You've already forked python-uncompyle6
mirror of
https://github.com/rocky/python-uncompyle6.git
synced 2025-08-04 09:22:40 +08:00
Start off with Anton Vorobyov's (DarkFenX) Python 3 scanner.
This commit is contained in:
rocky
@@ -118,7 +118,7 @@ class Scanner(object):
|
||||
print('%i\t%s\t' % (i, self.opc.opname[op]))
|
||||
|
||||
def first_instr(self, start, end, instr, target=None, exact=True):
|
||||
'''
|
||||
"""
|
||||
Find the first <instr> in the block from start to end.
|
||||
<instr> is any python bytecode instruction or a list of opcodes
|
||||
If <instr> is an opcode with a target (like a jump), a target
|
||||
@@ -127,32 +127,34 @@ class Scanner(object):
|
||||
closest to <target> will be returned.
|
||||
|
||||
Return index to it or None if not found.
|
||||
'''
|
||||
"""
|
||||
code = self.code
|
||||
assert(start>=0 and end<=len(code))
|
||||
assert(start >= 0 and end <= len(code))
|
||||
|
||||
try: None in instr
|
||||
except: instr = [instr]
|
||||
try:
|
||||
None in instr
|
||||
except:
|
||||
instr = [instr]
|
||||
|
||||
pos = None
|
||||
distance = len(code)
|
||||
for i in self.op_range(start, end):
|
||||
op = code[i]
|
||||
result_offset = None
|
||||
current_distance = len(code)
|
||||
for offset in self.op_range(start, end):
|
||||
op = code[offset]
|
||||
if op in instr:
|
||||
if target is None:
|
||||
return i
|
||||
dest = self.get_target(i, op)
|
||||
return offset
|
||||
dest = self.get_target(offset)
|
||||
if dest == target:
|
||||
return i
|
||||
return offset
|
||||
elif not exact:
|
||||
_distance = abs(target - dest)
|
||||
if _distance < distance:
|
||||
distance = _distance
|
||||
pos = i
|
||||
return pos
|
||||
new_distance = abs(target - dest)
|
||||
if new_distance < current_distance:
|
||||
current_distance = new_distance
|
||||
result_offset = offset
|
||||
return result_offset
|
||||
|
||||
def last_instr(self, start, end, instr, target=None, exact=True):
|
||||
'''
|
||||
"""
|
||||
Find the last <instr> in the block from start to end.
|
||||
<instr> is any python bytecode instruction or a list of opcodes
|
||||
If <instr> is an opcode with a target (like a jump), a target
|
||||
@@ -161,65 +163,74 @@ class Scanner(object):
|
||||
closest to <target> will be returned.
|
||||
|
||||
Return index to it or None if not found.
|
||||
'''
|
||||
"""
|
||||
|
||||
code = self.code
|
||||
# Make sure requested positions do not go out of
|
||||
# code bounds
|
||||
if not (start>=0 and end<=len(code)):
|
||||
return None
|
||||
|
||||
try: None in instr
|
||||
except: instr = [instr]
|
||||
try:
|
||||
None in instr
|
||||
except:
|
||||
instr = [instr]
|
||||
|
||||
pos = None
|
||||
distance = len(code)
|
||||
for i in self.op_range(start, end):
|
||||
op = code[i]
|
||||
result_offset = None
|
||||
current_distance = len(code)
|
||||
for offset in self.op_range(start, end):
|
||||
op = code[offset]
|
||||
if op in instr:
|
||||
if target is None:
|
||||
pos = i
|
||||
result_offset = offset
|
||||
else:
|
||||
dest = self.get_target(i, op)
|
||||
dest = self.get_target(offset)
|
||||
if dest == target:
|
||||
distance = 0
|
||||
pos = i
|
||||
current_distance = 0
|
||||
result_offset = offset
|
||||
elif not exact:
|
||||
_distance = abs(target - dest)
|
||||
if _distance <= distance:
|
||||
distance = _distance
|
||||
pos = i
|
||||
return pos
|
||||
new_distance = abs(target - dest)
|
||||
if new_distance <= current_distance:
|
||||
current_distance = new_distance
|
||||
result_offset = offset
|
||||
return result_offset
|
||||
|
||||
def all_instr(self, start, end, instr, target=None, include_beyond_target=False):
|
||||
'''
|
||||
"""
|
||||
Find all <instr> in the block from start to end.
|
||||
<instr> is any python bytecode instruction or a list of opcodes
|
||||
If <instr> is an opcode with a target (like a jump), a target
|
||||
destination can be specified which must match precisely.
|
||||
|
||||
Return a list with indexes to them or [] if none found.
|
||||
'''
|
||||
|
||||
"""
|
||||
code = self.code
|
||||
assert(start>=0 and end<=len(code))
|
||||
assert(start >= 0 and end <= len(code))
|
||||
|
||||
try: None in instr
|
||||
except: instr = [instr]
|
||||
try:
|
||||
None in instr
|
||||
except:
|
||||
instr = [instr]
|
||||
|
||||
result = []
|
||||
for i in self.op_range(start, end):
|
||||
op = code[i]
|
||||
for offset in self.op_range(start, end):
|
||||
op = code[offset]
|
||||
if op in instr:
|
||||
if target is None:
|
||||
result.append(i)
|
||||
result.append(offset)
|
||||
else:
|
||||
t = self.get_target(i, op)
|
||||
t = self.get_target(offset)
|
||||
if include_beyond_target and t >= target:
|
||||
result.append(i)
|
||||
result.append(offset)
|
||||
elif t == target:
|
||||
result.append(i)
|
||||
result.append(offset)
|
||||
return result
|
||||
|
||||
def op_size(self, op):
|
||||
"""
|
||||
Return size of operator with its arguments
|
||||
for given opcode <op>.
|
||||
"""
|
||||
if op < self.opc.HAVE_ARGUMENT and op not in self.opc.hasArgumentExtended:
|
||||
return 1
|
||||
else:
|
||||
@@ -229,6 +240,10 @@ class Scanner(object):
|
||||
return self.op_size(op) > 1
|
||||
|
||||
def op_range(self, start, end):
|
||||
"""
|
||||
Iterate through positions of opcodes, skipping
|
||||
arguments.
|
||||
"""
|
||||
while start < end:
|
||||
yield start
|
||||
start += self.op_size(self.code[start])
|
||||
@@ -243,14 +258,14 @@ class Scanner(object):
|
||||
return filtered
|
||||
|
||||
def rem_or(self, start, end, instr, target=None, include_beyond_target=False):
|
||||
'''
|
||||
"""
|
||||
Find all <instr> in the block from start to end.
|
||||
<instr> is any python bytecode instruction or a list of opcodes
|
||||
If <instr> is an opcode with a target (like a jump), a target
|
||||
destination can be specified which must match precisely.
|
||||
|
||||
Return a list with indexes to them or [] if none found.
|
||||
'''
|
||||
"""
|
||||
|
||||
code = self.code
|
||||
assert(start>=0 and end<=len(code))
|
||||
|
||||
@@ -8,552 +8,403 @@ from __future__ import print_function
|
||||
See main module for license.
|
||||
'''
|
||||
|
||||
import dis, types
|
||||
import dis, marshal
|
||||
from collections import namedtuple
|
||||
from array import array
|
||||
from operator import itemgetter
|
||||
|
||||
from uncompyle6.scanner import Token
|
||||
|
||||
|
||||
# Get all the opcodes into globals
|
||||
globals().update(dis.opmap)
|
||||
from uncompyle6.opcodes.opcode_27 import *
|
||||
import uncompyle6.scanner as scan
|
||||
import scanner as scan
|
||||
|
||||
class Scanner27(scan.Scanner):
|
||||
|
||||
class Scanner34(scan.Scanner):
|
||||
def __init__(self):
|
||||
self.Token = scan.Scanner.__init__(self, 2.7) # check
|
||||
|
||||
def disassemble(self, co, classname=None):
|
||||
'''
|
||||
Disassemble a code object, returning a list of 'Token'.
|
||||
The main part of this procedure is modelled after
|
||||
dis.disassemble().
|
||||
'''
|
||||
rv = []
|
||||
def run(self, bytecode):
|
||||
code_object = marshal.loads(bytecode)
|
||||
tokens = self.tokenize(code_object)
|
||||
return tokens
|
||||
|
||||
def disassemble(self, co):
|
||||
"""
|
||||
Convert code object <co> into a sequence of tokens.
|
||||
|
||||
Based on dis.disassemble() function.
|
||||
"""
|
||||
# Container for tokens
|
||||
tokens = []
|
||||
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.code = code = co.co_code
|
||||
codelen = len(code)
|
||||
self.build_lines_data(co)
|
||||
self.build_prev_op()
|
||||
# Get jump targets
|
||||
# Format: {target offset: [jump offsets]}
|
||||
jump_targets = self.find_jump_targets()
|
||||
# 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
|
||||
free = None
|
||||
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='{}_{}'.format(offset, jump_idx)))
|
||||
jump_idx += 1
|
||||
op = code[offset]
|
||||
# Create token and fill all the fields we can
|
||||
# w/o touching arguments
|
||||
current_token = Token()
|
||||
current_token.type = dis.opname[op]
|
||||
current_token.offset = offset
|
||||
current_token.linestart = True if offset in self.linestarts else False
|
||||
if op >= dis.HAVE_ARGUMENT:
|
||||
# Calculate op's argument value based on its argument and
|
||||
# preceding extended argument, if any
|
||||
oparg = code[offset+1] + code[offset+2]*256 + extended_arg
|
||||
extended_arg = 0
|
||||
if op == dis.EXTENDED_ARG:
|
||||
extended_arg = oparg*65536
|
||||
|
||||
self.prev = [0]
|
||||
# mapping adresses of instru & arg
|
||||
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 contains bloc code adresse (addr,block)
|
||||
linestarts = list(dis.findlinestarts(co))
|
||||
linestartoffsets = {a for (a, _) in linestarts}
|
||||
(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_start_byte, prev_line_no) = (start_byte, line_no)
|
||||
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:
|
||||
# Fill token's attr/pattr fields
|
||||
current_token.attr = oparg
|
||||
if op in dis.hasconst:
|
||||
current_token.pattr = repr(co.co_consts[oparg])
|
||||
elif op in dis.hasname:
|
||||
current_token.pattr = co.co_names[oparg]
|
||||
elif op in dis.hasjrel:
|
||||
current_token.pattr = repr(offset + 3 + oparg)
|
||||
elif op in dis.haslocal:
|
||||
current_token.pattr = co.co_varnames[oparg]
|
||||
elif op in dis.hascompare:
|
||||
current_token.pattr = dis.cmp_op[oparg]
|
||||
elif op in dis.hasfree:
|
||||
if free is None:
|
||||
free = co.co_cellvars + co.co_freevars
|
||||
names = co.co_names
|
||||
varnames = co.co_varnames
|
||||
current_token.pattr = free[oparg]
|
||||
tokens.append(current_token)
|
||||
return tokens, customize
|
||||
|
||||
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)
|
||||
def build_lines_data(self, code_obj):
|
||||
"""
|
||||
Generate various line-related helper data.
|
||||
"""
|
||||
# Offset: lineno pairs, only for offsets which start line.
|
||||
# Locally we use list for more convenient iteration using indices
|
||||
linestarts = list(dis.findlinestarts(code_obj))
|
||||
self.linestarts = dict(linestarts)
|
||||
# Plain set with offsets of first ops on line
|
||||
self.linestart_offsets = {a for (a, _) in linestarts}
|
||||
# 'List-map' which shows line number of current op and offset of
|
||||
# first op on following line, given offset of op as index
|
||||
self.lines = lines = []
|
||||
LineTuple = namedtuple('LineTuple', ['l_no', 'next'])
|
||||
# Iterate through available linestarts, and fill
|
||||
# the data for all code offsets encountered until
|
||||
# last linestart offset
|
||||
_, prev_line_no = linestarts[0]
|
||||
offset = 0
|
||||
for start_offset, line_no in linestarts[1:]:
|
||||
while offset < start_offset:
|
||||
lines.append(LineTuple(prev_line_no, start_offset))
|
||||
offset += 1
|
||||
prev_line_no = line_no
|
||||
# Fill remaining offsets with reference to last line number
|
||||
# and code length as start offset of following non-existing line
|
||||
codelen = len(self.code)
|
||||
while offset < codelen:
|
||||
lines.append(LineTuple(prev_line_no, codelen))
|
||||
offset += 1
|
||||
|
||||
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 * 65536
|
||||
continue
|
||||
if op in hasconst:
|
||||
const = co.co_consts[oparg]
|
||||
if isinstance(const, types.CodeType):
|
||||
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 comparism, since 'attr'
|
||||
# now holds Code(const) and thus can not be used
|
||||
# for comparism (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 not in replace:
|
||||
rv.append(Token(op_name, oparg, pattr, offset, linestart = offset in linestartoffsets))
|
||||
else:
|
||||
rv.append(Token(replace[offset], oparg, pattr, offset, linestart = offset in linestartoffsets))
|
||||
|
||||
if self.showasm:
|
||||
out = self.out # shortcut
|
||||
for t in rv:
|
||||
print(t, file=out)
|
||||
print(file=out)
|
||||
return rv, customize
|
||||
|
||||
def op_size(self, op):
|
||||
if op < self.opc.HAVE_ARGUMENT:
|
||||
return 1
|
||||
else:
|
||||
return 3
|
||||
|
||||
def build_stmt_indices(self):
|
||||
def build_prev_op(self):
|
||||
"""
|
||||
Compose 'list-map' which allows to jump to previous
|
||||
op, given offset of current op as index.
|
||||
"""
|
||||
code = self.code
|
||||
start = 0
|
||||
end = len(code)
|
||||
codelen = len(code)
|
||||
self.prev_op = [0]
|
||||
for offset in self.op_range(0, codelen):
|
||||
op = code[offset]
|
||||
for _ in range(self.op_size(op)):
|
||||
self.prev_op.append(offset)
|
||||
|
||||
stmt_opcodes = {
|
||||
def find_jump_targets(self):
|
||||
"""
|
||||
Detect all offsets in a byte code which are jump targets.
|
||||
|
||||
Return the list of offsets.
|
||||
|
||||
This procedure is modelled after dis.findlables(), but here
|
||||
for each target the number of jumps is counted.
|
||||
"""
|
||||
code = self.code
|
||||
codelen = len(code)
|
||||
self.structs = [{'type': 'root',
|
||||
'start': 0,
|
||||
'end': codelen-1}]
|
||||
|
||||
# All loop entry points
|
||||
#self.loops = []
|
||||
# Map fixed jumps to their real destination
|
||||
self.fixed_jumps = {}
|
||||
self.ignore_if = set()
|
||||
self.build_statement_indices()
|
||||
# Containers filled by detect_structure()
|
||||
self.not_continue = set()
|
||||
self.return_end_ifs = set()
|
||||
|
||||
targets = {}
|
||||
for offset in self.op_range(0, codelen):
|
||||
op = code[offset]
|
||||
|
||||
# Determine structures and fix jumps for 2.3+
|
||||
self.detect_structure(offset)
|
||||
|
||||
if op >= dis.HAVE_ARGUMENT:
|
||||
label = self.fixed_jumps.get(offset)
|
||||
oparg = code[offset+1] + code[offset+2] * 256
|
||||
|
||||
if label is None:
|
||||
if op in dis.hasjrel and op != FOR_ITER:
|
||||
label = offset + 3 + oparg
|
||||
elif op in dis.hasjabs:
|
||||
if op in (JUMP_IF_FALSE_OR_POP, JUMP_IF_TRUE_OR_POP):
|
||||
if oparg > offset:
|
||||
label = oparg
|
||||
|
||||
if label is not None and label != -1:
|
||||
targets[label] = targets.get(label, []) + [offset]
|
||||
elif op == END_FINALLY and offset in self.fixed_jumps:
|
||||
label = self.fixed_jumps[offset]
|
||||
targets[label] = targets.get(label, []) + [offset]
|
||||
return targets
|
||||
|
||||
|
||||
def build_statement_indices(self):
|
||||
code = self.code
|
||||
start = 0;
|
||||
end = codelen = len(code)
|
||||
|
||||
statement_opcodes = {
|
||||
SETUP_LOOP, BREAK_LOOP, CONTINUE_LOOP,
|
||||
SETUP_FINALLY, END_FINALLY, SETUP_EXCEPT, SETUP_WITH,
|
||||
POP_BLOCK, STORE_FAST, DELETE_FAST, STORE_DEREF,
|
||||
STORE_GLOBAL, DELETE_GLOBAL, STORE_NAME, DELETE_NAME,
|
||||
STORE_ATTR, DELETE_ATTR, STORE_SUBSCR, DELETE_SUBSCR,
|
||||
RETURN_VALUE, RAISE_VARARGS, POP_TOP,
|
||||
PRINT_EXPR, PRINT_ITEM, PRINT_NEWLINE, PRINT_ITEM_TO, PRINT_NEWLINE_TO,
|
||||
STORE_SLICE_0, STORE_SLICE_1, STORE_SLICE_2, STORE_SLICE_3,
|
||||
DELETE_SLICE_0, DELETE_SLICE_1, DELETE_SLICE_2, DELETE_SLICE_3,
|
||||
JUMP_ABSOLUTE, EXEC_STMT,
|
||||
RETURN_VALUE, RAISE_VARARGS, POP_TOP, PRINT_EXPR,
|
||||
JUMP_ABSOLUTE
|
||||
}
|
||||
|
||||
stmt_opcode_seqs = [(PJIF, JF), (PJIF, JA), (PJIT, JF), (PJIT, JA)]
|
||||
statement_opcode_sequences = [(POP_JUMP_IF_FALSE, JUMP_FORWARD), (POP_JUMP_IF_FALSE, JUMP_ABSOLUTE),
|
||||
(POP_JUMP_IF_TRUE, JUMP_FORWARD), (POP_JUMP_IF_TRUE, JUMP_ABSOLUTE)]
|
||||
|
||||
designator_ops = {
|
||||
STORE_FAST, STORE_NAME, STORE_GLOBAL, STORE_DEREF, STORE_ATTR,
|
||||
STORE_SLICE_0, STORE_SLICE_1, STORE_SLICE_2, STORE_SLICE_3,
|
||||
STORE_SUBSCR, UNPACK_SEQUENCE, JA
|
||||
STORE_SUBSCR, UNPACK_SEQUENCE, JUMP_ABSOLUTE
|
||||
}
|
||||
|
||||
prelim = self.all_instr(start, end, stmt_opcodes)
|
||||
# Compose preliminary list of indices with statements,
|
||||
# using plain statement opcodes
|
||||
prelim = self.all_instr(start, end, statement_opcodes)
|
||||
|
||||
# Initialize final container with statements with
|
||||
# preliminnary data
|
||||
stmts = self.stmts = set(prelim)
|
||||
|
||||
# Same for opcode sequences
|
||||
pass_stmts = set()
|
||||
for seq in stmt_opcode_seqs:
|
||||
for i in self.op_range(start, end-(len(seq)+1)):
|
||||
for sequence in statement_opcode_sequences:
|
||||
for i in self.op_range(start, end-(len(sequence)+1)):
|
||||
match = True
|
||||
for elem in seq:
|
||||
for elem in sequence:
|
||||
if elem != code[i]:
|
||||
match = False
|
||||
break
|
||||
i += self.op_size(code[i])
|
||||
|
||||
if match:
|
||||
i = self.prev[i]
|
||||
if match is True:
|
||||
i = self.prev_op[i]
|
||||
stmts.add(i)
|
||||
pass_stmts.add(i)
|
||||
|
||||
# Initialize statement list with the full data we've gathered so far
|
||||
if pass_stmts:
|
||||
stmt_list = list(stmts)
|
||||
stmt_list.sort()
|
||||
stmt_offset_list = list(stmts)
|
||||
stmt_offset_list.sort()
|
||||
else:
|
||||
stmt_list = prelim
|
||||
last_stmt = -1
|
||||
self.next_stmt = []
|
||||
slist = self.next_stmt = []
|
||||
stmt_offset_list = prelim
|
||||
# 'List-map' which contains offset of start of
|
||||
# next statement, when op offset is passed as index
|
||||
self.next_stmt = slist = []
|
||||
last_stmt_offset = -1
|
||||
i = 0
|
||||
for s in stmt_list:
|
||||
if code[s] == JA and s not in pass_stmts:
|
||||
target = self.get_target(s)
|
||||
if target > s or self.lines[last_stmt].l_no == self.lines[s].l_no:
|
||||
stmts.remove(s)
|
||||
# Go through all statement offsets
|
||||
for stmt_offset in stmt_offset_list:
|
||||
# Process absolute jumps, but do not remove 'pass' statements
|
||||
# from the set
|
||||
if code[stmt_offset] == JUMP_ABSOLUTE and stmt_offset not in pass_stmts:
|
||||
# If absolute jump occurs in forward direction or it takes off from the
|
||||
# same line as previous statement, this is not a statement
|
||||
target = self.get_target(stmt_offset)
|
||||
if target > stmt_offset or self.lines[last_stmt_offset].l_no == self.lines[stmt_offset].l_no:
|
||||
stmts.remove(stmt_offset)
|
||||
continue
|
||||
j = self.prev[s]
|
||||
while code[j] == JA:
|
||||
j = self.prev[j]
|
||||
if code[j] == LIST_APPEND: # list comprehension
|
||||
stmts.remove(s)
|
||||
# Rewing ops till we encounter non-JA one
|
||||
j = self.prev_op[stmt_offset]
|
||||
while code[j] == JUMP_ABSOLUTE:
|
||||
j = self.prev_op[j]
|
||||
# If we got here, then it's list comprehension which
|
||||
# is not a statement too
|
||||
if code[j] == LIST_APPEND:
|
||||
stmts.remove(stmt_offset)
|
||||
continue
|
||||
elif code[s] == POP_TOP and code[self.prev[s]] == ROT_TWO:
|
||||
stmts.remove(s)
|
||||
# Exclude ROT_TWO + POP_TOP
|
||||
elif code[stmt_offset] == POP_TOP and code[self.prev_op[stmt_offset]] == ROT_TWO:
|
||||
stmts.remove(stmt_offset)
|
||||
continue
|
||||
elif code[s] in designator_ops:
|
||||
j = self.prev[s]
|
||||
# Exclude FOR_ITER + designators
|
||||
elif code[stmt_offset] in designator_ops:
|
||||
j = self.prev_op[stmt_offset]
|
||||
while code[j] in designator_ops:
|
||||
j = self.prev[j]
|
||||
j = self.prev_op[j]
|
||||
if code[j] == FOR_ITER:
|
||||
stmts.remove(s)
|
||||
stmts.remove(stmt_offset)
|
||||
continue
|
||||
last_stmt = s
|
||||
slist += [s] * (s-i)
|
||||
i = s
|
||||
slist += [end] * (end-len(slist))
|
||||
# Add to list another list with offset of current statement,
|
||||
# equal to length of previous statement
|
||||
slist += [stmt_offset] * (stmt_offset-i)
|
||||
last_stmt_offset = stmt_offset
|
||||
i = stmt_offset
|
||||
# Finish filling the list for last statement
|
||||
slist += [codelen] * (codelen-len(slist))
|
||||
|
||||
def remove_mid_line_ifs(self, ifs):
|
||||
filtered = []
|
||||
for i in ifs:
|
||||
if self.lines[i].l_no == self.lines[i+3].l_no:
|
||||
if self.code[self.prev[self.lines[i].next]] in (PJIT, PJIF):
|
||||
continue
|
||||
filtered.append(i)
|
||||
return filtered
|
||||
def get_target(self, offset):
|
||||
"""
|
||||
Get target offset for op located at given <offset>.
|
||||
"""
|
||||
op = self.code[offset]
|
||||
target = self.code[offset+1] + self.code[offset+2] * 256
|
||||
if op in dis.hasjrel:
|
||||
target += offset + 3
|
||||
return target
|
||||
|
||||
def next_except_jump(self, start):
|
||||
'''
|
||||
Return the next jump that was generated by an except SomeException:
|
||||
construct in a try...except...else clause or None if not found.
|
||||
'''
|
||||
|
||||
if self.code[start] == DUP_TOP:
|
||||
except_match = self.first_instr(start, len(self.code), POP_JUMP_IF_FALSE)
|
||||
if except_match:
|
||||
jmp = self.prev[self.get_target(except_match)]
|
||||
self.ignore_if.add(except_match)
|
||||
self.not_continue.add(jmp)
|
||||
return jmp
|
||||
|
||||
count_END_FINALLY = 0
|
||||
count_SETUP_ = 0
|
||||
for i in self.op_range(start, len(self.code)):
|
||||
op = self.code[i]
|
||||
if op == END_FINALLY:
|
||||
if count_END_FINALLY == count_SETUP_:
|
||||
assert self.code[self.prev[i]] in (JA, JF, RETURN_VALUE)
|
||||
self.not_continue.add(self.prev[i])
|
||||
return self.prev[i]
|
||||
count_END_FINALLY += 1
|
||||
elif op in (SETUP_EXCEPT, SETUP_WITH, SETUP_FINALLY):
|
||||
count_SETUP_ += 1
|
||||
|
||||
def detect_structure(self, pos, op=None):
|
||||
'''
|
||||
Detect type of block structures and their boundaries to fix optimizied jumps
|
||||
def detect_structure(self, offset):
|
||||
"""
|
||||
Detect structures and their boundaries to fix optimizied jumps
|
||||
in python2.3+
|
||||
'''
|
||||
|
||||
# TODO: check the struct boundaries more precisely -Dan
|
||||
|
||||
"""
|
||||
code = self.code
|
||||
# Ev remove this test and make op a mandatory argument -Dan
|
||||
if op is None:
|
||||
op = code[pos]
|
||||
|
||||
op = code[offset]
|
||||
# Detect parent structure
|
||||
parent = self.structs[0]
|
||||
start = parent['start']
|
||||
end = parent['end']
|
||||
for s in self.structs:
|
||||
_start = s['start']
|
||||
_end = s['end']
|
||||
if (_start <= pos < _end) and (_start >= start and _end <= end):
|
||||
start = _start
|
||||
end = _end
|
||||
parent = s
|
||||
|
||||
if op == SETUP_LOOP:
|
||||
start = pos+3
|
||||
target = self.get_target(pos, op)
|
||||
end = self.restrict_to_parent(target, parent)
|
||||
# Pick inner-most parent for our offset
|
||||
for struct in self.structs:
|
||||
curent_start = struct['start']
|
||||
curent_end = struct['end']
|
||||
if (curent_start <= offset < curent_end) and (curent_start >= start and curent_end <= end):
|
||||
start = curent_start
|
||||
end = curent_end
|
||||
parent = struct
|
||||
|
||||
if target != end:
|
||||
self.fixed_jumps[pos] = end
|
||||
(line_no, next_line_byte) = self.lines[pos]
|
||||
jump_back = self.last_instr(start, end, JA,
|
||||
next_line_byte, False)
|
||||
|
||||
if jump_back and jump_back != self.prev[end] and code[jump_back+3] in (JA, JF):
|
||||
if code[self.prev[end]] == RETURN_VALUE or \
|
||||
(code[self.prev[end]] == POP_BLOCK and code[self.prev[self.prev[end]]] == RETURN_VALUE):
|
||||
jump_back = None
|
||||
if not jump_back: # loop suite ends in return. wtf right?
|
||||
jump_back = self.last_instr(start, end, RETURN_VALUE) + 1
|
||||
if not jump_back:
|
||||
return
|
||||
if code[self.prev[next_line_byte]] not in (PJIF, PJIT):
|
||||
loop_type = 'for'
|
||||
else:
|
||||
loop_type = 'while'
|
||||
self.ignore_if.add(self.prev[next_line_byte])
|
||||
target = next_line_byte
|
||||
end = jump_back + 3
|
||||
else:
|
||||
if self.get_target(jump_back) >= next_line_byte:
|
||||
jump_back = self.last_instr(start, end, JA, start, False)
|
||||
if end > jump_back+4 and code[end] in (JF, JA):
|
||||
if code[jump_back+4] in (JA, JF):
|
||||
if self.get_target(jump_back+4) == self.get_target(end):
|
||||
self.fixed_jumps[pos] = jump_back+4
|
||||
end = jump_back+4
|
||||
elif target < pos:
|
||||
self.fixed_jumps[pos] = jump_back+4
|
||||
end = jump_back+4
|
||||
target = self.get_target(jump_back, JA)
|
||||
|
||||
if code[target] in (FOR_ITER, GET_ITER):
|
||||
loop_type = 'for'
|
||||
else:
|
||||
loop_type = 'while'
|
||||
test = self.prev[next_line_byte]
|
||||
if test == pos:
|
||||
loop_type = 'while 1'
|
||||
elif self.code[test] in hasjabs+hasjrel:
|
||||
self.ignore_if.add(test)
|
||||
test_target = self.get_target(test)
|
||||
if test_target > (jump_back+3):
|
||||
jump_back = test_target
|
||||
self.not_continue.add(jump_back)
|
||||
self.loops.append(target)
|
||||
self.structs.append({'type': loop_type + '-loop',
|
||||
'start': target,
|
||||
'end': jump_back})
|
||||
if jump_back+3 != end:
|
||||
self.structs.append({'type': loop_type + '-else',
|
||||
'start': jump_back+3,
|
||||
'end': end})
|
||||
elif op == SETUP_EXCEPT:
|
||||
start = pos+3
|
||||
target = self.get_target(pos, op)
|
||||
end = self.restrict_to_parent(target, parent)
|
||||
if target != end:
|
||||
self.fixed_jumps[pos] = end
|
||||
# print target, end, parent
|
||||
# Add the try block
|
||||
self.structs.append({'type': 'try',
|
||||
'start': start,
|
||||
'end': end-4})
|
||||
# Now isolate the except and else blocks
|
||||
end_else = start_else = self.get_target(self.prev[end])
|
||||
|
||||
# Add the except blocks
|
||||
i = end
|
||||
while self.code[i] != END_FINALLY:
|
||||
jmp = self.next_except_jump(i)
|
||||
if self.code[jmp] == RETURN_VALUE:
|
||||
self.structs.append({'type': 'except',
|
||||
'start': i,
|
||||
'end': jmp+1})
|
||||
i = jmp + 1
|
||||
else:
|
||||
if self.get_target(jmp) != start_else:
|
||||
end_else = self.get_target(jmp)
|
||||
if self.code[jmp] == JF:
|
||||
self.fixed_jumps[jmp] = -1
|
||||
self.structs.append({'type': 'except',
|
||||
'start': i,
|
||||
'end': jmp})
|
||||
i = jmp + 3
|
||||
|
||||
# Add the try-else block
|
||||
if end_else != start_else:
|
||||
r_end_else = self.restrict_to_parent(end_else, parent)
|
||||
self.structs.append({'type': 'try-else',
|
||||
'start': i+1,
|
||||
'end': r_end_else})
|
||||
self.fixed_jumps[i] = r_end_else
|
||||
else:
|
||||
self.fixed_jumps[i] = i+1
|
||||
|
||||
elif op in (PJIF, PJIT):
|
||||
start = pos+3
|
||||
target = self.get_target(pos, op)
|
||||
if op in (POP_JUMP_IF_FALSE, POP_JUMP_IF_TRUE):
|
||||
start = offset + self.op_size(op)
|
||||
target = self.get_target(offset)
|
||||
rtarget = self.restrict_to_parent(target, parent)
|
||||
pre = self.prev
|
||||
prev_op = self.prev_op
|
||||
|
||||
# Do not let jump to go out of parent struct bounds
|
||||
if target != rtarget and parent['type'] == 'and/or':
|
||||
self.fixed_jumps[pos] = rtarget
|
||||
self.fixed_jumps[offset] = rtarget
|
||||
return
|
||||
# does this jump to right after another cond jump?
|
||||
# if so, it's part of a larger conditional
|
||||
if (code[pre[target]] in (JUMP_IF_FALSE_OR_POP, JUMP_IF_TRUE_OR_POP,
|
||||
PJIF, PJIT)) and (target > pos):
|
||||
self.fixed_jumps[pos] = pre[target]
|
||||
|
||||
# Does this jump to right after another cond jump?
|
||||
# If so, it's part of a larger conditional
|
||||
if (code[prev_op[target]] in (JUMP_IF_FALSE_OR_POP, JUMP_IF_TRUE_OR_POP,
|
||||
POP_JUMP_IF_FALSE, POP_JUMP_IF_TRUE)) and (target > offset):
|
||||
self.fixed_jumps[offset] = prev_op[target]
|
||||
self.structs.append({'type': 'and/or',
|
||||
'start': start,
|
||||
'end': pre[target]})
|
||||
'end': prev_op[target]})
|
||||
return
|
||||
|
||||
# is this an if and
|
||||
if op == PJIF:
|
||||
match = self.rem_or(start, self.next_stmt[pos], PJIF, target)
|
||||
# Is it an and inside if block
|
||||
if op == POP_JUMP_IF_FALSE:
|
||||
# Search for other POP_JUMP_IF_FALSE targetting the same op,
|
||||
# in current statement, starting from current offset, and filter
|
||||
# everything inside inner 'or' jumps and midline ifs
|
||||
match = self.rem_or(start, self.next_stmt[offset], POP_JUMP_IF_FALSE, target)
|
||||
match = self.remove_mid_line_ifs(match)
|
||||
|
||||
# If we still have any offsets in set, start working on it
|
||||
if match:
|
||||
if code[pre[rtarget]] in (JF, JA) \
|
||||
and pre[rtarget] not in self.stmts \
|
||||
and self.restrict_to_parent(self.get_target(pre[rtarget]), parent) == rtarget:
|
||||
if code[pre[pre[rtarget]]] == JA \
|
||||
and self.remove_mid_line_ifs([pos]) \
|
||||
and target == self.get_target(pre[pre[rtarget]]) \
|
||||
and (pre[pre[rtarget]] not in self.stmts or self.get_target(pre[pre[rtarget]]) > pre[pre[rtarget]])\
|
||||
and 1 == len(self.remove_mid_line_ifs(self.rem_or(start, pre[pre[rtarget]], (PJIF, PJIT), target))):
|
||||
if (code[prev_op[rtarget]] in (JUMP_FORWARD, JUMP_ABSOLUTE) and prev_op[rtarget] not in self.stmts and
|
||||
self.restrict_to_parent(self.get_target(prev_op[rtarget]), parent) == rtarget):
|
||||
if (code[prev_op[prev_op[rtarget]]] == JUMP_ABSOLUTE and self.remove_mid_line_ifs([offset]) and
|
||||
target == self.get_target(prev_op[prev_op[rtarget]]) and
|
||||
(prev_op[prev_op[rtarget]] not in self.stmts or self.get_target(prev_op[prev_op[rtarget]]) > prev_op[prev_op[rtarget]]) and
|
||||
1 == len(self.remove_mid_line_ifs(self.rem_or(start, prev_op[prev_op[rtarget]], (POP_JUMP_IF_FALSE, POP_JUMP_IF_TRUE), target)))):
|
||||
pass
|
||||
elif code[pre[pre[rtarget]]] == RETURN_VALUE \
|
||||
and self.remove_mid_line_ifs([pos]) \
|
||||
and 1 == (len(set(self.remove_mid_line_ifs(self.rem_or(start,
|
||||
pre[pre[rtarget]],
|
||||
(PJIF, PJIT), target)))
|
||||
| set(self.remove_mid_line_ifs(self.rem_or(start, pre[pre[rtarget]],
|
||||
(PJIF, PJIT, JA), pre[rtarget], True))))):
|
||||
elif (code[prev_op[prev_op[rtarget]]] == RETURN_VALUE and self.remove_mid_line_ifs([offset]) and
|
||||
1 == (len(set(self.remove_mid_line_ifs(self.rem_or(start, prev_op[prev_op[rtarget]],
|
||||
(POP_JUMP_IF_FALSE, POP_JUMP_IF_TRUE), target))) |
|
||||
set(self.remove_mid_line_ifs(self.rem_or(start, prev_op[prev_op[rtarget]],
|
||||
(POP_JUMP_IF_FALSE, POP_JUMP_IF_TRUE, JUMP_ABSOLUTE),
|
||||
prev_op[rtarget], True)))))):
|
||||
pass
|
||||
else:
|
||||
fix = None
|
||||
jump_ifs = self.all_instr(start, self.next_stmt[pos], PJIF)
|
||||
jump_ifs = self.all_instr(start, self.next_stmt[offset], POP_JUMP_IF_FALSE)
|
||||
last_jump_good = True
|
||||
for j in jump_ifs:
|
||||
if target == self.get_target(j):
|
||||
if self.lines[j].next == j+3 and last_jump_good:
|
||||
if self.lines[j].next == j + 3 and last_jump_good:
|
||||
fix = j
|
||||
break
|
||||
else:
|
||||
last_jump_good = False
|
||||
self.fixed_jumps[pos] = fix or match[-1]
|
||||
self.fixed_jumps[offset] = fix or match[-1]
|
||||
return
|
||||
else:
|
||||
self.fixed_jumps[pos] = match[-1]
|
||||
self.fixed_jumps[offset] = match[-1]
|
||||
return
|
||||
else: # op == PJIT
|
||||
if (pos+3) in self.load_asserts:
|
||||
if code[pre[rtarget]] == RAISE_VARARGS:
|
||||
return
|
||||
self.load_asserts.remove(pos+3)
|
||||
|
||||
next = self.next_stmt[pos]
|
||||
if pre[next] == pos:
|
||||
pass
|
||||
elif code[next] in (JF, JA) and target == self.get_target(next):
|
||||
if code[pre[next]] == PJIF:
|
||||
if code[next] == JF or target != rtarget or code[pre[pre[rtarget]]] not in (JA, RETURN_VALUE):
|
||||
self.fixed_jumps[pos] = pre[next]
|
||||
return
|
||||
elif code[next] == JA and code[target] in (JA, JF):
|
||||
next_target = self.get_target(next)
|
||||
if self.get_target(target) == next_target:
|
||||
self.fixed_jumps[pos] = pre[next]
|
||||
return
|
||||
elif code[next_target] in (JA, JF) and self.get_target(next_target) == self.get_target(target):
|
||||
self.fixed_jumps[pos] = pre[next]
|
||||
return
|
||||
|
||||
# don't add a struct for a while test, it's already taken care of
|
||||
if pos in self.ignore_if:
|
||||
return
|
||||
|
||||
if code[pre[rtarget]] == JA and pre[rtarget] in self.stmts \
|
||||
and pre[rtarget] != pos and pre[pre[rtarget]] != pos:
|
||||
if code[rtarget] == JA and code[rtarget+3] == POP_BLOCK:
|
||||
if code[pre[pre[rtarget]]] != JA:
|
||||
pass
|
||||
elif self.get_target(pre[pre[rtarget]]) != target:
|
||||
pass
|
||||
# op == POP_JUMP_IF_TRUE
|
||||
else:
|
||||
rtarget = pre[rtarget]
|
||||
else:
|
||||
rtarget = pre[rtarget]
|
||||
# does the if jump just beyond a jump op, then this is probably an if statement
|
||||
if code[pre[rtarget]] in (JA, JF):
|
||||
if_end = self.get_target(pre[rtarget])
|
||||
next = self.next_stmt[offset]
|
||||
if prev_op[next] == offset:
|
||||
pass
|
||||
elif code[next] in (JUMP_FORWARD, JUMP_ABSOLUTE) and target == self.get_target(next):
|
||||
if code[prev_op[next]] == POP_JUMP_IF_FALSE:
|
||||
if code[next] == JUMP_FORWARD or target != rtarget or code[prev_op[prev_op[rtarget]]] not in (JUMP_ABSOLUTE, RETURN_VALUE):
|
||||
self.fixed_jumps[offset] = prev_op[next]
|
||||
return
|
||||
elif (code[next] == JUMP_ABSOLUTE and code[target] in (JUMP_ABSOLUTE, JUMP_FORWARD) and
|
||||
self.get_target(target) == self.get_target(next)):
|
||||
self.fixed_jumps[offset] = prev_op[next]
|
||||
return
|
||||
|
||||
# is this a loop not an if?
|
||||
if (if_end < pre[rtarget]) and (code[pre[if_end]] == SETUP_LOOP):
|
||||
# Don't add a struct for a while test, it's already taken care of
|
||||
if offset in self.ignore_if:
|
||||
return
|
||||
|
||||
if (code[prev_op[rtarget]] == JUMP_ABSOLUTE and prev_op[rtarget] in self.stmts and
|
||||
prev_op[rtarget] != offset and prev_op[prev_op[rtarget]] != offset and
|
||||
not (code[rtarget] == JUMP_ABSOLUTE and code[rtarget+3] == POP_BLOCK and code[prev_op[prev_op[rtarget]]] != JUMP_ABSOLUTE)):
|
||||
rtarget = prev_op[rtarget]
|
||||
|
||||
# Does the if jump just beyond a jump op, then this is probably an if statement
|
||||
if code[prev_op[rtarget]] in (JUMP_ABSOLUTE, JUMP_FORWARD):
|
||||
if_end = self.get_target(prev_op[rtarget])
|
||||
|
||||
# Is this a loop not an if?
|
||||
if (if_end < prev_op[rtarget]) and (code[prev_op[if_end]] == SETUP_LOOP):
|
||||
if(if_end > start):
|
||||
return
|
||||
|
||||
@@ -561,65 +412,66 @@ class Scanner27(scan.Scanner):
|
||||
|
||||
self.structs.append({'type': 'if-then',
|
||||
'start': start,
|
||||
'end': pre[rtarget]})
|
||||
self.not_continue.add(pre[rtarget])
|
||||
'end': prev_op[rtarget]})
|
||||
self.not_continue.add(prev_op[rtarget])
|
||||
|
||||
if rtarget < end:
|
||||
self.structs.append({'type': 'if-else',
|
||||
'start': rtarget,
|
||||
'end': end})
|
||||
elif code[pre[rtarget]] == RETURN_VALUE:
|
||||
elif code[prev_op[rtarget]] == RETURN_VALUE:
|
||||
self.structs.append({'type': 'if-then',
|
||||
'start': start,
|
||||
'end': rtarget})
|
||||
self.return_end_ifs.add(pre[rtarget])
|
||||
self.return_end_ifs.add(prev_op[rtarget])
|
||||
|
||||
elif op in (JUMP_IF_FALSE_OR_POP, JUMP_IF_TRUE_OR_POP):
|
||||
target = self.get_target(pos, op)
|
||||
self.fixed_jumps[pos] = self.restrict_to_parent(target, parent)
|
||||
target = self.get_target(offset)
|
||||
if target > offset:
|
||||
unop_target = self.last_instr(offset, target, JUMP_FORWARD, target)
|
||||
if unop_target and code[unop_target+3] != ROT_TWO:
|
||||
self.fixed_jumps[offset] = unop_target
|
||||
else:
|
||||
self.fixed_jumps[offset] = self.restrict_to_parent(target, parent)
|
||||
|
||||
def find_jump_targets(self, code):
|
||||
'''
|
||||
Detect all offsets in a byte code which are jump targets.
|
||||
def restrict_to_parent(self, target, parent):
|
||||
"""Restrict target to parent structure boundaries."""
|
||||
if not (parent['start'] < target < parent['end']):
|
||||
target = parent['end']
|
||||
return target
|
||||
|
||||
Return the list of offsets.
|
||||
def rem_or(self, start, end, instr, target=None, include_beyond_target=False):
|
||||
"""
|
||||
Find offsets of all requested <instr> between <start> and <end>,
|
||||
optionally <target>ing specified offset, and return list found
|
||||
<instr> offsets which are not within any POP_JUMP_IF_TRUE jumps.
|
||||
"""
|
||||
# Find all offsets of requested instructions
|
||||
instr_offsets = self.all_instr(start, end, instr, target, include_beyond_target)
|
||||
# Get all POP_JUMP_IF_TRUE (or) offsets
|
||||
pjit_offsets = self.all_instr(start, end, POP_JUMP_IF_TRUE)
|
||||
filtered = []
|
||||
for pjit_offset in pjit_offsets:
|
||||
pjit_tgt = self.get_target(pjit_offset) - 3
|
||||
for instr_offset in instr_offsets:
|
||||
if instr_offset <= pjit_offset or instr_offset >= pjit_tgt:
|
||||
filtered.append(instr_offset)
|
||||
instr_offsets = filtered
|
||||
filtered = []
|
||||
return instr_offsets
|
||||
|
||||
This procedure is modelled after dis.findlables(), but here
|
||||
for each target the number of jumps are counted.
|
||||
'''
|
||||
|
||||
n = len(code)
|
||||
self.structs = [{'type': 'root',
|
||||
'start': 0,
|
||||
'end': n-1}]
|
||||
self.loops = [] # All loop entry points
|
||||
self.fixed_jumps = {} # Map fixed jumps to their real destination
|
||||
self.ignore_if = set()
|
||||
self.build_stmt_indices()
|
||||
self.not_continue = set()
|
||||
self.return_end_ifs = set()
|
||||
|
||||
targets = {}
|
||||
for i in self.op_range(0, n):
|
||||
op = code[i]
|
||||
|
||||
# Determine structures and fix jumps for 2.3+
|
||||
self.detect_structure(i, op)
|
||||
|
||||
if op >= HAVE_ARGUMENT:
|
||||
label = self.fixed_jumps.get(i)
|
||||
oparg = code[i+1] + code[i+2] * 256
|
||||
if label is None:
|
||||
if op in hasjrel and op != FOR_ITER:
|
||||
label = i + 3 + oparg
|
||||
elif op in hasjabs:
|
||||
if op in (JUMP_IF_FALSE_OR_POP, JUMP_IF_TRUE_OR_POP):
|
||||
if (oparg > i):
|
||||
label = oparg
|
||||
|
||||
if label is not None and label != -1:
|
||||
targets[label] = targets.get(label, []) + [i]
|
||||
elif op == END_FINALLY and i in self.fixed_jumps:
|
||||
label = self.fixed_jumps[i]
|
||||
targets[label] = targets.get(label, []) + [i]
|
||||
return targets
|
||||
def remove_mid_line_ifs(self, ifs):
|
||||
"""
|
||||
Go through passed offsets, filtering ifs
|
||||
located somewhere mid-line.
|
||||
"""
|
||||
filtered = []
|
||||
for if_ in ifs:
|
||||
# For each offset, if line number of current and next op
|
||||
# is the same
|
||||
if self.lines[if_].l_no == self.lines[if_+3].l_no:
|
||||
# Check if last op on line is PJIT or PJIF, and if it is - skip it
|
||||
if self.code[self.prev_op[self.lines[if_].next]] in (POP_JUMP_IF_TRUE, POP_JUMP_IF_FALSE):
|
||||
continue
|
||||
filtered.append(if_)
|
||||
return filtered
|
||||
|
||||
Reference in New Issue
Block a user