = 'code_object @ 0x%x %s->%s' %\
# (id(const), const.co_filename, const.co_name)
pattr = '<code_object ' + const.co_name + '>'
else:
pattr = const
elif op in hasname:
pattr = names[oparg]
elif op in hasjrel:
pattr = repr(offset + 3 + oparg)
elif op in hasjabs:
pattr = repr(oparg)
elif op in haslocal:
pattr = varnames[oparg]
elif op in hascompare:
pattr = cmp_op[oparg]
elif op in hasfree:
pattr = free[oparg]
if op in (BUILD_LIST, BUILD_TUPLE, BUILD_SET, BUILD_SLICE,
UNPACK_SEQUENCE, MAKE_FUNCTION, CALL_FUNCTION,
MAKE_CLOSURE, CALL_FUNCTION_VAR, CALL_FUNCTION_KW,
CALL_FUNCTION_VAR_KW, DUP_TOPX, RAISE_VARARGS):
# CE - Hack for >= 2.5
# Now all values loaded via LOAD_CLOSURE are packed into
# a tuple before calling MAKE_CLOSURE.
if op == BUILD_TUPLE and \
self.code[self.prev[offset]] == LOAD_CLOSURE:
continue
else:
op_name = '%s_%d' % (op_name, oparg)
if op != BUILD_SLICE:
customize[op_name] = oparg
elif op == JA:
target = self.get_target(offset)
if target < offset:
if offset in self.stmts and self.code[offset+3] not in (END_FINALLY, POP_BLOCK) \
and offset not in self.not_continue:
op_name = 'CONTINUE'
else:
op_name = 'JUMP_BACK'
elif op == LOAD_GLOBAL:
if offset in self.load_asserts:
op_name = 'LOAD_ASSERT'
elif op == RETURN_VALUE:
if offset in self.return_end_ifs:
op_name = 'RETURN_END_IF'
if offset in linestartoffsets:
linestart = linestartoffsets[offset]
else:
linestart = None
if offset not in replace:
rv.append(Token(op_name, oparg, pattr, offset, linestart))
else:
rv.append(
Token(replace[offset], oparg, pattr, offset, linestart))
return rv, customize
def disassemble_built_in(self, co, classname=None,
code_objects={}):
# Container for tokens
tokens = []
customize = {}
self.code = array('B', co.co_code)
self.build_lines_data(co)
self.build_prev_op()
# Get jump targets
# Format: {target offset: [jump offsets]}
jump_targets = self.find_jump_targets()
bytecode = dis.Bytecode(co)
# self.lines contains (block,addrLastInstr)
if classname:
classname = '_' + classname.lstrip('_') + '__'
def unmangle(name):
if name.startswith(classname) and name[-2:] != '__':
return name[len(classname) - 2:]
return name
# free = [ unmangle(name) for name in (co.co_cellvars + co.co_freevars) ]
# names = [ unmangle(name) for name in co.co_names ]
# varnames = [ unmangle(name) for name in co.co_varnames ]
else:
# free = co.co_cellvars + co.co_freevars
# names = co.co_names
# varnames = co.co_varnames
pass
# Scan for assertions. Later we will
# turn 'LOAD_GLOBAL' to 'LOAD_ASSERT' for those
# assertions
self.load_asserts = set()
bs = list(bytecode)
n = len(bs)
for i in range(n):
inst = bs[i]
if inst.opname == 'POP_JUMP_IF_TRUE' and i+1 < n:
next_inst = bs[i+1]
if (next_inst.opname == 'LOAD_GLOBAL' and
next_inst.argval == 'AssertionError'):
self.load_asserts.add(next_inst.offset)
for inst in bytecode:
if inst.offset in jump_targets:
jump_idx = 0
for jump_offset in jump_targets[inst.offset]:
tokens.append(Token('COME_FROM', None, repr(jump_offset),
offset='%s_%s' % (inst.offset, jump_idx)))
jump_idx += 1
pass
pass
pattr = inst.argrepr
opname = inst.opname
if opname in ['LOAD_CONST']:
const = inst.argval
if iscode(const):
if const.co_name == '<lambda>':
opname = 'LOAD_LAMBDA'
elif const.co_name == '<genexpr>':
opname = 'LOAD_GENEXPR'
elif const.co_name == '<dictcomp>':
opname = 'LOAD_DICTCOMP'
elif const.co_name == '<setcomp>':
opname = 'LOAD_SETCOMP'
elif const.co_name == '<listcomp>':
opname = 'LOAD_LISTCOMP'
# verify() uses 'pattr' for comparison, since 'attr'
# now holds Code(const) and thus can not be used
# for comparison (todo: think about changing this)
# pattr = 'code_object @ 0x%x %s->%s' %\
# (id(const), const.co_filename, const.co_name)
pattr = '<code_object ' + const.co_name + '>'
else:
pattr = const
pass
elif opname in ('BUILD_LIST', 'BUILD_TUPLE', 'BUILD_SET', 'BUILD_SLICE',
'UNPACK_SEQUENCE',
'MAKE_FUNCTION', 'MAKE_CLOSURE',
'DUP_TOPX', 'RAISE_VARARGS'
):
# if opname == 'BUILD_TUPLE' and \
# self.code[self.prev[offset]] == LOAD_CLOSURE:
# continue
# else:
# op_name = '%s_%d' % (op_name, oparg)
# if opname != BUILD_SLICE:
# customize[op_name] = oparg
opname = '%s_%d' % (opname, inst.argval)
if inst.opname != 'BUILD_SLICE':
customize[opname] = inst.argval
elif opname == 'JUMP_ABSOLUTE':
pattr = inst.argval
target = self.get_target(inst.offset)
if target < inst.offset:
if (inst.offset in self.stmts and
self.code[inst.offset+3] not in (END_FINALLY, POP_BLOCK)
and offset not in self.not_continue):
opname = 'CONTINUE'
else:
opname = 'JUMP_BACK'
elif inst.offset in self.load_asserts:
opname = 'LOAD_ASSERT'
tokens.append(
Token(
type_ = opname,
attr = inst.argval,
pattr = pattr,
offset = inst.offset,
linestart = inst.starts_line,
)
)
pass
return tokens, {}
def disassemble(self, co, classname=None,
code_objects={}):
# imoprt dis; dis.disassemble(co) # DEBUG
# Container for tokens
tokens = []
customize = {}
self.code = array('B', co.co_code)
self.build_lines_data(co)
self.build_prev_op()
bytecode = dis35.Bytecode(co)
# self.lines contains (block,addrLastInstr)
if classname:
classname = '_' + classname.lstrip('_') + '__'
def unmangle(name):
if name.startswith(classname) and name[-2:] != '__':
return name[len(classname) - 2:]
return name
else:
pass
# Scan for assertions. Later we will
# turn 'LOAD_GLOBAL' to 'LOAD_ASSERT' for those
# assertions
self.load_asserts = set()
bs = list(bytecode)
n = len(bs)
for i in range(n):
inst = bs[i]
if inst.opname == 'POP_JUMP_IF_TRUE' and i+1 < n:
next_inst = bs[i+1]
if (next_inst.opname == 'LOAD_GLOBAL' and
next_inst.argval == 'AssertionError'):
self.load_asserts.add(next_inst.offset)
# Get jump targets
# Format: {target offset: [jump offsets]}
jump_targets = self.find_jump_targets()
for inst in bytecode:
if inst.offset in jump_targets:
jump_idx = 0
for jump_offset in jump_targets[inst.offset]:
tokens.append(Token('COME_FROM', None, repr(jump_offset),
offset='%s_%s' % (inst.offset, jump_idx)))
jump_idx += 1
pass
pass
pattr = inst.argrepr
opname = inst.opname
if opname in ['LOAD_CONST']:
const = inst.argval
if iscode(const):
if const.co_name == '<lambda>':
opname = 'LOAD_LAMBDA'
elif const.co_name == '<genexpr>':
opname = 'LOAD_GENEXPR'
elif const.co_name == '<dictcomp>':
opname = 'LOAD_DICTCOMP'
elif const.co_name == '<setcomp>':
opname = 'LOAD_SETCOMP'
elif const.co_name == '<listcomp>':
opname = 'LOAD_LISTCOMP'
# verify() uses 'pattr' for comparison, since 'attr'
# now holds Code(const) and thus can not be used
# for comparison (todo: think about changing this)
# pattr = 'code_object @ 0x%x %s->%s' %\
# (id(const), const.co_filename, const.co_name)
pattr = '<code_object ' + const.co_name + '>'
else:
pattr = const
pass
elif opname == 'MAKE_FUNCTION':
argc = inst.argval
attr = ((argc & 0xFF), (argc >> 8) & 0xFF, (argc >> 16) & 0x7FFF)
pos_args, name_pair_args, annotate_args = attr
if name_pair_args > 0:
opname = 'MAKE_FUNCTION_N%d' % name_pair_args
pass
if annotate_args > 0:
opname = '%s_A_%d' % [op_name, annotate_args]
pass
opname = '%s_%d' % (opname, pos_args)
pattr = ("%d positional, %d keyword pair, %d annotated" %
(pos_args, name_pair_args, annotate_args))
tokens.append(
Token(
type_ = opname,
attr = (pos_args, name_pair_args, annotate_args),
pattr = pattr,
offset = inst.offset,
linestart = inst.starts_line)
)
continue
elif opname in ('BUILD_LIST', 'BUILD_TUPLE', 'BUILD_SET', 'BUILD_SLICE',
'BUILD_MAP', 'UNPACK_SEQUENCE', 'MAKE_CLOSURE',
'RAISE_VARARGS'
):
pos_args = inst.argval
if inst.opname != 'BUILD_SLICE':
customize[opname] = pos_args
pass
opname = '%s_%d' % (opname, pos_args)
elif opname == 'JUMP_ABSOLUTE':
pattr = inst.argval
target = self.get_target(inst.offset)
if target < inst.offset:
if (inst.offset in self.stmts and
self.code[inst.offset+3] not in (END_FINALLY, POP_BLOCK)
and inst.offset not in self.not_continue):
opname = 'CONTINUE'
else:
opname = 'JUMP_BACK'
elif inst.offset in self.load_asserts:
opname = 'LOAD_ASSERT'
tokens.append(
Token(
type_ = opname,
attr = inst.argval,
pattr = pattr,
offset = inst.offset,
linestart = inst.starts_line,
)
)
pass
return tokens, {}
def cmp_code_objects(version, code_obj1, code_obj2, name=''):
"""
Compare two code-objects.
This is the main part of this module.
"""
# print code_obj1, type(code_obj2)
assert iscode(code_obj1), \
"cmp_code_object first object type is %s, not code" % type(code_obj1)
assert iscode(code_obj2), \
"cmp_code_object second object type is %s, not code" % type(code_obj2)
# print dir(code_obj1)
if isinstance(code_obj1, object):
# new style classes (Python 2.2)
# assume _both_ code objects to be new stle classes
assert dir(code_obj1) == dir(code_obj2)
else:
# old style classes
assert dir(code_obj1) == code_obj1.__members__
assert dir(code_obj2) == code_obj2.__members__
assert code_obj1.__members__ == code_obj2.__members__
if name == '__main__':
name = code_obj1.co_name
else:
name = '%s.%s' % (name, code_obj1.co_name)
if name == '.?': name = '__main__'
if isinstance(code_obj1, object) and code_equal(code_obj1, code_obj2):
# use the new style code-classes' __cmp__ method, which
# should be faster and more sophisticated
# if this compare fails, we use the old routine to
# find out, what exactly is nor equal
# if this compare succeds, simply return
# return
pass
if isinstance(code_obj1, object):
members = [x for x in dir(code_obj1) if x.startswith('co_')]
else:
members = dir(code_obj1)
members.sort() # ; members.reverse()
tokens1 = None
for member in members:
if member in __IGNORE_CODE_MEMBERS__:
pass
elif member == 'co_code':
if version == 2.5:
import uncompyle6.scanners.scanner25 as scan
scanner = scan.Scanner25(version)
elif version == 2.6:
import uncompyle6.scanners.scanner26 as scan
scanner = scan.Scanner26(version)
elif version == 2.7:
import uncompyle6.scanners.scanner27 as scan
scanner = scan.Scanner27(version)
elif version == 3.2:
import uncompyle6.scanners.scanner32 as scan
scanner = scan.Scanner32(version)
elif version == 3.3:
import uncompyle6.scanners.scanner33 as scan
scanner = scan.Scanner33(version)
elif version == 3.4:
import uncompyle6.scanners.scanner34 as scan
scanner = scan.Scanner34(version)
global JUMP_OPs
JUMP_OPs = list(scan.JUMP_OPs) + ['JUMP_BACK']
# use changed Token class
# we (re)set this here to save exception handling,
# which would get 'unubersichtlich'
scanner.setTokenClass(Token)
try:
# disassemble both code-objects
tokens1, customize = scanner.disassemble(code_obj1)
del customize # save memory
tokens2, customize = scanner.disassemble(code_obj2)
del customize # save memory
finally:
scanner.resetTokenClass() # restore Token class
targets1 = dis.findlabels(code_obj1.co_code)
tokens1 = [t for t in tokens1 if t.type != 'COME_FROM']
tokens2 = [t for t in tokens2 if t.type != 'COME_FROM']
i1 = 0; i2 = 0
offset_map = {}; check_jumps = {}
while i1 < len(tokens1):
if i2 >= len(tokens2):
if len(tokens1) == len(tokens2) + 2 \
and tokens1[-1].type == 'RETURN_VALUE' \
and tokens1[-2].type == 'LOAD_CONST' \
and tokens1[-2].pattr is None \
and tokens1[-3].type == 'RETURN_VALUE':
break
else:
raise CmpErrorCodeLen(name, tokens1, tokens2)
offset_map[tokens1[i1].offset] = tokens2[i2].offset
for idx1, idx2, offset2 in check_jumps.get(tokens1[i1].offset, []):
if offset2 != tokens2[i2].offset:
raise CmpErrorCode(name, tokens1[idx1].offset, tokens1[idx1],
tokens2[idx2], tokens1, tokens2)
if tokens1[i1].type != tokens2[i2].type:
if tokens1[i1].type == 'LOAD_CONST' == tokens2[i2].type:
i = 1
while tokens1[i1+i].type == 'LOAD_CONST':
i += 1
if tokens1[i1+i].type.startswith(('BUILD_TUPLE', 'BUILD_LIST')) \
and i == int(tokens1[i1+i].type.split('_')[-1]):
t = tuple([ elem.pattr for elem in tokens1[i1:i1+i] ])
if t != tokens2[i2].pattr:
raise CmpErrorCode(name, tokens1[i1].offset, tokens1[i1],
tokens2[i2], tokens1, tokens2)
i1 += i + 1
i2 += 1
continue
elif i == 2 and tokens1[i1+i].type == 'ROT_TWO' and tokens2[i2+1].type == 'UNPACK_SEQUENCE_2':
i1 += 3
i2 += 2
continue
elif i == 2 and tokens1[i1+i].type in BIN_OP_FUNCS:
f = BIN_OP_FUNCS[tokens1[i1+i].type]
if f(tokens1[i1].pattr, tokens1[i1+1].pattr) == tokens2[i2].pattr:
i1 += 3
i2 += 1
continue
elif tokens1[i1].type == 'UNARY_NOT':
if tokens2[i2].type == 'POP_JUMP_IF_TRUE':
if tokens1[i1+1].type == 'POP_JUMP_IF_FALSE':
i1 += 2
i2 += 1
continue
elif tokens2[i2].type == 'POP_JUMP_IF_FALSE':
if tokens1[i1+1].type == 'POP_JUMP_IF_TRUE':
i1 += 2
i2 += 1
continue
elif tokens1[i1].type in ('JUMP_FORWARD', 'JUMP_BACK') \
and tokens1[i1-1].type == 'RETURN_VALUE' \
and tokens2[i2-1].type in ('RETURN_VALUE', 'RETURN_END_IF') \
and int(tokens1[i1].offset) not in targets1:
i1 += 1
continue
elif tokens1[i1].type == 'JUMP_FORWARD' and tokens2[i2].type == 'JUMP_BACK' \
and tokens1[i1+1].type == 'JUMP_BACK' and tokens2[i2+1].type == 'JUMP_BACK' \
and int(tokens1[i1].pattr) == int(tokens1[i1].offset) + 3:
if int(tokens1[i1].pattr) == int(tokens1[i1+1].offset):
i1 += 2
i2 += 2
continue
raise CmpErrorCode(name, tokens1[i1].offset, tokens1[i1],
tokens2[i2], tokens1, tokens2)
elif tokens1[i1].type in JUMP_OPs and tokens1[i1].pattr != tokens2[i2].pattr:
dest1 = int(tokens1[i1].pattr)
dest2 = int(tokens2[i2].pattr)
if tokens1[i1].type == 'JUMP_BACK':
if offset_map[dest1] != dest2:
raise CmpErrorCode(name, tokens1[i1].offset, tokens1[i1],
tokens2[i2], tokens1, tokens2)
else:
# import pdb; pdb.set_trace()
if dest1 in check_jumps:
check_jumps[dest1].append((i1, i2, dest2))
else:
check_jumps[dest1] = [(i1, i2, dest2)]
i1 += 1
i2 += 1
del tokens1, tokens2 # save memory
elif member == 'co_consts':
# partial optimization can make the co_consts look different,
# so we'll just compare the code consts
codes1 = ( c for c in code_obj1.co_consts if hasattr(c, 'co_consts') )
codes2 = ( c for c in code_obj2.co_consts if hasattr(c, 'co_consts') )
for c1, c2 in zip(codes1, codes2):
cmp_code_objects(version, c1, c2, name=name)
else:
# all other members must be equal
if getattr(code_obj1, member) != getattr(code_obj2, member):
raise CmpErrorMember(name, member,
getattr(code_obj1, member),
getattr(code_obj2, member))
