angr 9.2.166__tar.gz → 9.2.167__tar.gz

{angr-9.2.166 → angr-9.2.167}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: angr
-Version: 9.2.166
+Version: 9.2.167
 Summary: A multi-architecture binary analysis toolkit, with the ability to perform dynamic symbolic execution and various static analyses on binaries
 License: BSD-2-Clause
 Project-URL: Homepage, https://angr.io/
@@ -16,12 +16,12 @@ Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: cxxheaderparser
 Requires-Dist: GitPython
-Requires-Dist: archinfo==9.2.166
+Requires-Dist: archinfo==9.2.167
 Requires-Dist: cachetools
 Requires-Dist: capstone==5.0.3
 Requires-Dist: cffi>=1.14.0
-Requires-Dist: claripy==9.2.166
-Requires-Dist: cle==9.2.166
+Requires-Dist: claripy==9.2.167
+Requires-Dist: cle==9.2.167
 Requires-Dist: mulpyplexer
 Requires-Dist: networkx!=2.8.1,>=2.0
 Requires-Dist: protobuf>=5.28.2
@@ -30,7 +30,7 @@ Requires-Dist: pycparser>=2.18
 Requires-Dist: pydemumble
 Requires-Dist: pyformlang
 Requires-Dist: pypcode<4.0,>=3.2.1
-Requires-Dist: pyvex==9.2.166
+Requires-Dist: pyvex==9.2.167
 Requires-Dist: rich>=13.1.0
 Requires-Dist: sortedcontainers
 Requires-Dist: sympy

{angr-9.2.166 → angr-9.2.167}/angr/__init__.py

@@ -2,7 +2,7 @@
 # pylint: disable=wrong-import-position
 from __future__ import annotations
 
-__version__ = "9.2.166"
+__version__ = "9.2.167"
 
 if bytes is str:
     raise Exception(

{angr-9.2.166 → angr-9.2.167}/angr/analyses/cfg/indirect_jump_resolvers/jumptable.py

@@ -920,32 +920,32 @@ class JumpTableResolver(IndirectJumpResolver):
         # more sanity checks
 
         # for a typical jump table, the current block has only one predecessor, and the predecessor to the current
-        # block has two successors (not including itself)
+        # block has two successors
         # for a typical vtable call (or jump if at the end of a function), the block as two predecessors that form a
         # diamond shape
         curr_node = func.get_node(addr)
-        if curr_node is None or curr_node not in func.graph:
+        if curr_node is None or curr_node not in func.transition_graph:
             l.debug("Could not find the node %#x in the function transition graph", addr)
             return False, None
-        preds = list(func.graph.predecessors(curr_node))
+        preds = list(func.transition_graph.predecessors(curr_node))
         pred_endaddrs = {pred.addr + pred.size for pred in preds}  # handle non-normalized CFGs
         if func_graph_complete and not is_arm and not potential_call_table:
             # on ARM you can do a single-block jump table...
             if len(pred_endaddrs) == 1:
-                pred_succs = [succ for succ in func.graph.successors(preds[0]) if succ.addr != preds[0].addr]
+                pred_succs = [succ for succ in func.transition_graph.successors(preds[0]) if succ.addr != preds[0].addr]
                 if len(pred_succs) != 2:
                     l.debug("Expect two successors to the single predecessor, found %d.", len(pred_succs))
                     return False, None
             elif len(pred_endaddrs) == 2 and len(preds) == 2:
                 pred_succs = set(
-                    [succ for succ in func.graph.successors(preds[0]) if succ.addr != preds[0].addr]
-                    + [succ for succ in func.graph.successors(preds[1]) if succ.addr != preds[1].addr]
+                    [succ for succ in func.transition_graph.successors(preds[0]) if succ.addr != preds[0].addr]
+                    + [succ for succ in func.transition_graph.successors(preds[1]) if succ.addr != preds[1].addr]
                 )
                 is_diamond = False
                 if len(pred_succs) == 2:
                     non_node_succ = next(iter(pred_succ for pred_succ in pred_succs if pred_succ is not curr_node))
-                    while func.graph.out_degree[non_node_succ] == 1:
-                        non_node_succ = next(iter(func.graph.successors(non_node_succ)))
+                    while func.transition_graph.out_degree[non_node_succ] == 1:
+                        non_node_succ = next(iter(func.transition_graph.successors(non_node_succ)))
                         if non_node_succ == curr_node:
                             is_diamond = True
                             break

{angr-9.2.166 → angr-9.2.167}/angr/analyses/decompiler/clinic.py

@@ -142,6 +142,7 @@ class Clinic(Analysis):
         optimization_scratch: dict[str, Any] | None = None,
         desired_variables: set[str] | None = None,
         force_loop_single_exit: bool = True,
+        refine_loops_with_single_successor: bool = False,
         complete_successors: bool = False,
         max_type_constraints: int = 100_000,
         type_constraint_set_degradation_threshold: int = 150,
@@ -212,6 +213,7 @@ class Clinic(Analysis):
         self._inlining_parents = inlining_parents or ()
         self._desired_variables = desired_variables
         self._force_loop_single_exit = force_loop_single_exit
+        self._refine_loops_with_single_successor = refine_loops_with_single_successor
         self._complete_successors = complete_successors
 
         self._register_save_areas_removed: bool = False
@@ -1550,6 +1552,7 @@ class Clinic(Analysis):
                 entry_node_addr=self.entry_node_addr,
                 scratch=self.optimization_scratch,
                 force_loop_single_exit=self._force_loop_single_exit,
+                refine_loops_with_single_successor=self._refine_loops_with_single_successor,
                 complete_successors=self._complete_successors,
                 stack_pointer_tracker=stack_pointer_tracker,
                 **kwargs,

{angr-9.2.166 → angr-9.2.167}/angr/analyses/decompiler/condition_processor.py

@@ -239,6 +239,24 @@ class ConditionProcessor:
         condition translation if possible.
         """
 
+        if isinstance(src, SequenceNode) and src.nodes and isinstance(src.nodes[-1], ConditionNode):
+            cond_node = src.nodes[-1]
+            if (
+                isinstance(cond_node.true_node, ailment.Block)
+                and isinstance(cond_node.false_node, ailment.Block)
+                and cond_node.true_node.statements
+                and cond_node.false_node.statements
+            ):
+                last_stmt_true = self.get_last_statement(cond_node.true_node)
+                last_stmt_false = self.get_last_statement(cond_node.false_node)
+                if (
+                    isinstance(last_stmt_true, ailment.Stmt.Jump)
+                    and isinstance(last_stmt_false, ailment.Stmt.Jump)
+                    and isinstance(last_stmt_true.target, ailment.Expr.Const)
+                    and isinstance(last_stmt_false.target, ailment.Expr.Const)
+                ):
+                    return {last_stmt_true.target.value, last_stmt_false.target.value} == {dst0.addr, dst1.addr}
+
         if src in graph and graph.out_degree[src] == 2 and graph.has_edge(src, dst0) and graph.has_edge(src, dst1):
             # sometimes the last statement is the conditional jump. sometimes it's the first statement of the block
             if isinstance(src, ailment.Block) and src.statements and is_head_controlled_loop_block(src):
@@ -247,7 +265,10 @@ class ConditionProcessor:
                 )
                 assert last_stmt is not None
             else:
-                last_stmt = self.get_last_statement(src)
+                try:
+                    last_stmt = self.get_last_statement(src)
+                except EmptyBlockNotice:
+                    last_stmt = None
 
             if isinstance(last_stmt, ailment.Stmt.ConditionalJump):
                 return True
@@ -258,6 +279,28 @@ class ConditionProcessor:
         return claripy.is_true(claripy.Not(edge_cond_left) == edge_cond_right)  # type: ignore
 
     def recover_edge_condition(self, graph: networkx.DiGraph, src, dst):
+
+        def _check_condnode_and_get_condition(cond_node: ConditionNode) -> claripy.ast.Bool | None:
+            for cond_block, negate in [(cond_node.true_node, False), (cond_node.false_node, True)]:
+                if isinstance(cond_block, ailment.Block) and cond_block.statements:
+                    last_stmt = self.get_last_statement(cond_block)
+                    if (
+                        isinstance(last_stmt, ailment.Stmt.Jump)
+                        and isinstance(last_stmt.target, ailment.Expr.Const)
+                        and last_stmt.target.value == dst.addr
+                    ):
+                        return claripy.Not(cond_node.condition) if negate else cond_node.condition
+            return None
+
+        if isinstance(src, SequenceNode) and src.nodes and isinstance(src.nodes[-1], ConditionNode):
+            predicate = _check_condnode_and_get_condition(src.nodes[-1])
+            if predicate is not None:
+                return predicate
+        if isinstance(src, ConditionNode):
+            predicate = _check_condnode_and_get_condition(src)
+            if predicate is not None:
+                return predicate
+
         edge = src, dst
         edge_data = graph.get_edge_data(*edge)
         edge_type = edge_data.get("type", "transition") if edge_data is not None else "transition"

{angr-9.2.166 → angr-9.2.167}/angr/analyses/decompiler/decompiler.py

@@ -222,6 +222,7 @@ class Decompiler(Analysis):
         # determine a few arguments according to the structuring algorithm
         fold_callexprs_into_conditions = False
         self._force_loop_single_exit = True
+        self._refine_loops_with_single_successor = False
         self._complete_successors = False
         self._recursive_structurer_params = self.options_to_params(self.options_by_class["recursive_structurer"])
         if "structurer_cls" not in self._recursive_structurer_params:
@@ -229,6 +230,7 @@ class Decompiler(Analysis):
         # is the algorithm based on Phoenix (a schema-based algorithm)?
         if issubclass(self._recursive_structurer_params["structurer_cls"], PhoenixStructurer):
             self._force_loop_single_exit = False
+            # self._refine_loops_with_single_successor = True
             self._complete_successors = True
             fold_callexprs_into_conditions = True
 
@@ -261,6 +263,7 @@ class Decompiler(Analysis):
                 desired_variables=self._desired_variables,
                 optimization_scratch=self._optimization_scratch,
                 force_loop_single_exit=self._force_loop_single_exit,
+                refine_loops_with_single_successor=self._refine_loops_with_single_successor,
                 complete_successors=self._complete_successors,
                 ail_graph=self._clinic_graph,
                 arg_vvars=self._clinic_arg_vvars,
@@ -396,6 +399,7 @@ class Decompiler(Analysis):
             cond_proc=condition_processor,
             update_graph=update_graph,
             force_loop_single_exit=self._force_loop_single_exit,
+            refine_loops_with_single_successor=self._refine_loops_with_single_successor,
             complete_successors=self._complete_successors,
             entry_node_addr=self.clinic.entry_node_addr,
             **self.options_to_params(self.options_by_class["region_identifier"]),
@@ -444,6 +448,7 @@ class Decompiler(Analysis):
                 entry_node_addr=self.clinic.entry_node_addr,
                 scratch=self._optimization_scratch,
                 force_loop_single_exit=self._force_loop_single_exit,
+                refine_loops_with_single_successor=self._refine_loops_with_single_successor,
                 complete_successors=self._complete_successors,
                 **kwargs,
             )
@@ -507,6 +512,7 @@ class Decompiler(Analysis):
                 entry_node_addr=self.clinic.entry_node_addr,
                 scratch=self._optimization_scratch,
                 force_loop_single_exit=self._force_loop_single_exit,
+                refine_loops_with_single_successor=self._refine_loops_with_single_successor,
                 complete_successors=self._complete_successors,
                 peephole_optimizations=self._peephole_optimizations,
                 avoid_vvar_ids=self._copied_var_ids,

angr-9.2.167/angr/analyses/decompiler/node_replacer.py

@@ -0,0 +1,42 @@
+from __future__ import annotations
+
+from angr.ailment import Block
+from .sequence_walker import SequenceWalker
+from .structuring.structurer_nodes import BaseNode, SequenceNode, MultiNode
+
+
+class NodeReplacer(SequenceWalker):
+    """
+    Replaces nodes in a node with new nodes based on a mapping.
+    """
+
+    def __init__(self, root: BaseNode, replacements: dict) -> None:
+        super().__init__(update_seqnode_in_place=False)
+
+        self.root = root
+        self.replacements = replacements
+        self.result: BaseNode = self.walk(self.root)  # type:ignore
+
+    def _handle(self, node: BaseNode, **kwargs):
+        return self.replacements[node] if node in self.replacements else super()._handle(node, **kwargs)
+
+    def _handle_MultiNode(self, node: MultiNode, **kwargs):
+        changed = False
+        nodes_copy = list(node.nodes)
+
+        i = len(nodes_copy) - 1
+        has_non_block = False
+        while i > -1:
+            node_ = nodes_copy[i]
+            new_node = self._handle(node_, parent=node, index=i)
+            if new_node is not None:
+                changed = True
+                nodes_copy[i] = new_node
+                if not isinstance(new_node, Block):
+                    has_non_block = True
+            i -= 1
+        if not changed:
+            return None
+        if has_non_block:
+            return SequenceNode(node.addr, nodes=nodes_copy)
+        return MultiNode(nodes_copy, addr=node.addr, idx=node.idx)

{angr-9.2.166 → angr-9.2.167}/angr/analyses/decompiler/optimization_passes/lowered_switch_simplifier.py

@@ -163,7 +163,7 @@ class LoweredSwitchSimplifier(StructuringOptimizationPass):
             require_gotos=False,
             prevent_new_gotos=False,
             simplify_ail=False,
-            must_improve_rel_quality=True,
+            must_improve_rel_quality=False,
             **kwargs,
         )
 

{angr-9.2.166 → angr-9.2.167}/angr/analyses/decompiler/optimization_passes/optimization_pass.py

@@ -135,6 +135,7 @@ class OptimizationPass(BaseOptimizationPass):
         entry_node_addr=None,
         scratch: dict[str, Any] | None = None,
         force_loop_single_exit: bool = True,
+        refine_loops_with_single_successor: bool = False,
         complete_successors: bool = False,
         avoid_vvar_ids: set[int] | None = None,
         arg_vvars: set[int] | None = None,
@@ -158,6 +159,7 @@ class OptimizationPass(BaseOptimizationPass):
             entry_node_addr if entry_node_addr is not None else (func.addr, None)
         )
         self._force_loop_single_exit = force_loop_single_exit
+        self._refine_loops_with_single_successor = refine_loops_with_single_successor
         self._complete_successors = complete_successors
         self._avoid_vvar_ids = avoid_vvar_ids or set()
         self._peephole_optimizations = peephole_optimizations
@@ -397,6 +399,7 @@ class OptimizationPass(BaseOptimizationPass):
             cond_proc=condition_processor or ConditionProcessor(self.project.arch),
             update_graph=update_graph,
             force_loop_single_exit=self._force_loop_single_exit,
+            refine_loops_with_single_successor=self._refine_loops_with_single_successor,
             complete_successors=self._complete_successors,
             entry_node_addr=self.entry_node_addr,
         )

{angr-9.2.166 → angr-9.2.167}/angr/analyses/decompiler/optimization_passes/return_duplicator_low.py

@@ -6,7 +6,7 @@ from typing import Any
 import networkx
 
 from angr.ailment import Block
-from angr.ailment.statement import ConditionalJump, Label
+from angr.ailment.statement import ConditionalJump
 
 from .return_duplicator_base import ReturnDuplicatorBase
 from .optimization_pass import StructuringOptimizationPass
@@ -53,7 +53,7 @@ class ReturnDuplicatorLow(StructuringOptimizationPass, ReturnDuplicatorBase):
         prevent_new_gotos: bool = True,
         minimize_copies_for_regions: bool = True,
         region_identifier=None,
-        vvar_id_start: int | None = None,
+        vvar_id_start: int = 0,
         scratch: dict[str, Any] | None = None,
         max_func_blocks: int = 500,
         **kwargs,
@@ -91,8 +91,9 @@ class ReturnDuplicatorLow(StructuringOptimizationPass, ReturnDuplicatorBase):
         self,
         src: Block,
         dst: Block,
-        graph: networkx.DiGraph = None,
         max_level_check=1,
+        *,
+        graph: networkx.DiGraph,
     ):
         """
         TODO: Implement a more principled way of checking if an edge is a goto edge with Phoenix's structuring info
@@ -100,6 +101,7 @@ class ReturnDuplicatorLow(StructuringOptimizationPass, ReturnDuplicatorBase):
         above a goto edge as the goto src.
         """
         # Do a simple and fast check first
+        assert self._goto_manager is not None
         is_simple_goto = self._goto_manager.is_goto_edge(src, dst)
         if is_simple_goto:
             return True
@@ -155,79 +157,6 @@ class ReturnDuplicatorLow(StructuringOptimizationPass, ReturnDuplicatorBase):
                 # keep testing the next edge
                 node = succ
 
-            # Special case 3: In Phoenix, regions full of only if-stmts can be collapsed and moved. This causes
-            # the goto manager to report gotos that are at the top of the region instead of ones in the middle of it.
-            # Because of this, we need to gather all the nodes above the original src and check if any of them
-            # go to the destination. Additionally, we need to do this on the supergraph to get rid of
-            # goto edges that are removed by Phoenix.
-            # This case is observed in the test case `TestDecompiler.test_tail_tail_bytes_ret_dup`.
-            if self._supergraph is None:
-                return False
-
-            super_to_og_nodes = {n: self._supergraph.nodes[n]["original_nodes"] for n in self._supergraph.nodes}
-            og_to_super_nodes = {og: super_n for super_n, ogs in super_to_og_nodes.items() for og in ogs}
-            super_src = og_to_super_nodes.get(src)
-            super_dst = og_to_super_nodes.get(dst)
-            if super_src is None or super_dst is None:
-                return False
-
-            # collect all nodes which have only an if-stmt in them that are ancestors of super_src
-            check_blks = {super_src}
-            level_blocks = {super_src}
-            for _ in range(10):
-                done = False
-                if_blks = set()
-                for lblock in level_blocks:
-                    preds = list(self._supergraph.predecessors(lblock))
-                    for pred in preds:
-                        only_cond_jump = all(isinstance(s, (ConditionalJump, Label)) for s in pred.statements)
-                        if only_cond_jump:
-                            if_blks.add(pred)
-
-                    done = len(if_blks) == 0
-
-                if done:
-                    break
-
-                check_blks |= if_blks
-                level_blocks = if_blks
-
-            # convert all the found if-only super-blocks back into their original blocks
-            og_check_blocks = set()
-            for blk in check_blks:
-                og_check_blocks |= set(super_to_og_nodes[blk])
-
-            # check if any of the original blocks are gotos to the destination
-            goto_hits = 0
-            for block in og_check_blocks:
-                if self._goto_manager.is_goto_edge(block, dst):
-                    goto_hits += 1
-
-            # Although it is good to find a goto in the if-only block region, having more than a single goto
-            # existing that goes to the same dst is a bad sign. This can be seen in the the following test:
-            # TestDecompiler.test_dd_iread_ret_dup_region
-            #
-            # It occurs when you have something like:
-            # ```
-            # if (a || c)
-            #     goto target;
-            # target:
-            # return 0;
-            # ```
-            #
-            #
-            # This looks like an edge from (a, target) and (c, target) but it is actually a single edge.
-            # If you allow both to duplicate you get the following:
-            # ```
-            # if (a):
-            #    return
-            # if (c):
-            #    return
-            # ```
-            # This is not the desired behavior.
-            # So we need to check if there is only a single goto that goes to the destination.
-            return goto_hits == 1
-
         return False
 
     def _analyze(self, cache=None):

{angr-9.2.166 → angr-9.2.167}/angr/analyses/decompiler/region_identifier.py

@@ -43,6 +43,7 @@ class RegionIdentifier(Analysis):
         update_graph=True,
         largest_successor_tree_outside_loop=True,
         force_loop_single_exit=True,
+        refine_loops_with_single_successor=False,
         complete_successors=False,
         entry_node_addr: tuple[int, int | None] | None = None,
     ):
@@ -70,6 +71,7 @@ class RegionIdentifier(Analysis):
         self.regions_by_block_addrs = []
         self._largest_successor_tree_outside_loop = largest_successor_tree_outside_loop
         self._force_loop_single_exit = force_loop_single_exit
+        self._refine_loops_with_single_successor = refine_loops_with_single_successor
         self._complete_successors = complete_successors
         # we keep a dictionary of node and their traversal order in a quasi-topological traversal and update this
         # dictionary as we update the graph
@@ -265,13 +267,18 @@ class RegionIdentifier(Analysis):
 
         # special case: any node with more than two non-self successors are probably the head of a switch-case. we
         # should include all successors into the loop subgraph.
+        # we must be extra careful here to not include nodes that are reachable from outside the loop subgraph. an
+        # example is in binary 064e1d62c8542d658d83f7e231cc3b935a1f18153b8aea809dcccfd446a91c93, loop 0x40d7b0 should
+        # not include block 0x40d9d5 because this node has a out-of-loop-body predecessor (block 0x40d795).
         while True:
             updated = False
             for node in list(loop_subgraph):
                 nonself_successors = [succ for succ in graph.successors(node) if succ is not node]
                 if len(nonself_successors) > 2:
                     for succ in nonself_successors:
-                        if not loop_subgraph.has_edge(node, succ):
+                        if not loop_subgraph.has_edge(node, succ) and all(
+                            pred in loop_subgraph for pred in graph.predecessors(succ)
+                        ):
                             updated = True
                             loop_subgraph.add_edge(node, succ)
             if not updated:
@@ -280,7 +287,9 @@ class RegionIdentifier(Analysis):
         return set(loop_subgraph)
 
     def _refine_loop(self, graph: networkx.DiGraph, head, initial_loop_nodes, initial_exit_nodes):
-        if len(initial_exit_nodes) <= 1:
+        if (self._refine_loops_with_single_successor and len(initial_exit_nodes) == 0) or (
+            not self._refine_loops_with_single_successor and len(initial_exit_nodes) <= 1
+        ):
             return initial_loop_nodes, initial_exit_nodes
 
         refined_loop_nodes = initial_loop_nodes.copy()
@@ -713,7 +722,7 @@ class RegionIdentifier(Analysis):
 
         # visit the nodes in post-order
         region_created = False
-        for node in list(networkx.dfs_postorder_nodes(graph_copy, source=head)):
+        for node in list(GraphUtils.dfs_postorder_nodes_deterministic(graph_copy, head)):
             if node is dummy_endnode:
                 # skip the dummy endnode
                 continue

{angr-9.2.166 → angr-9.2.167}/angr/analyses/decompiler/sequence_walker.py

@@ -110,24 +110,28 @@ class SequenceWalker:
 
     def _handle_MultiNode(self, node, **kwargs):
         changed = False
-        nodes_copy = list(node.nodes)
+        nodes = node.nodes if self._update_seqnode_in_place else list(node.nodes)
 
         if self._force_forward_scan:
-            for i, node_ in enumerate(nodes_copy):
+            for i, node_ in enumerate(nodes):
                 new_node = self._handle(node_, parent=node, index=i)
                 if new_node is not None:
                     changed = True
-                    node.nodes[i] = new_node
+                    nodes[i] = new_node
         else:
-            i = len(nodes_copy) - 1
+            i = len(nodes) - 1
             while i > -1:
-                node_ = nodes_copy[i]
+                node_ = nodes[i]
                 new_node = self._handle(node_, parent=node, index=i)
                 if new_node is not None:
                     changed = True
-                    node.nodes[i] = new_node
+                    nodes[i] = new_node
                 i -= 1
-        return None if not changed else node
+        if not changed:
+            return None
+        if self._update_seqnode_in_place:
+            return node
+        return MultiNode(nodes, addr=node.addr, idx=node.idx)
 
     def _handle_SwitchCase(self, node, **kwargs):
         self._handle(node.switch_expr, parent=node, label="switch_expr")