[Restructuring] Resolve unresolved reaching conditions (#407)

* Create draft PR for #28

* start with issue

* some minor changes

* no endless recursion

* fix tests

* add test

* modify to always pair two

* fix format

* some refactoring and docstrings

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Eva-Maria Behner <[email protected]>

decompiler/pipeline/controlflowanalysis/restructuring_commons/acyclic_restructuring.py

@@ -94,7 +94,10 @@ def _construct_refined_ast(self, seq_node_root: SeqNode) -> AbstractSyntaxTreeNo
         ConditionBasedRefinement.refine(self.asforest)
         acyclic_processor.preprocess_condition_aware_refinement()
         if self.options.reconstruct_switch:
-            ConditionAwareRefinement.refine(self.asforest, self.options)
+            updated_switch_nodes = ConditionAwareRefinement.refine(self.asforest, self.options)
+            for switch_node in updated_switch_nodes:
+                for sequence_case in (c for c in switch_node.cases if isinstance(c.child, SeqNode)):
+                    ConditionBasedRefinement.refine(self.asforest, sequence_case.child)
         acyclic_processor.postprocess_condition_refinement()
         root = self.asforest.current_root
         self.asforest.remove_current_root()

decompiler/pipeline/controlflowanalysis/restructuring_commons/ast_processor.py

@@ -232,7 +232,7 @@ def _group_by_reaching_conditions(self, nodes: Tuple[AbstractSyntaxTreeNode]) ->
 
         :param nodes: The AST nodes that we want to group.
         :return: A dictionary that assigns to a reaching condition the list of AST code nodes with this reaching condition,
-                 if it are at least two with the same.
+                 if there are at least two with the same.
         """
         initial_groups: Dict[LogicCondition, List[AbstractSyntaxTreeNode]] = dict()
         for node in nodes:

decompiler/pipeline/controlflowanalysis/restructuring_commons/condition_aware_refinement.py

@@ -2,6 +2,8 @@
 Module for Condition Aware Refinement
 """
 
+from typing import Set
+
 from decompiler.pipeline.controlflowanalysis.restructuring_commons.condition_aware_refinement_commons.base_class_car import (
     BaseClassConditionAwareRefinement,
 )
@@ -21,6 +23,7 @@
     SwitchExtractor,
 )
 from decompiler.pipeline.controlflowanalysis.restructuring_options import RestructuringOptions
+from decompiler.structures.ast.ast_nodes import SwitchNode
 from decompiler.structures.ast.syntaxforest import AbstractSyntaxForest
 
 
@@ -35,13 +38,14 @@ class ConditionAwareRefinement(BaseClassConditionAwareRefinement):
     ]
 
     @classmethod
-    def refine(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions):
+    def refine(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions) -> Set[SwitchNode]:
         condition_aware_refinement = cls(asforest, options)
         for stage in condition_aware_refinement.REFINEMENT_PIPELINE:
             asforest.clean_up(asforest.current_root)
-            stage(asforest, options)
+            condition_aware_refinement.updated_switch_nodes.update(stage(asforest, options))
             condition_aware_refinement._remove_redundant_reaching_condition_from_switch_nodes()
         asforest.clean_up(asforest.current_root)
+        return set(switch for switch in condition_aware_refinement.updated_switch_nodes if switch in asforest)
 
     def _remove_redundant_reaching_condition_from_switch_nodes(self):
         """Remove the reaching condition from all switch nodes if it is redundant."""

decompiler/pipeline/controlflowanalysis/restructuring_commons/condition_aware_refinement_commons/base_class_car.py

@@ -1,5 +1,5 @@
 from dataclasses import dataclass
-from typing import Iterator, Optional, Tuple
+from typing import Iterator, Optional, Set, Tuple
 
 from decompiler.pipeline.controlflowanalysis.restructuring_options import LoopBreakOptions, RestructuringOptions
 from decompiler.structures.ast.ast_nodes import AbstractSyntaxTreeNode, CaseNode, FalseNode, SwitchNode, TrueNode
@@ -63,6 +63,7 @@ def __init__(self, asforest: AbstractSyntaxForest, options: RestructuringOptions
         self.asforest: AbstractSyntaxForest = asforest
         self.condition_handler: ConditionHandler = asforest.condition_handler
         self.options: RestructuringOptions = options
+        self.updated_switch_nodes: Set[SwitchNode] = set()
 
     def _get_constant_equality_check_expressions_and_conditions(
         self, condition: LogicCondition

decompiler/pipeline/controlflowanalysis/restructuring_commons/condition_aware_refinement_commons/initial_switch_node_constructer.py

@@ -214,13 +214,14 @@ class InitialSwitchNodeConstructor(BaseClassConditionAwareRefinement):
     """Class that constructs switch nodes."""
 
     @classmethod
-    def construct(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions):
+    def construct(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions) -> Set[SwitchNode]:
         """Constructs initial switch nodes if possible."""
         initial_switch_constructor = cls(asforest, options)
         for cond_node in asforest.get_condition_nodes_post_order(asforest.current_root):
             initial_switch_constructor._extract_case_nodes_from_nested_condition(cond_node)
         for seq_node in asforest.get_sequence_nodes_post_order(asforest.current_root):
             initial_switch_constructor._try_to_construct_initial_switch_node_for(seq_node)
+        return initial_switch_constructor.updated_switch_nodes
 
     def _extract_case_nodes_from_nested_condition(self, cond_node: ConditionNode) -> None:
         """
@@ -336,6 +337,7 @@ def _try_to_construct_initial_switch_node_for(self, seq_node: SeqNode) -> None:
             sibling_reachability = self.asforest.get_sibling_reachability_of_children_of(seq_node)
             switch_cases = list(possible_switch_node.construct_switch_cases())
             switch_node = self.asforest.create_switch_node_with(possible_switch_node.expression, switch_cases)
+            self.updated_switch_nodes.add(switch_node)
             case_dependency = CaseDependencyGraph.construct_case_dependency_for(self.asforest.children(switch_node), sibling_reachability)
             self._update_reaching_condition_for_case_node_children(switch_node)
             self._add_constants_to_cases(switch_node, case_dependency)
@@ -393,7 +395,7 @@ def _update_reaching_condition_for_case_node_children(self, switch_node: SwitchN
                 case_node.reaching_condition.is_disjunction_of_literals
             ), f"The condition of a case node should be a disjunction, but it is {case_node.reaching_condition}!"
 
-            if isinstance(cond_node := case_node.child, ConditionNode) and cond_node.false_branch is None:
+            if (cond_node := case_node.child).is_single_branch:
                 self._update_condition_for(cond_node, case_node)
 
             case_node.child.reaching_condition = case_node.child.reaching_condition.substitute_by_true(case_node.reaching_condition)

decompiler/pipeline/controlflowanalysis/restructuring_commons/condition_aware_refinement_commons/missing_case_finder.py

@@ -33,6 +33,7 @@ def _insert_case_node(self, new_case_node: AbstractSyntaxTreeNode, case_constant
         if default_case := switch_node.default:
             new_children.append(default_case)
         switch_node._sorted_cases = tuple(new_children)
+        self.updated_switch_nodes.add(switch_node)
 
     def _new_case_nodes_for(
         self, new_case_node: AbstractSyntaxTreeNode, switch_node: SwitchNode, sorted_case_constants: List[Constant]

decompiler/pipeline/controlflowanalysis/restructuring_commons/condition_aware_refinement_commons/missing_case_finder_condition.py

@@ -28,7 +28,7 @@ class MissingCaseFinderCondition(MissingCaseFinder):
     """
 
     @classmethod
-    def find(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions):
+    def find(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions) -> Set[SwitchNode]:
         """Try to find missing cases that are branches of condition nodes."""
         missing_case_finder = cls(asforest, options)
         for condition_node in asforest.get_condition_nodes_post_order(asforest.current_root):
@@ -40,6 +40,7 @@ def find(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions):
                     asforest.extract_switch_from_sequence(case_candidate_information.switch_node)
                 else:
                     asforest.replace_condition_node_by_single_branch(condition_node)
+        return missing_case_finder.updated_switch_nodes
 
     def _can_insert_missing_case_node(self, condition_node: ConditionNode) -> Optional[CaseCandidateInformation]:
         """

decompiler/pipeline/controlflowanalysis/restructuring_commons/condition_aware_refinement_commons/missing_case_finder_intersecting_constants.py

@@ -47,7 +47,7 @@ def insert(self, possible_case: CaseNodeCandidate):
           first fallthrough-cases.
         - If the possible-case node is reached by the switch-node, then the content must be after any other code.
           Thus, it must contain all constants from a block of fallthrough-cases. But here, it can contain more.
-        - If neither one reaches the other, then it can be insert anywhere, at long as it can be archived by only
+        - If neither one reaches the other, then it can be inserted anywhere, as long as it can be archived by only
           resorting fallthrough-cases all leading to the same code-execution.
         """
         cases_of_switch_node = {case.constant for case in self._switch_node.children}
@@ -70,6 +70,7 @@ def insert(self, possible_case: CaseNodeCandidate):
                 return
 
         self._sibling_reachability_graph.update_when_inserting_new_case_node(compare_node, self._switch_node)
+        self.updated_switch_nodes.add(self._switch_node)
         compare_node.clean()
 
     def _add_case_before(self, intersecting_linear_case: Tuple[CaseNode], possible_case_properties: IntersectingCaseNodeProperties) -> bool:

decompiler/pipeline/controlflowanalysis/restructuring_commons/condition_aware_refinement_commons/missing_case_finder_sequence.py

@@ -38,7 +38,7 @@ def __init__(self, asforest: AbstractSyntaxForest, options: RestructuringOptions
         self._switch_node_of_expression: Dict[ExpressionUsages, SwitchNode] = dict()
 
     @classmethod
-    def find(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions):
+    def find(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions) -> Set[SwitchNode]:
         """
         Try to find missing cases that are children of sequence nodes.
 
@@ -58,6 +58,7 @@ def find(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions):
 
             if seq_node in asforest:
                 seq_node.clean()
+        return missing_case_finder.updated_switch_nodes
 
     def _initialize_switch_node_of_expression_dictionary(self) -> None:
         """

decompiler/pipeline/controlflowanalysis/restructuring_commons/condition_aware_refinement_commons/switch_extractor.py

@@ -1,4 +1,4 @@
-from typing import Optional, Union
+from typing import Optional, Set, Union
 
 from decompiler.pipeline.controlflowanalysis.restructuring_commons.condition_aware_refinement_commons.base_class_car import (
     BaseClassConditionAwareRefinement,
@@ -19,6 +19,7 @@ def extract(cls, asforest: AbstractSyntaxForest, options: RestructuringOptions):
         for switch_node in list(asforest.get_switch_nodes_post_order(asforest.current_root)):
             while switch_extractor._successfully_extracts_switch_nodes(switch_node):
                 pass
+        return switch_extractor.updated_switch_nodes
 
     def _successfully_extracts_switch_nodes(self, switch_node: SwitchNode) -> bool:
         """