angr 9.2.166__cp310-abi3-manylinux_2_28_aarch64.whl

angr/analyses/typehoon/simple_solver.py

@@ -0,0 +1,1666 @@
+# pylint:disable=missing-class-docstring,too-many-boolean-expressions
+from __future__ import annotations
+import enum
+from collections import defaultdict
+from contextlib import suppress
+import logging
+
+import networkx
+from sortedcontainers import SortedDict
+
+from angr.utils.constants import MAX_POINTSTO_BITS
+from .typevars import (
+    Existence,
+    Subtype,
+    Equivalence,
+    Add,
+    TypeVariable,
+    DerivedTypeVariable,
+    HasField,
+    IsArray,
+    TypeConstraint,
+    Load,
+    Store,
+    BaseLabel,
+    FuncIn,
+    FuncOut,
+    ConvertTo,
+    new_dtv,
+)
+from .typeconsts import (
+    BottomType,
+    TopType,
+    TypeConstant,
+    Int,
+    Int8,
+    Int16,
+    Int32,
+    Int64,
+    Pointer,
+    Pointer32,
+    Pointer64,
+    Struct,
+    Array,
+    Function,
+    int_type,
+    Float,
+    Float32,
+    Float64,
+)
+from .variance import Variance
+from .dfa import DFAConstraintSolver, EmptyEpsilonNFAError
+
+_l = logging.getLogger(__name__)
+
+
+Top_ = TopType()
+Int_ = Int()
+Int64_ = Int64()
+Int32_ = Int32()
+Int16_ = Int16()
+Int8_ = Int8()
+Bottom_ = BottomType()
+Pointer64_ = Pointer64()
+Pointer32_ = Pointer32()
+Struct_ = Struct()
+Array_ = Array()
+Float_ = Float()
+Float32_ = Float32()
+Float64_ = Float64()
+
+
+PRIMITIVE_TYPES = {
+    Top_,
+    Int_,
+    Int8_,
+    Int16_,
+    Int32_,
+    Int64_,
+    Pointer32_,
+    Pointer64_,
+    Bottom_,
+    Struct_,
+    Array_,
+    Float_,
+    Float32_,
+    Float64_,
+}
+
+
+# lattice for 64-bit binaries
+BASE_LATTICE_64 = networkx.DiGraph()
+BASE_LATTICE_64.add_edge(Top_, Int_)
+BASE_LATTICE_64.add_edge(Int_, Int64_)
+BASE_LATTICE_64.add_edge(Int_, Int32_)
+BASE_LATTICE_64.add_edge(Int_, Int16_)
+BASE_LATTICE_64.add_edge(Int_, Int8_)
+BASE_LATTICE_64.add_edge(Int32_, Bottom_)
+BASE_LATTICE_64.add_edge(Int16_, Bottom_)
+BASE_LATTICE_64.add_edge(Int8_, Bottom_)
+BASE_LATTICE_64.add_edge(Int64_, Pointer64_)
+BASE_LATTICE_64.add_edge(Pointer64_, Bottom_)
+
+# lattice for 32-bit binaries
+BASE_LATTICE_32 = networkx.DiGraph()
+BASE_LATTICE_32.add_edge(Top_, Int_)
+BASE_LATTICE_32.add_edge(Int_, Int64_)
+BASE_LATTICE_32.add_edge(Int_, Int32_)
+BASE_LATTICE_32.add_edge(Int_, Int16_)
+BASE_LATTICE_32.add_edge(Int_, Int8_)
+BASE_LATTICE_32.add_edge(Int32_, Pointer32_)
+BASE_LATTICE_32.add_edge(Int64_, Bottom_)
+BASE_LATTICE_32.add_edge(Pointer32_, Bottom_)
+BASE_LATTICE_32.add_edge(Int16_, Bottom_)
+BASE_LATTICE_32.add_edge(Int8_, Bottom_)
+
+BASE_LATTICES = {
+    32: BASE_LATTICE_32,
+    64: BASE_LATTICE_64,
+}
+
+
+#
+# Sketch
+#
+
+
+class SketchNodeBase:
+    """
+    The base class for nodes in a sketch.
+    """
+
+    __slots__ = ()
+
+
+class SketchNode(SketchNodeBase):
+    """
+    Represents a node in a sketch graph.
+    """
+
+    __slots__ = ("lower_bound", "typevar", "upper_bound")
+
+    def __init__(self, typevar: TypeVariable | DerivedTypeVariable):
+        self.typevar: TypeVariable | DerivedTypeVariable = typevar
+        self.upper_bound: TypeConstant = TopType()
+        self.lower_bound: TypeConstant = BottomType()
+
+    def __repr__(self):
+        return f"{self.lower_bound} <: {self.typevar} <: {self.upper_bound}"
+
+    def __eq__(self, other):
+        return isinstance(other, SketchNode) and self.typevar == other.typevar
+
+    def __hash__(self):
+        return hash((SketchNode, self.typevar))
+
+    @property
+    def size(self) -> int | None:
+        """
+        Best-effort estimation of the size of the typevar (in bits). Returns None if we cannot determine.
+        """
+
+        if isinstance(self.typevar, DerivedTypeVariable):
+            last_label = self.typevar.labels[-1]
+            if isinstance(last_label, HasField) and last_label.bits != MAX_POINTSTO_BITS:
+                return last_label.bits
+        if isinstance(self.lower_bound, TypeConstant) and not isinstance(self.lower_bound, (TopType, BottomType)):
+            with suppress(NotImplementedError):
+                return self.lower_bound.size * 8
+        if isinstance(self.upper_bound, TypeConstant) and not isinstance(self.upper_bound, (TopType, BottomType)):
+            with suppress(NotImplementedError):
+                return self.upper_bound.size * 8
+        return None
+
+
+class RecursiveRefNode(SketchNodeBase):
+    """
+    Represents a cycle in a sketch graph.
+
+    This is equivalent to sketches.LabelNode in the reference implementation of retypd.
+    """
+
+    def __init__(self, target: DerivedTypeVariable):
+        self.target: DerivedTypeVariable = target
+
+    def __hash__(self):
+        return hash((RecursiveRefNode, self.target))
+
+    def __eq__(self, other):
+        return type(other) is RecursiveRefNode and other.target == self.target
+
+
+class Sketch:
+    """
+    Describes the sketch of a type variable.
+    """
+
+    __slots__ = (
+        "graph",
+        "node_mapping",
+        "root",
+        "solver",
+    )
+
+    def __init__(self, solver: SimpleSolver, root: TypeVariable):
+        self.root: SketchNode = SketchNode(root)
+        self.graph = networkx.MultiDiGraph()
+        self.node_mapping: dict[TypeVariable | DerivedTypeVariable, SketchNodeBase] = {}
+        self.solver = solver
+
+        # add the root node
+        self.graph.add_node(self.root)
+        self.node_mapping[root] = self.root
+
+    def lookup(self, typevar: TypeVariable | DerivedTypeVariable) -> SketchNodeBase | None:
+        if typevar in self.node_mapping:
+            return self.node_mapping[typevar]
+        node: SketchNodeBase | None = None
+        if isinstance(typevar, DerivedTypeVariable):
+            t = SimpleSolver._to_typevar_or_typeconst(typevar.type_var)
+            assert isinstance(t, TypeVariable)
+            node = self.node_mapping[t]
+            for label in typevar.labels:
+                succs = []
+                for _, dst, data in self.graph.out_edges(node, data=True):
+                    if "label" in data and data["label"] == label and dst not in succs:
+                        succs.append(dst)
+                if len(succs) > 1:
+                    _l.warning(
+                        "Multiple successors found for node %s with label %s. Picking the first one.", node, label
+                    )
+                    succs = succs[:1]
+                if not succs:
+                    return None
+                node = succs[0]
+                if isinstance(node, RecursiveRefNode):
+                    node = self.lookup(node.target)
+        return node
+
+    def add_edge(self, src: SketchNodeBase, dst: SketchNodeBase, label) -> None:
+        # ensure the label does not already exist in existing edges
+        if self.graph.has_edge(src, dst):
+            for data in self.graph.get_edge_data(src, dst).values():
+                if "label" in data and data["label"] == label:
+                    return
+        if isinstance(src, SketchNode) and src.typevar not in self.node_mapping:
+            self.node_mapping[src.typevar] = src
+        if isinstance(dst, SketchNode) and dst.typevar not in self.node_mapping:
+            self.node_mapping[dst.typevar] = dst
+        self.graph.add_edge(src, dst, label=label)
+
+    def add_constraint(self, constraint: TypeConstraint) -> None:
+        # sub <: super
+        if not isinstance(constraint, Subtype):
+            return
+        subtype, _ = self.flatten_typevar(constraint.sub_type)
+        supertype, try_maxsize = self.flatten_typevar(constraint.super_type)
+
+        if (
+            try_maxsize
+            and isinstance(subtype, TypeVariable)
+            and subtype in self.solver.stackvar_max_sizes
+            and isinstance(supertype, TypeConstant)
+            and not isinstance(supertype, BottomType)
+        ):
+            basetype = supertype
+            assert basetype.size is not None
+            max_size = self.solver.stackvar_max_sizes.get(subtype, None)
+            if max_size not in {0, None} and max_size // basetype.size > 0:  # type: ignore
+                supertype = Array(element=basetype, count=max_size // basetype.size)  # type: ignore
+
+        if SimpleSolver._typevar_inside_set(subtype, PRIMITIVE_TYPES) and not SimpleSolver._typevar_inside_set(
+            supertype, PRIMITIVE_TYPES
+        ):
+            assert isinstance(supertype, (TypeVariable, DerivedTypeVariable))
+            super_node = self.lookup(supertype)
+            assert super_node is None or isinstance(super_node, SketchNode)
+            if super_node is not None:
+                super_node.lower_bound = self.solver.join(super_node.lower_bound, subtype)
+        elif SimpleSolver._typevar_inside_set(supertype, PRIMITIVE_TYPES) and not SimpleSolver._typevar_inside_set(
+            subtype, PRIMITIVE_TYPES
+        ):
+            assert isinstance(subtype, (TypeVariable, DerivedTypeVariable))
+            sub_node = self.lookup(subtype)
+            assert sub_node is None or isinstance(sub_node, SketchNode)
+            # assert sub_node is not None
+            if sub_node is not None:
+                sub_node.upper_bound = self.solver.meet(sub_node.upper_bound, supertype)
+
+    @staticmethod
+    def flatten_typevar(
+        derived_typevar: TypeVariable | TypeConstant | DerivedTypeVariable,
+    ) -> tuple[DerivedTypeVariable | TypeVariable | TypeConstant, bool]:
+        # pylint:disable=too-many-boolean-expressions
+        if (
+            isinstance(derived_typevar, DerivedTypeVariable)
+            and isinstance(derived_typevar.type_var, Pointer)
+            and SimpleSolver._typevar_inside_set(derived_typevar.type_var.basetype, PRIMITIVE_TYPES)
+            and len(derived_typevar.labels) == 2
+            and isinstance(derived_typevar.labels[0], Load)
+            and isinstance(derived_typevar.labels[1], HasField)
+            and derived_typevar.labels[1].offset == 0
+            and derived_typevar.labels[1].bits == MAX_POINTSTO_BITS
+        ):
+            bt = derived_typevar.type_var.basetype
+            assert bt is not None
+            return bt, True
+        return derived_typevar, False
+
+
+#
+# Constraint graph
+#
+
+
+class ConstraintGraphTag(enum.Enum):
+    LEFT = 0
+    RIGHT = 1
+    UNKNOWN = 2
+
+
+class FORGOTTEN(enum.Enum):
+    PRE_FORGOTTEN = 0
+    POST_FORGOTTEN = 1
+
+
+class ConstraintGraphNode:
+    __slots__ = ("forgotten", "tag", "typevar", "variance")
+
+    def __init__(
+        self,
+        typevar: TypeVariable | DerivedTypeVariable,
+        variance: Variance,
+        tag: ConstraintGraphTag,
+        forgotten: FORGOTTEN,
+    ):
+        self.typevar = typevar
+        self.variance = variance
+        self.tag = tag
+        self.forgotten = forgotten
+
+    def __repr__(self):
+        variance_str = "CO" if self.variance == Variance.COVARIANT else "CONTRA"
+        if self.tag == ConstraintGraphTag.LEFT:
+            tag_str = "L"
+        elif self.tag == ConstraintGraphTag.RIGHT:
+            tag_str = "R"
+        else:
+            tag_str = "U"
+        forgotten_str = "PRE" if self.forgotten == FORGOTTEN.PRE_FORGOTTEN else "POST"
+        s = f"{self.typevar}#{variance_str}.{tag_str}.{forgotten_str}"
+        if ":" in s:
+            return '"' + s + '"'
+        return s
+
+    def __eq__(self, other):
+        if not isinstance(other, ConstraintGraphNode):
+            return False
+        return (
+            self.typevar == other.typevar
+            and self.variance == other.variance
+            and self.tag == other.tag
+            and self.forgotten == other.forgotten
+        )
+
+    def __hash__(self):
+        return hash((ConstraintGraphNode, self.typevar, self.variance, self.tag, self.forgotten))
+
+    def forget_last_label(self) -> tuple[ConstraintGraphNode, BaseLabel] | None:
+        if isinstance(self.typevar, DerivedTypeVariable) and self.typevar.labels:
+            last_label = self.typevar.labels[-1]
+            if len(self.typevar.labels) == 1:
+                prefix = self.typevar.type_var
+            else:
+                prefix = DerivedTypeVariable(self.typevar.type_var, None, labels=self.typevar.labels[:-1])
+            variance = Variance.COVARIANT if self.variance == last_label.variance else Variance.CONTRAVARIANT
+            if not isinstance(prefix, (TypeVariable, DerivedTypeVariable)):
+                # we may see incorrectly generated type constraints that attempt to load from an int:
+                #   int64.load
+                # we don't want to entertain such constraints
+                return None
+            return (
+                ConstraintGraphNode(prefix, variance, self.tag, FORGOTTEN.PRE_FORGOTTEN),
+                self.typevar.labels[-1],
+            )
+        return None
+
+    def recall(self, label: BaseLabel) -> ConstraintGraphNode:
+        if isinstance(self.typevar, DerivedTypeVariable):
+            labels = (*self.typevar.labels, label)
+            typevar = self.typevar.type_var
+        elif isinstance(self.typevar, (TypeVariable, TypeConstant)):
+            labels = (label,)
+            typevar = self.typevar
+        else:
+            raise TypeError(f"Unsupported type {type(self.typevar)}")
+        variance = Variance.COVARIANT if self.variance == label.variance else Variance.CONTRAVARIANT
+        var = typevar if not labels else DerivedTypeVariable(typevar, None, labels=labels)
+        assert isinstance(var, (TypeVariable, DerivedTypeVariable))
+        return ConstraintGraphNode(var, variance, self.tag, FORGOTTEN.PRE_FORGOTTEN)
+
+    def inverse(self) -> ConstraintGraphNode:
+        if self.tag == ConstraintGraphTag.LEFT:
+            tag = ConstraintGraphTag.RIGHT
+        elif self.tag == ConstraintGraphTag.RIGHT:
+            tag = ConstraintGraphTag.LEFT
+        else:
+            tag = ConstraintGraphTag.UNKNOWN
+
+        variance = Variance.CONTRAVARIANT if self.variance == Variance.COVARIANT else Variance.COVARIANT
+
+        return ConstraintGraphNode(self.typevar, variance, tag, self.forgotten)
+
+    def inverse_wo_tag(self) -> ConstraintGraphNode:
+        """
+        Invert the variance only.
+        """
+        variance = Variance.CONTRAVARIANT if self.variance == Variance.COVARIANT else Variance.COVARIANT
+
+        return ConstraintGraphNode(self.typevar, variance, self.tag, self.forgotten)
+
+
+#
+# The solver
+#
+
+
+class SimpleSolver:
+    """
+    SimpleSolver is, by its name, a simple solver. Most of this solver is based on the (complex) simplification logic
+    that the retypd paper describes and the retypd re-implementation (https://github.com/GrammaTech/retypd) implements.
+    Additionally, we add some improvements to allow type propagation of known struct names, among a few other
+    improvements.
+    """
+
+    def __init__(
+        self,
+        bits: int,
+        constraints,
+        typevars,
+        constraint_set_degradation_threshold: int = 150,
+        stackvar_max_sizes: dict[TypeVariable, int] | None = None,
+    ):
+        if bits not in (32, 64):
+            raise ValueError(f"Pointer size {bits} is not supported. Expect 32 or 64.")
+
+        self.bits = bits
+        self._constraints: dict[TypeVariable, set[TypeConstraint]] = constraints
+        self._typevars: set[TypeVariable] = typevars
+        self.stackvar_max_sizes = stackvar_max_sizes if stackvar_max_sizes is not None else {}
+        self._constraint_set_degradation_threshold = constraint_set_degradation_threshold
+        self._base_lattice = BASE_LATTICES[bits]
+        self._base_lattice_inverted = networkx.DiGraph()
+        for src, dst in self._base_lattice.edges:
+            self._base_lattice_inverted.add_edge(dst, src)
+
+        # statistics
+        self.processed_constraints_count: int = 0
+        self.simplified_constraints_count: int = 0
+        self.eqclass_constraints_count: list[int] = []
+
+        #
+        # Solving state
+        #
+        self._equivalence = defaultdict(dict)
+        for typevar in list(self._constraints):
+            if self._constraints[typevar]:
+                self.processed_constraints_count += len(self._constraints[typevar])
+
+                self._constraints[typevar] |= self._eq_constraints_from_add(typevar)
+                self._constraints[typevar] |= self._discover_equivalence(self._constraints[typevar])
+                new_constraints, replacements = self._handle_equivalence(self._constraints[typevar])
+                self._equivalence |= replacements
+                self._constraints[typevar] = new_constraints
+                self._constraints[typevar] = self._filter_constraints(self._constraints[typevar])
+
+                self.simplified_constraints_count += len(self._constraints[typevar])
+
+        self.solution = {}
+        for tv, sol in self._equivalence.items():
+            if isinstance(tv, TypeVariable) and isinstance(sol, TypeConstant):
+                self.solution[tv] = sol
+
+        self._solution_cache = {}
+        self.solve()
+        for typevar in list(self._constraints):
+            self._convert_arrays(self._constraints[typevar])
+
+    def solve(self):
+        """
+        Steps:
+
+        For each type variable,
+        - Infer the shape in its sketch
+        - Build the constraint graph
+        - Collect all constraints
+        - Apply constraints to derive the lower and upper bounds
+        - Determine a solution for type variables with constraints
+        - Rewrite the constraint graph by replacing determined type variables with their solutions
+        - Solve repeatedly until all interesting type variables have solutions
+
+        By repeatedly solving until exhausting interesting type variables, we ensure the S-Trans rule is applied.
+        """
+
+        prem_typevars = set(self._constraints) | self._typevars
+        typevars = set()
+        for tv in prem_typevars:
+            if tv not in self._equivalence:
+                typevars.add(tv)
+            else:
+                repl = self._equivalence[tv]
+                if isinstance(repl, TypeVariable):
+                    typevars.add(repl)
+
+        constraints = set()
+        for tv in typevars:
+            if tv in self._constraints:
+                constraints |= self._constraints[tv]
+
+        # collect typevars used in the constraint set
+        constrained_typevars = set()
+        for constraint in constraints:
+            if isinstance(constraint, Subtype):
+                for t in (constraint.sub_type, constraint.super_type):
+                    if isinstance(t, DerivedTypeVariable):
+                        if t.type_var in typevars:
+                            constrained_typevars.add(t.type_var)
+                    elif isinstance(t, TypeVariable) and t in typevars:
+                        constrained_typevars.add(t)
+
+        _, sketches = self.infer_shapes(typevars, constraints)
+        constraintset2tvs = defaultdict(set)
+        tvs_seen = set()
+        for idx, tv in enumerate(constrained_typevars):
+            _l.debug("Collecting constraints for type variable %r (%d/%d)", tv, idx + 1, len(constrained_typevars))
+            if tv in tvs_seen:
+                continue
+            # build a sub constraint set for the type variable
+            constraint_subset, related_tvs = self._generate_constraint_subset(constraints, {tv})
+            # drop all type vars outside constrained_typevars
+            related_tvs = related_tvs.intersection(constrained_typevars)
+            tvs_seen |= related_tvs
+            frozen_constraint_subset = frozenset(constraint_subset)
+            constraintset2tvs[frozen_constraint_subset] = related_tvs
+
+        for idx, (constraint_subset, tvs) in enumerate(constraintset2tvs.items()):
+            _l.debug(
+                "Solving %d constraints for %d type variables %r (%d/%d)",
+                len(constraint_subset),
+                len(tvs),
+                tvs,
+                idx + 1,
+                len(constraintset2tvs),
+            )
+            self.eqclass_constraints_count.append(len(constraint_subset))
+
+            if len(constraint_subset) > self._constraint_set_degradation_threshold:
+                _l.debug(
+                    "Constraint subset contains %d constraints, which is over the limit of %d. Enter degradation.",
+                    len(constraint_subset),
+                    self._constraint_set_degradation_threshold,
+                )
+                constraint_subset = self._degrade_constraint_set(constraint_subset)
+                _l.debug("Degraded constraint subset to %d constraints.", len(constraint_subset))
+
+            while constraint_subset:
+
+                _l.debug("Working with %d constraints.", len(constraint_subset))
+
+                # remove constraints that are a <: b where a only appears once; in this case, the solution fo a is
+                # entirely determined by the solution of b (which is the upper bound of a)
+                filtered_constraint_subset, ub_subtypes = self._filter_leaf_typevars(constraint_subset, tvs)
+                _l.debug(
+                    "Filtered %d leaf typevars; %d constraints remain.",
+                    len(ub_subtypes),
+                    len(filtered_constraint_subset),
+                )
+
+                base_constraint_graph = self._generate_constraint_graph(
+                    filtered_constraint_subset, tvs | PRIMITIVE_TYPES
+                )
+                primitive_constraints = self._generate_primitive_constraints(tvs, base_constraint_graph)
+                tvs_with_primitive_constraints = set()
+                for primitive_constraint in primitive_constraints:
+                    tv = self._typevar_from_primitive_constraint(primitive_constraint)
+                    tvs_with_primitive_constraints.add(tv)
+                    assert tv is not None, f"Cannot find type variable in primitive constraint {primitive_constraint}"
+                    sketches[tv].add_constraint(primitive_constraint)
+                solutions = {}
+                self.determine(sketches, tvs_with_primitive_constraints, solutions)
+                _l.debug("Determined solutions for %d type variable(s).", len(tvs_with_primitive_constraints))
+
+                leaf_solutions = 0
+                for tv_, ub_tv in ub_subtypes.items():
+                    if ub_tv in solutions:
+                        solutions[tv_] = solutions[ub_tv]
+                        leaf_solutions += 1
+                    elif isinstance(ub_tv, TypeConstant):
+                        solutions[tv_] = ub_tv
+                        leaf_solutions += 1
+                _l.debug("Determined solutions for %d leaf type variable(s).", leaf_solutions)
+
+                if not solutions:
+                    break
+                self.solution |= solutions
+
+                tvs = {tv for tv in tvs if tv not in solutions}
+                if not tvs:
+                    break
+                # rewrite existing constraints
+                new_constraint_subset = set()
+                for constraint in constraint_subset:
+                    rewritten = self._rewrite_constraint(constraint, solutions)
+                    new_constraint_subset.add(rewritten)
+                constraint_subset = self._filter_constraints(new_constraint_subset)
+
+        # set the solution for missing type vars to TOP
+        self.determine(sketches, set(sketches).difference(set(self.solution)), self.solution)
+
+    def infer_shapes(
+        self, typevars: set[TypeVariable], constraints: set[TypeConstraint]
+    ) -> tuple[dict, dict[TypeVariable, Sketch]]:
+        """
+        Computing sketches from constraint sets. Implements Algorithm E.1 in the retypd paper.
+        """
+
+        equivalence_classes, quotient_graph = self.compute_quotient_graph(constraints)
+
+        sketches: dict[TypeVariable, Sketch] = {}
+        for tv in typevars:
+            sketches[tv] = Sketch(self, tv)
+
+        for tv, sketch in sketches.items():
+            sketch_node = sketch.lookup(tv)
+            graph_node = equivalence_classes.get(tv, None)
+            # assert graph_node is not None
+            if graph_node is None:
+                continue
+            visited = {graph_node: sketch_node}
+            self._get_all_paths(quotient_graph, sketch, graph_node, visited)
+        return equivalence_classes, sketches
+
+    def compute_quotient_graph(self, constraints: set[TypeConstraint]):
+        """
+        Compute the quotient graph (the constraint graph modulo ~ in Algorithm E.1 in the retypd paper) with respect to
+        a given set of type constraints.
+        """
+
+        g = networkx.DiGraph()
+        # collect all derived type variables
+        typevars = self._typevars_from_constraints(constraints)
+        g.add_nodes_from(typevars)
+        # add paths for each derived type variable into the graph
+        for tv in typevars:
+            last_node = tv
+            prefix = tv
+            while isinstance(prefix, DerivedTypeVariable) and prefix.labels:
+                prefix = prefix.longest_prefix()
+                if prefix is None:
+                    continue
+                g.add_edge(prefix, last_node, label=last_node.labels[-1])
+                last_node = prefix
+
+        # compute the constraint graph modulo ~
+        equivalence_classes = {node: node for node in g}
+
+        load = Load()
+        store = Store()
+        for node in g.nodes:
+            lbl_to_node = {}
+            for succ in g.successors(node):
+                lbl_to_node[succ.labels[-1]] = succ
+                if load in lbl_to_node and store in lbl_to_node:
+                    self._unify(equivalence_classes, lbl_to_node[load], lbl_to_node[store], g)
+
+        for constraint in constraints:
+            if isinstance(constraint, Subtype):
+                if self._typevar_inside_set(constraint.super_type, PRIMITIVE_TYPES) or self._typevar_inside_set(
+                    constraint.sub_type, PRIMITIVE_TYPES
+                ):
+                    continue
+                self._unify(equivalence_classes, constraint.super_type, constraint.sub_type, g)
+
+        out_graph = networkx.MultiDiGraph()  # there can be multiple edges between two nodes, each edge is associated
+        # with a different label
+        for src, dst, data in g.edges(data=True):
+            src_cls = equivalence_classes[src]
+            dst_cls = equivalence_classes[dst]
+            label = None if not data else data["label"]
+            if label is not None and out_graph.has_edge(src_cls, dst_cls):
+                # do not add the same edge twice
+                existing_labels = {
+                    data_["label"]
+                    for _, dst_cls_, data_ in out_graph.out_edges(src_cls, data=True)
+                    if dst_cls_ == dst_cls and data
+                }
+                if label in existing_labels:
+                    continue
+            out_graph.add_edge(src_cls, dst_cls, label=label)
+
+        return equivalence_classes, out_graph
+
+    def _generate_primitive_constraints(
+        self,
+        non_primitive_endpoints: set[TypeVariable | DerivedTypeVariable],
+        constraint_graph,
+    ) -> set[TypeConstraint]:
+        constraints_0 = self._solve_constraints_between(constraint_graph, non_primitive_endpoints, PRIMITIVE_TYPES)
+        constraints_1 = self._solve_constraints_between(constraint_graph, PRIMITIVE_TYPES, non_primitive_endpoints)
+        return constraints_0 | constraints_1
+
+    @staticmethod
+    def _typevars_from_constraints(constraints: set[TypeConstraint]) -> set[TypeVariable | DerivedTypeVariable]:
+        """
+        Collect derived type variables from a set of constraints.
+        """
+
+        typevars: set[TypeVariable | DerivedTypeVariable] = set()
+        for constraint in constraints:
+            if isinstance(constraint, Subtype):
+                typevars.add(constraint.sub_type)
+                typevars.add(constraint.super_type)
+            # TODO: Other types of constraints?
+        return typevars
+
+    @staticmethod
+    def _typevar_from_primitive_constraint(constraint: TypeConstraint) -> TypeVariable | None:
+        if isinstance(constraint, Subtype):
+            if (
+                isinstance(constraint.sub_type, DerivedTypeVariable)
+                and type(constraint.sub_type.type_var) is TypeVariable
+            ):
+                return constraint.sub_type.type_var
+            if type(constraint.sub_type) is TypeVariable:
+                return constraint.sub_type
+            if (
+                isinstance(constraint.super_type, DerivedTypeVariable)
+                and type(constraint.super_type.type_var) is TypeVariable
+            ):
+                return constraint.super_type.type_var
+            if type(constraint.super_type) is TypeVariable:
+                return constraint.super_type
+        return None
+
+    @staticmethod
+    def _get_all_paths(
+        graph: networkx.DiGraph[TypeVariable | DerivedTypeVariable],
+        sketch: Sketch,
+        node: DerivedTypeVariable,
+        visited: dict[TypeVariable | DerivedTypeVariable, SketchNode],
+    ):
+        if node not in graph:
+            return
+        curr_node = visited[node]
+        for _, succ, data in sorted(graph.out_edges(node, data=True), key=lambda x: str(x[1])):
+            label = data["label"]
+            if succ not in visited:
+                if isinstance(curr_node.typevar, DerivedTypeVariable):
+                    base_typevar = curr_node.typevar.type_var
+                    labels = curr_node.typevar.labels
+                elif isinstance(curr_node.typevar, TypeVariable):
+                    base_typevar = curr_node.typevar
+                    labels = ()
+                else:
+                    raise TypeError("Unexpected")
+                labels += (label,)
+                succ_derived_typevar = new_dtv(
+                    base_typevar,
+                    labels=labels,
+                )
+                succ_node = SketchNode(succ_derived_typevar)
+                sketch.add_edge(curr_node, succ_node, label)
+                visited[succ] = succ_node
+                SimpleSolver._get_all_paths(graph, sketch, succ, visited)
+                del visited[succ]
+            else:
+                # a cycle exists
+                ref_node = RecursiveRefNode(visited[succ].typevar)
+                sketch.add_edge(curr_node, ref_node, label)
+
+    @staticmethod
+    def _unify(
+        equivalence_classes: dict, cls0: DerivedTypeVariable, cls1: DerivedTypeVariable, graph: networkx.DiGraph
+    ) -> None:
+        # first convert cls0 and cls1 to their equivalence classes
+        cls0 = equivalence_classes[cls0]
+        cls1 = equivalence_classes[cls1]
+
+        # unify if needed
+        if cls0 != cls1:
+            # MakeEquiv
+            existing_elements = {key for key, item in equivalence_classes.items() if item in {cls0, cls1}}
+            rep_cls = cls0
+            for elem in existing_elements:
+                equivalence_classes[elem] = rep_cls
+            # the logic below refers to the retypd reference implementation. it is different from Algorithm E.1
+            # note that graph is used read-only in this method, so we do not need to make copy of edges
+            for _, dst0, data0 in graph.out_edges(cls0, data=True):
+                if "label" in data0 and data0["label"] is not None:
+                    for _, dst1, data1 in graph.out_edges(cls1, data=True):
+                        if data0["label"] == data1["label"] or (
+                            isinstance(data0["label"], Load) and isinstance(data1["label"], Store)
+                        ):
+                            SimpleSolver._unify(
+                                equivalence_classes, equivalence_classes[dst0], equivalence_classes[dst1], graph
+                            )
+
+    def _eq_constraints_from_add(self, typevar: TypeVariable):
+        """
+        Handle Add constraints.
+        """
+        new_constraints = set()
+        for constraint in self._constraints[typevar]:
+            if isinstance(constraint, Add):
+                if (
+                    isinstance(constraint.type_0, TypeVariable)
+                    and not isinstance(constraint.type_0, DerivedTypeVariable)
+                    and isinstance(constraint.type_r, TypeVariable)
+                    and not isinstance(constraint.type_r, DerivedTypeVariable)
+                ):
+                    new_constraints.add(Equivalence(constraint.type_0, constraint.type_r))
+                if (
+                    isinstance(constraint.type_1, TypeVariable)
+                    and not isinstance(constraint.type_1, DerivedTypeVariable)
+                    and isinstance(constraint.type_r, TypeVariable)
+                    and not isinstance(constraint.type_r, DerivedTypeVariable)
+                ):
+                    new_constraints.add(Equivalence(constraint.type_1, constraint.type_r))
+        return new_constraints
+
+    @staticmethod
+    def _discover_equivalence(constraints: set[TypeConstraint]) -> set[Equivalence]:
+        """
+        a <:b && b <: a  ==>  a == b
+        """
+
+        new_eq_constraints: set[Equivalence] = set()
+        subtypes = defaultdict(set)
+        for constraint in constraints:
+            if isinstance(constraint, Subtype):
+                sub_type = constraint.sub_type
+                super_type = constraint.super_type
+                subtypes[sub_type].add(super_type)
+
+        # check everything
+        seen = set()
+        for tv, tv_supers in subtypes.items():
+            for tv_super in tv_supers:
+                if tv_super in subtypes and tv in subtypes[tv_super]:  # noqa: SIM102
+                    # we have a pair of subtypes that are equivalent
+                    if (tv, tv_super) not in seen and (tv_super, tv) not in seen:
+                        new_eq_constraints.add(Equivalence(tv, tv_super))
+                        seen.add((tv, tv_super))
+
+        _l.debug(
+            "Discovered %d equivalence constraints from %d constraints.", len(new_eq_constraints), len(constraints)
+        )
+        return new_eq_constraints
+
+    @staticmethod
+    def _handle_equivalence(
+        constraint_set: set[TypeConstraint],
+    ) -> tuple[set[TypeConstraint], dict[TypeVariable, TypeVariable | TypeConstant]]:
+        graph = networkx.Graph()
+
+        replacements: dict[TypeVariable, TypeVariable | TypeConstant] = {}
+
+        # collect equivalence relations
+        for constraint in constraint_set:
+            if isinstance(constraint, Equivalence):
+                # | type_a == type_b
+                # we apply unification and removes one of them
+                ta, tb = constraint.type_a, constraint.type_b
+                if isinstance(ta, TypeConstant) and isinstance(tb, TypeVariable):
+                    # replace tb with ta
+                    replacements[tb] = ta
+                elif isinstance(ta, TypeVariable) and isinstance(tb, TypeConstant):
+                    # replace ta with tb
+                    replacements[ta] = tb
+                else:
+                    # they are both type variables. we will determine a representative later
+                    graph.add_edge(ta, tb)
+
+        for components in networkx.connected_components(graph):
+            components_lst = sorted(components, key=lambda x: str(x))  # pylint:disable=unnecessary-lambda
+            representative = components_lst[0]
+            for tv in components_lst[1:]:
+                replacements[tv] = representative
+
+        constraints = SimpleSolver._rewrite_constraints_with_replacements(constraint_set, replacements)
+
+        # import pprint
+        # print("Replacements")
+        # pprint.pprint(replacements)
+        # print("Constraints (after replacement)")
+        # pprint.pprint(constraints)
+
+        return constraints, replacements
+
+    @staticmethod
+    def _rewrite_constraints_with_replacements(
+        constraints: set[TypeConstraint], replacements: dict[TypeVariable, TypeVariable]
+    ) -> set[TypeConstraint]:
+        # replace constraints according to a dictionary of type variable replacements
+        replaced_constraints = set()
+        for constraint in constraints:
+            if isinstance(constraint, Existence):
+                replaced, new_constraint = constraint.replace(replacements)
+
+                if replaced:
+                    replaced_constraints.add(new_constraint)
+                else:
+                    replaced_constraints.add(constraint)
+
+            elif isinstance(constraint, Subtype):
+                # subtype <: supertype
+                # replace type variables
+                replaced, new_constraint = constraint.replace(replacements)
+
+                if replaced:
+                    replaced_constraints.add(new_constraint)
+                else:
+                    replaced_constraints.add(constraint)
+        return replaced_constraints
+
+    @staticmethod
+    def _filter_constraints(constraints: set[TypeConstraint]) -> set[TypeConstraint]:
+        """
+        Filter out constraints that we don't yet support.
+        """
+
+        filtered_constraints = set()
+        for constraint in constraints:
+            dropped = False
+            if isinstance(constraint, Subtype) and (
+                (isinstance(constraint.sub_type, TypeConstant) and isinstance(constraint.super_type, TypeConstant))
+                or (
+                    isinstance(constraint.sub_type, DerivedTypeVariable)
+                    and isinstance(constraint.sub_type.labels[-1], ConvertTo)
+                )
+                or (
+                    isinstance(constraint.sub_type, TypeVariable)
+                    and isinstance(constraint.super_type, TypeVariable)
+                    and constraint.sub_type == constraint.super_type
+                )
+            ):
+                dropped = True
+
+            if not dropped:
+                filtered_constraints.add(constraint)
+
+        return filtered_constraints
+
+    @staticmethod
+    def _filter_leaf_typevars(
+        constraints: set[TypeConstraint], tvs_to_solve: set[TypeVariable]
+    ) -> tuple[set[TypeConstraint], dict[TypeVariable, TypeVariable]]:
+        """
+        Filter out leaf type variables that only appear once in the constraints. These type variables are not
+        interesting and can be removed from the constraints.
+        """
+
+        sub_typevars = defaultdict(set)
+        tv_to_dtvs: dict[TypeVariable, set[TypeVariable | DerivedTypeVariable]] = defaultdict(set)
+        for constraint in constraints:
+            if isinstance(constraint, Subtype):
+                if isinstance(constraint.sub_type, TypeVariable):
+                    sub_typevars[constraint.sub_type].add(constraint.super_type)
+                for tv in [constraint.sub_type, constraint.super_type]:
+                    if isinstance(tv, DerivedTypeVariable):
+                        tv_to_dtvs[tv.type_var].add(constraint.sub_type)
+                    elif isinstance(tv, TypeVariable):
+                        tv_to_dtvs[tv].add(constraint.sub_type)
+
+        ub_subtypes: dict[TypeVariable, TypeVariable] = {}
+        for tv, dtvs in tv_to_dtvs.items():
+            if len(dtvs) == 1 and tv in sub_typevars and len(sub_typevars[tv]) == 1:
+                ub = next(iter(sub_typevars[tv]))
+                if ub in tvs_to_solve:
+                    ub_subtypes[tv] = ub
+
+        filtered_constraints = set()
+        for constraint in constraints:
+            if isinstance(constraint, Subtype) and constraint.sub_type in ub_subtypes:
+                continue
+            filtered_constraints.add(constraint)
+
+        return filtered_constraints, ub_subtypes
+
+    def _degrade_constraint_set(self, constraints: set[TypeConstraint]) -> set[TypeConstraint]:
+        """
+        Degrade the constraint set to a smaller set of constraints to speed up the DFA generation process.
+        """
+
+        tv_with_ls = defaultdict(set)  # tv_with_ls are type variables with Loads or Stores
+        graph = networkx.Graph()
+
+        for constraint in constraints:
+            if isinstance(constraint, Subtype):
+                if isinstance(constraint.sub_type, DerivedTypeVariable) and isinstance(
+                    constraint.sub_type.labels[0], (Load, Store)
+                ):
+                    tv_with_ls[constraint.sub_type.type_var].add(constraint.sub_type)
+                if type(constraint.sub_type) is TypeVariable and type(constraint.super_type) is TypeVariable:
+                    graph.add_edge(constraint.sub_type, constraint.super_type)
+
+        tv_to_degrade = set()
+        for tv, dtvs in tv_with_ls.items():
+            if len(dtvs) > 5:
+                # degrade all subtype relationships involving this type variable to equivalence
+                tv_to_degrade.add(tv)
+
+        replacements = {}
+        for components in networkx.connected_components(graph):
+            if len(components) == 1:
+                continue
+            if any(tv in tv_to_degrade for tv in components):
+                components_lst = sorted(components, key=str)
+                representative = components_lst[0]
+                for tv in components_lst[1:]:
+                    replacements[tv] = representative
+
+        degraded_constraints = self._rewrite_constraints_with_replacements(constraints, replacements)
+
+        # discover more equivalence relations
+        eq_constraints = self._discover_equivalence(degraded_constraints)
+        _l.debug("Discovered %d equivalence constraints from degraded constraints.", len(eq_constraints))
+        if eq_constraints:
+            degraded_constraints, eq_replacements = self._handle_equivalence(degraded_constraints | eq_constraints)
+            self._equivalence |= eq_replacements
+
+        # filter them
+        degraded_constraints = self._filter_constraints(degraded_constraints)
+
+        self._equivalence |= replacements
+        return degraded_constraints
+
+    def _convert_arrays(self, constraints):
+        for constraint in constraints:
+            if not isinstance(constraint, Existence):
+                continue
+            inner = constraint.type_
+            if (
+                isinstance(inner, DerivedTypeVariable)
+                and isinstance(inner.one_label(), IsArray)
+                and inner.type_var in self.solution
+            ):
+                curr_type = self.solution[inner.type_var]
+                if (
+                    isinstance(curr_type, Pointer)
+                    and isinstance(curr_type.basetype, Struct)
+                    and 0 in curr_type.basetype.fields
+                ):
+                    # replace all fields with the first field
+                    first_field = curr_type.basetype.fields[0]
+                    for offset in curr_type.basetype.fields:
+                        curr_type.basetype.fields[offset] = first_field
+
+    #
+    # Constraint graph
+    #
+
+    @staticmethod
+    def _generate_constraint_subset(
+        constraints: set[TypeConstraint], typevars: set[TypeVariable]
+    ) -> tuple[set[TypeConstraint], set[TypeVariable]]:
+        subset = set()
+        related_typevars = set(typevars)
+        while True:
+            new = set()
+            for constraint in constraints:
+                if constraint in subset:
+                    continue
+                if isinstance(constraint, Subtype):
+                    if isinstance(constraint.sub_type, DerivedTypeVariable):
+                        subt = constraint.sub_type.type_var
+                    elif isinstance(constraint.sub_type, TypeVariable):
+                        subt = constraint.sub_type
+                    else:
+                        subt = None
+                    if isinstance(constraint.super_type, DerivedTypeVariable):
+                        supert = constraint.super_type.type_var
+                    elif isinstance(constraint.super_type, TypeVariable):
+                        supert = constraint.super_type
+                    else:
+                        supert = None
+                    if subt in related_typevars or supert in related_typevars:
+                        new.add(constraint)
+                        if subt is not None:
+                            related_typevars.add(subt)
+                        if supert is not None:
+                            related_typevars.add(supert)
+            if not new:
+                break
+            subset |= new
+        return subset, related_typevars
+
+    def _generate_constraint_graph(
+        self, constraints: set[TypeConstraint], interesting_variables: set[DerivedTypeVariable]
+    ) -> networkx.DiGraph:
+        """
+        A constraint graph is the same as the finite state transducer that is presented in Appendix D in the retypd
+        paper.
+        """
+
+        graph = networkx.DiGraph()
+        for constraint in constraints:
+            if isinstance(constraint, Subtype):
+                self._constraint_graph_add_edges(
+                    graph, constraint.sub_type, constraint.super_type, interesting_variables
+                )
+        self._constraint_graph_saturate(graph)
+        self._constraint_graph_remove_self_loops(graph)
+        self._constraint_graph_recall_forget_split(graph)
+        return graph
+
+    @staticmethod
+    def _constraint_graph_add_recall_edges(graph: networkx.DiGraph, node: ConstraintGraphNode) -> None:
+        while True:
+            r = node.forget_last_label()
+            if r is None:
+                break
+            prefix, last_label = r
+            graph.add_edge(prefix, node, label=(last_label, "recall"))
+            node = prefix
+
+    @staticmethod
+    def _constraint_graph_add_forget_edges(graph: networkx.DiGraph, node: ConstraintGraphNode) -> None:
+        while True:
+            r = node.forget_last_label()
+            if r is None:
+                break
+            prefix, last_label = r
+            graph.add_edge(node, prefix, label=(last_label, "forget"))
+            node = prefix
+
+    def _constraint_graph_add_edges(
+        self,
+        graph: networkx.DiGraph,
+        subtype: TypeVariable | DerivedTypeVariable,
+        supertype: TypeVariable | DerivedTypeVariable,
+        interesting_variables: set[DerivedTypeVariable],
+    ):
+        # left and right tags
+        if self._typevar_inside_set(self._to_typevar_or_typeconst(subtype), interesting_variables):
+            left_tag = ConstraintGraphTag.LEFT
+        else:
+            left_tag = ConstraintGraphTag.UNKNOWN
+        if self._typevar_inside_set(self._to_typevar_or_typeconst(supertype), interesting_variables):
+            right_tag = ConstraintGraphTag.RIGHT
+        else:
+            right_tag = ConstraintGraphTag.UNKNOWN
+        # nodes
+        forward_src = ConstraintGraphNode(subtype, Variance.COVARIANT, left_tag, FORGOTTEN.PRE_FORGOTTEN)
+        forward_dst = ConstraintGraphNode(supertype, Variance.COVARIANT, right_tag, FORGOTTEN.PRE_FORGOTTEN)
+        graph.add_edge(forward_src, forward_dst)
+        # add recall edges and forget edges
+        self._constraint_graph_add_recall_edges(graph, forward_src)
+        self._constraint_graph_add_forget_edges(graph, forward_dst)
+
+        # backward edges
+        backward_src = forward_dst.inverse()
+        backward_dst = forward_src.inverse()
+        graph.add_edge(backward_src, backward_dst)
+        self._constraint_graph_add_recall_edges(graph, backward_src)
+        self._constraint_graph_add_forget_edges(graph, backward_dst)
+
+    @staticmethod
+    def _constraint_graph_saturate(graph: networkx.DiGraph) -> None:
+        """
+        The saturation algorithm D.2 as described in Appendix of the retypd paper.
+        """
+        R: defaultdict[ConstraintGraphNode, set[tuple[BaseLabel, ConstraintGraphNode]]] = defaultdict(set)
+
+        # initialize the reaching-push sets R(x)
+        for x, y, data in graph.edges(data=True):
+            if "label" in data and data.get("label")[1] == "forget":
+                d = data["label"][0], x
+                R[y].add(d)
+
+        # repeat ... until fixed point
+        changed = True
+        while changed:
+            changed = False
+            for x, y, data in graph.edges(data=True):
+                if "label" not in data:
+                    if R[y].issuperset(R[x]):
+                        continue
+                    changed = True
+                    R[y] |= R[x]
+            for x, y, data in graph.edges(data=True):
+                lbl = data.get("label")
+                if lbl and lbl[1] == "recall":
+                    for _label, z in R[x]:
+                        if not graph.has_edge(z, y):
+                            changed = True
+                            graph.add_edge(z, y)
+            v_contravariant = []
+            for node in graph.nodes:
+                node: ConstraintGraphNode
+                if node.variance == Variance.CONTRAVARIANT:
+                    v_contravariant.append(node)
+            # lazily apply saturation rules corresponding to S-Pointer
+            for x in v_contravariant:
+                for z_label, z in R[x]:
+                    label = None
+                    if isinstance(z_label, Store):
+                        label = Load()
+                    elif isinstance(z_label, Load):
+                        label = Store()
+                    if label is not None:
+                        x_inverse = x.inverse_wo_tag()
+                        d = label, z
+                        if d not in R[x_inverse]:
+                            changed = True
+                            R[x_inverse].add(d)
+
+    @staticmethod
+    def _constraint_graph_remove_self_loops(graph: networkx.DiGraph):
+        for node in list(graph.nodes):
+            if graph.has_edge(node, node):
+                graph.remove_edge(node, node)
+
+    @staticmethod
+    def _constraint_graph_recall_forget_split(graph: networkx.DiGraph):
+        """
+        Ensure that recall edges are not reachable after traversing a forget node.
+        """
+        for src, dst, data in list(graph.edges(data=True)):
+            src: ConstraintGraphNode
+            dst: ConstraintGraphNode
+            if "label" in data and data["label"][1] == "recall":
+                continue
+            forget_src = ConstraintGraphNode(src.typevar, src.variance, src.tag, FORGOTTEN.POST_FORGOTTEN)
+            forget_dst = ConstraintGraphNode(dst.typevar, dst.variance, dst.tag, FORGOTTEN.POST_FORGOTTEN)
+            if "label" in data and data["label"][1] == "forget":
+                graph.remove_edge(src, dst)
+                graph.add_edge(src, forget_dst, **data)
+            graph.add_edge(forget_src, forget_dst, **data)
+
+    @staticmethod
+    def _to_typevar_or_typeconst(obj: TypeVariable | DerivedTypeVariable | TypeConstant) -> TypeVariable | TypeConstant:
+        if isinstance(obj, DerivedTypeVariable):
+            return SimpleSolver._to_typevar_or_typeconst(obj.type_var)
+        if isinstance(obj, (TypeVariable, TypeConstant)):
+            return obj
+        raise TypeError(f"Unsupported type {type(obj)}")
+
+    #
+    # Graph solver
+    #
+
+    @staticmethod
+    def _typevar_inside_set(typevar, typevar_set: set[TypeConstant | TypeVariable | DerivedTypeVariable]) -> bool:
+        if typevar in typevar_set:
+            return True
+        if isinstance(typevar, Struct) and Struct_ in typevar_set:
+            return True
+        if isinstance(typevar, Array) and Array_ in typevar_set:
+            return SimpleSolver._typevar_inside_set(typevar.element, typevar_set)
+        if isinstance(typevar, Pointer) and (Pointer32_ in typevar_set or Pointer64_ in typevar_set):
+            return SimpleSolver._typevar_inside_set(typevar.basetype, typevar_set)
+        return False
+
+    def _solve_constraints_between(
+        self,
+        graph: networkx.DiGraph,
+        starts: set[TypeConstant | TypeVariable | DerivedTypeVariable],
+        ends: set[TypeConstant | TypeVariable | DerivedTypeVariable],
+    ) -> set[TypeConstraint]:
+        start_nodes = set()
+        end_nodes = set()
+        for node in graph.nodes:
+            node: ConstraintGraphNode
+            if (
+                self._typevar_inside_set(self._to_typevar_or_typeconst(node.typevar), starts)
+                and node.tag == ConstraintGraphTag.LEFT
+            ):
+                start_nodes.add(node)
+            if (
+                self._typevar_inside_set(self._to_typevar_or_typeconst(node.typevar), ends)
+                and node.tag == ConstraintGraphTag.RIGHT
+            ):
+                end_nodes.add(node)
+
+        if not start_nodes or not end_nodes:
+            return set()
+
+        dfa_solver = DFAConstraintSolver()
+        try:
+            return dfa_solver.generate_constraints_between(graph, start_nodes, end_nodes)
+        except EmptyEpsilonNFAError:
+            return set()
+
+    #
+    # Type lattice
+    #
+
+    def join(self, t1: TypeConstant | TypeVariable, t2: TypeConstant | TypeVariable) -> TypeConstant:
+        abstract_t1 = self.abstract(t1)
+        abstract_t2 = self.abstract(t2)
+        if abstract_t1 in self._base_lattice and abstract_t2 in self._base_lattice:
+            ancestor = networkx.lowest_common_ancestor(self._base_lattice, abstract_t1, abstract_t2)
+
+            if (
+                isinstance(ancestor, Pointer)
+                and isinstance(abstract_t1, Pointer)
+                and isinstance(abstract_t2, Pointer)
+                and isinstance(t1, Pointer)
+                and isinstance(t2, Pointer)
+            ):
+                return ancestor.__class__(self.join(t1.basetype, t2.basetype))
+
+            if ancestor == abstract_t1:
+                return t1
+            if ancestor == abstract_t2:
+                return t2
+            return ancestor
+        if t1 == Bottom_:
+            return t2
+        if t2 == Bottom_:
+            return t1
+        return Bottom_
+
+    def meet(self, t1: TypeConstant | TypeVariable, t2: TypeConstant | TypeVariable) -> TypeConstant:
+        abstract_t1 = self.abstract(t1)
+        abstract_t2 = self.abstract(t2)
+        if abstract_t1 in self._base_lattice_inverted and abstract_t2 in self._base_lattice_inverted:
+            ancestor = networkx.lowest_common_ancestor(self._base_lattice_inverted, abstract_t1, abstract_t2)
+
+            if (
+                isinstance(ancestor, Pointer)
+                and isinstance(abstract_t1, Pointer)
+                and isinstance(abstract_t2, Pointer)
+                and isinstance(t1, Pointer)
+                and isinstance(t2, Pointer)
+            ):
+                return ancestor.__class__(self.meet(t1.basetype, t2.basetype))
+
+            if ancestor == abstract_t1:
+                return t1
+            if ancestor == abstract_t2:
+                return t2
+            return ancestor
+        if t1 == Top_:
+            return t2
+        if t2 == Top_:
+            return t1
+        return Top_
+
+    @staticmethod
+    def abstract(t: TypeConstant | TypeVariable) -> TypeConstant | TypeVariable:
+        if isinstance(t, Pointer32):
+            return Pointer32()
+        if isinstance(t, Pointer64):
+            return Pointer64()
+        return t
+
+    @staticmethod
+    def _rewrite_constraint(constraint: TypeConstraint, solutions: dict) -> TypeConstraint:
+        if isinstance(constraint, Subtype):
+            replaced = False
+            if isinstance(constraint.sub_type, TypeVariable) and constraint.sub_type in solutions:
+                sub_type = solutions[constraint.sub_type]
+                replaced = True
+            else:
+                sub_type = constraint.sub_type
+            if isinstance(constraint.super_type, TypeVariable) and constraint.super_type in solutions:
+                super_type = solutions[constraint.super_type]
+                replaced = True
+            else:
+                super_type = constraint.super_type
+            return Subtype(sub_type, super_type) if replaced else constraint
+        return constraint
+
+    def determine(
+        self,
+        sketches,
+        tvs,
+        solution: dict,
+        nodes: set[SketchNode] | None = None,
+    ) -> None:
+        """
+        Determine C-like types from sketches.
+
+        :param sketches:            A dictionary storing sketches for each type variable.
+        :param solution:            The dictionary storing C-like types for each type variable. Output.
+        :param nodes:               Optional. Nodes that should be considered in the sketch.
+        :return:                    None
+        """
+
+        for typevar in tvs:
+            self._determine(typevar, sketches[typevar], solution, nodes=nodes)
+
+        for v, eq in self._equivalence.items():
+            if v not in solution and eq in solution:
+                solution[v] = solution[eq]
+
+    def _determine(self, the_typevar, sketch, solution: dict, nodes: set[SketchNode] | None = None):
+        """
+        Return the solution from sketches
+        """
+
+        if not nodes:
+            # TODO: resolve references
+            node = sketch.lookup(the_typevar)
+            assert node is not None
+            nodes = {node}
+
+        # consult the cache
+        cached_results = set()
+        for node in nodes:
+            if node.typevar in self._solution_cache:
+                cached_results.add(self._solution_cache[node.typevar])
+        if len(cached_results) == 1:
+            return next(iter(cached_results))
+        if len(cached_results) > 1:
+            # we get nodes for multiple type variables?
+            raise RuntimeError("Getting nodes for multiple type variables. Unexpected.")
+
+        # collect all successors and the paths (labels) of this type variable
+        path_and_successors = []
+        last_labels = []
+        for node in nodes:
+            path_and_successors += self._collect_sketch_paths(node, sketch)
+        for labels, _ in path_and_successors:
+            if labels:
+                last_labels.append(labels[-1])
+
+        # now, what is this variable?
+        result = None
+
+        if last_labels and all(isinstance(label, (FuncIn, FuncOut)) for label in last_labels):
+            # create a dummy result and dump it to the cache
+            func_type = Function([], [])
+            result = self._pointer_class()(basetype=func_type)
+            for node in nodes:
+                self._solution_cache[node.typevar] = result
+
+            # this is a function variable
+            func_inputs = defaultdict(set)
+            func_outputs = defaultdict(set)
+
+            for labels, succ in path_and_successors:
+                last_label = labels[-1] if labels else None
+
+                if isinstance(last_label, FuncIn):
+                    func_inputs[last_label.loc].add(succ)
+                elif isinstance(last_label, FuncOut):
+                    func_outputs[last_label.loc].add(succ)
+                else:
+                    raise RuntimeError("Unreachable")
+
+            input_args = []
+            output_values = []
+            for vals, out in [(func_inputs, input_args), (func_outputs, output_values)]:
+                for idx in range(max(vals) + 1):
+                    if idx in vals:
+                        sol = self._determine(the_typevar, sketch, solution, nodes=vals[idx])
+                        out.append(sol)
+                    else:
+                        out.append(None)
+
+            # back patch
+            func_type.params = input_args
+            func_type.outputs = output_values
+
+            for node in nodes:
+                solution[node.typevar] = result
+
+        elif path_and_successors:
+            # maybe this is a pointer to a struct?
+            if len(nodes) == 1:
+                the_node = next(iter(nodes))
+                if (
+                    isinstance(the_node.upper_bound, self._pointer_class())
+                    and isinstance(the_node.upper_bound.basetype, Struct)
+                    and the_node.upper_bound.basetype.name
+                ):
+                    # handle pointers to known struct types
+                    result = (
+                        the_node.lower_bound
+                        if not isinstance(the_node.lower_bound, BottomType)
+                        else the_node.upper_bound
+                    )
+                    for node in nodes:
+                        solution[node.typevar] = result
+                        self._solution_cache[node.typevar] = result
+                    return result
+
+            # create a dummy result and shove it into the cache
+            struct_type = Struct(fields={})
+            result = self._pointer_class()(struct_type)
+            for node in nodes:
+                self._solution_cache[node.typevar] = result
+
+            # this might be a struct
+            fields = {}
+
+            candidate_bases = SortedDict()
+
+            for labels, _succ in path_and_successors:
+                last_label = labels[-1] if labels else None
+                if isinstance(last_label, HasField):
+                    # TODO: Really determine the maximum possible size of the field when MAX_POINTSTO_BITS is in use
+                    if last_label.offset not in candidate_bases:
+                        candidate_bases[last_label.offset] = set()
+                    candidate_bases[last_label.offset].add(
+                        1 if last_label.bits == MAX_POINTSTO_BITS else (last_label.bits // 8)
+                    )
+
+            # determine possible bases and map each offset to its base
+            offset_to_base = SortedDict()
+            for start_offset, sizes in candidate_bases.items():
+                for size in sizes:
+                    for i in range(size):
+                        access_off = start_offset + i
+                        if access_off not in offset_to_base:
+                            offset_to_base[access_off] = start_offset
+
+            # determine again the maximum size of each field (at each offset)
+            offset_to_maxsize = defaultdict(int)
+            offset_to_sizes = defaultdict(set)  # we do not consider offsets to each base offset
+            for labels, _succ in path_and_successors:
+                last_label = labels[-1] if labels else None
+                if isinstance(last_label, HasField):
+                    base = offset_to_base[last_label.offset]
+                    access_size = 1 if last_label.bits == MAX_POINTSTO_BITS else (last_label.bits // 8)
+                    offset_to_maxsize[base] = max(offset_to_maxsize[base], (last_label.offset - base) + access_size)
+                    offset_to_sizes[base].add(access_size)
+
+            idx_to_base = {}
+
+            for idx, (labels, _) in enumerate(path_and_successors):
+                last_label = labels[-1] if labels else None
+                if isinstance(last_label, HasField):
+                    prev_offset = next(offset_to_base.irange(maximum=last_label.offset, reverse=True))
+                    idx_to_base[idx] = offset_to_base[prev_offset]
+
+            node_by_offset = defaultdict(set)
+
+            for idx, (labels, succ) in enumerate(path_and_successors):
+                last_label = labels[-1] if labels else None
+                if isinstance(last_label, HasField):
+                    if idx in idx_to_base:
+                        node_by_offset[idx_to_base[idx]].add(succ)
+                    else:
+                        node_by_offset[last_label.offset].add(succ)
+
+            sorted_offsets: list[int] = sorted(node_by_offset)
+            for i in range(len(sorted_offsets)):  # pylint:disable=consider-using-enumerate
+                offset = sorted_offsets[i]
+
+                child_nodes = node_by_offset[offset]
+                sol = self._determine(the_typevar, sketch, solution, nodes=child_nodes)
+                if isinstance(sol, TopType):
+                    # make it an array if possible
+                    elem_size = min(offset_to_sizes[offset])
+                    array_size = offset_to_maxsize[offset]
+                    if array_size % elem_size != 0:
+                        # fall back to byte_t
+                        elem_size = 1
+                    elem_type = int_type(elem_size * 8)
+                    sol = elem_type if array_size == elem_size else Array(elem_type, array_size // elem_size)
+                fields[offset] = sol
+
+            if not fields:
+                result = Top_
+                for node in nodes:
+                    self._solution_cache[node.typevar] = result
+                    solution[node.typevar] = result
+            elif any(off < 0 for off in fields):
+                result = self._pointer_class()(Bottom_)
+                for node in nodes:
+                    self._solution_cache[node.typevar] = result
+                    solution[node.typevar] = result
+            else:
+                # back-patch
+                struct_type.fields = fields
+                for node in nodes:
+                    solution[node.typevar] = result
+
+        if not path_and_successors or result in {Top_, None}:
+            # this is probably a primitive variable
+            lower_bound = Bottom_
+            upper_bound = Top_
+
+            node_sizes = set()
+            for node in nodes:
+                node_size = node.size
+                if node_size is not None:
+                    node_sizes.add(node_size)
+            if len(node_sizes) > 1:
+                # multi-sized reads - cannot converge to a reasonable type
+                result = Bottom_
+            else:
+                for node in nodes:
+                    lower_bound = self.join(lower_bound, node.lower_bound)
+                    upper_bound = self.meet(upper_bound, node.upper_bound)
+                result = lower_bound if not isinstance(lower_bound, BottomType) else upper_bound
+
+            for node in nodes:
+                solution[node.typevar] = result
+                self._solution_cache[node.typevar] = result
+
+        # import pprint
+
+        # print("Solution")
+        # pprint.pprint(result)
+        return result
+
+    @staticmethod
+    def _collect_sketch_paths(node: SketchNodeBase, sketch: Sketch) -> list[tuple[list[BaseLabel], SketchNodeBase]]:
+        """
+        Collect all paths that go from `typevar` to its leaves.
+        """
+        paths = []
+        visited: set[SketchNodeBase] = set()
+        queue: list[tuple[list[BaseLabel], SketchNodeBase]] = [([], node)]
+
+        while queue:
+            curr_labels, curr_node = queue.pop(0)
+            if curr_node in visited:
+                continue
+            visited.add(curr_node)
+
+            out_edges = sketch.graph.out_edges(curr_node, data=True)
+            for _, succ, data in sorted(out_edges, key=lambda x: str(x[1])):
+                if isinstance(succ, RecursiveRefNode):
+                    ref = succ
+                    succ: SketchNode | None = sketch.lookup(succ.target)  # type: ignore
+                    if succ is None:
+                        # failed to resolve...
+                        _l.warning(
+                            "Failed to resolve reference node to a real sketch node for type variable %s", ref.target
+                        )
+                        continue
+                label = data["label"]
+                if isinstance(label, ConvertTo):
+                    # drop conv labels for now
+                    continue
+                if isinstance(label, IsArray):
+                    continue
+                new_labels = [*curr_labels, label]
+                succ: SketchNode
+                if isinstance(succ.typevar, DerivedTypeVariable) and isinstance(succ.typevar.labels[-1], (Load, Store)):
+                    queue.append((new_labels, succ))
+                else:
+                    paths.append((new_labels, succ))
+
+        return paths
+
+    def _pointer_class(self) -> type[Pointer32] | type[Pointer64]:
+        if self.bits == 32:
+            return Pointer32
+        if self.bits == 64:
+            return Pointer64
+        raise NotImplementedError(f"Unsupported bits {self.bits}")
+
+    @staticmethod
+    def dump_constraint_graph(graph: networkx.DiGraph, filename: str) -> None:
+        """
+        Dump the constraint graph to a file.
+        """
+        from networkx.drawing.nx_agraph import write_dot  # pylint:disable=import-outside-toplevel
+
+        write_dot(graph, filename)