mirror of
https://github.com/revng/revng
synced 2026-06-21 14:07:57 +00:00
b6520330fb
This commit fixes a use-after-poison bug. DLADeduplicateUnionFields::runOnTypeSystem iterates over the DLA graph while mutating it. Due to iterations on copies of the edges, there were some edges to dead nodes that were still hanging around and could be accessed. This commit fixes the problem by not saving copies of the edges, but saving LayoutTypeSystemNode * and iterating lazily on the edges starting from the node itself every time. In this way, if the graph is mutated, we never have outdated edges pointing to dead nodes, and we don't have use-after-poison.
585 lines
21 KiB
C++
585 lines
21 KiB
C++
//
|
|
// Copyright (c) rev.ng Srls. See LICENSE.md for details.
|
|
//
|
|
|
|
#include <algorithm>
|
|
#include <compare>
|
|
#include <cstdint>
|
|
#include <iterator>
|
|
|
|
#include "llvm/ADT/GraphTraits.h"
|
|
#include "llvm/ADT/PostOrderIterator.h"
|
|
#include "llvm/ADT/STLExtras.h"
|
|
#include "llvm/ADT/SetVector.h"
|
|
#include "llvm/Support/Debug.h"
|
|
|
|
#include "revng/Support/Assert.h"
|
|
#include "revng/Support/Debug.h"
|
|
|
|
#include "revng-c/DataLayoutAnalysis/DLATypeSystem.h"
|
|
|
|
#include "../DLAHelpers.h"
|
|
#include "DLAStep.h"
|
|
|
|
using LTSN = dla::LayoutTypeSystemNode;
|
|
using order = std::strong_ordering;
|
|
using Link = dla::LayoutTypeSystemNode::Link;
|
|
using EdgeList = std::vector<Link>;
|
|
using Tag = dla::TypeLinkTag;
|
|
using NonPointerFilterT = EdgeFilteredGraph<LTSN *, dla::isNotPointerEdge>;
|
|
|
|
using namespace llvm;
|
|
|
|
static Logger<> Log("dla-deduplicate-union-fields");
|
|
static Logger<> CmpLog("dla-duf-comparisons");
|
|
|
|
namespace dla {
|
|
|
|
///\brief Strong ordering for nodes: order by size, then by number of successors
|
|
static order cmpNodes(const LTSN *A, const LTSN *B) {
|
|
if (A == B)
|
|
return order::equal;
|
|
|
|
if (not A)
|
|
return order::less;
|
|
|
|
if (not B)
|
|
return order::greater;
|
|
|
|
const auto SizeCmp = A->Size <=> B->Size;
|
|
if (SizeCmp != order::equal) {
|
|
revng_log(CmpLog, "Different sizes");
|
|
return SizeCmp;
|
|
}
|
|
|
|
size_t NChild1 = A->Successors.size();
|
|
size_t NChild2 = B->Successors.size();
|
|
const auto NChildCmp = NChild1 <=> NChild2;
|
|
if (NChildCmp != order::equal) {
|
|
revng_log(CmpLog,
|
|
"Different number of successors: node "
|
|
<< A->ID << " has " << NChild1 << " successors, node " << B->ID
|
|
<< " has" << NChild2 << " successors");
|
|
return NChildCmp;
|
|
}
|
|
|
|
return order::equal;
|
|
}
|
|
|
|
///\brief Strong ordering for edges: order by kind, then by offset expression
|
|
///
|
|
///\note Inheritance and instance at offset 0 can be considered equivalent when
|
|
/// comparing subtrees.
|
|
static order
|
|
cmpEdgeTags(const Tag *A, const Tag *B, bool IgnoreInheritance = true) {
|
|
if (A == B)
|
|
return order::equal;
|
|
revng_assert(A != nullptr and B != nullptr);
|
|
|
|
auto KindA = A->getKind();
|
|
auto KindB = B->getKind();
|
|
|
|
// If at least one is a pointer, only look at the kind
|
|
if (KindA == TypeLinkTag::LK_Pointer or KindB == TypeLinkTag::LK_Pointer)
|
|
return KindA <=> KindB;
|
|
|
|
// If A is an inheritance edge, consider it as an instance-offset-0 edge
|
|
OffsetExpression OffA;
|
|
if (KindA == TypeLinkTag::LK_Inheritance) {
|
|
if (IgnoreInheritance)
|
|
KindA = TypeLinkTag::LK_Instance;
|
|
OffA.Offset = 0;
|
|
} else {
|
|
OffA = A->getOffsetExpr();
|
|
}
|
|
|
|
// If B is an inheritance edge, consider it as an instance-offset-0 edge
|
|
OffsetExpression OffB;
|
|
if (KindB == TypeLinkTag::LK_Inheritance) {
|
|
if (IgnoreInheritance)
|
|
KindB = TypeLinkTag::LK_Instance;
|
|
OffB.Offset = 0;
|
|
} else {
|
|
OffB = B->getOffsetExpr();
|
|
}
|
|
|
|
const auto KindCmp = KindA <=> KindB;
|
|
if (KindCmp != order::equal)
|
|
return KindCmp;
|
|
|
|
const auto OffsetCmp = OffA.Offset <=> OffB.Offset;
|
|
if (OffsetCmp != order::equal)
|
|
return OffsetCmp;
|
|
|
|
const auto StrideSizeCmp = OffA.Strides.size() <=> OffB.Strides.size();
|
|
if (StrideSizeCmp != order::equal)
|
|
return StrideSizeCmp;
|
|
for (const auto &[StrA, StrB] : llvm::zip(OffA.Strides, OffB.Strides)) {
|
|
const auto StrideCmp = StrA <=> StrB;
|
|
if (StrideCmp != order::equal)
|
|
return StrideCmp;
|
|
}
|
|
|
|
const auto TCSizeCmp = OffA.TripCounts.size() <=> OffB.TripCounts.size();
|
|
if (TCSizeCmp != order::equal)
|
|
return TCSizeCmp;
|
|
for (const auto &[TCA, TCB] : llvm::zip(OffA.TripCounts, OffB.TripCounts)) {
|
|
if (not TCA and not TCB)
|
|
continue;
|
|
if (TCA and not TCB)
|
|
return order::less;
|
|
if (not TCA and TCB)
|
|
return order::greater;
|
|
|
|
if (TCA and TCB) {
|
|
const auto TCCmp = *TCA <=> *TCB;
|
|
if (TCCmp != order::equal)
|
|
return TCCmp;
|
|
}
|
|
}
|
|
|
|
return order::equal;
|
|
}
|
|
|
|
///\brief Strong ordering for links: compare edge tags and destination node
|
|
static order cmpLinks(const Link &A, const Link &B, bool IgnoreInheritance) {
|
|
const order EdgeOrder = cmpEdgeTags(A.second, B.second, IgnoreInheritance);
|
|
if (EdgeOrder != order::equal)
|
|
return EdgeOrder;
|
|
|
|
// Pointer edges are equivalent only if they correspond to the same node
|
|
if (isPointerEdge(A)) {
|
|
revng_assert(isPointerEdge(B));
|
|
return A.first->ID <=> B.first->ID;
|
|
}
|
|
|
|
const order NodeOrder = cmpNodes(A.first, B.first);
|
|
if (NodeOrder != order::equal)
|
|
return NodeOrder;
|
|
|
|
return order::equal;
|
|
}
|
|
|
|
///\brief Compare two subtrees, saving the visited nodes onto two stacks
|
|
static std::tuple<order, EdgeList, EdgeList>
|
|
exploreAndCompare(const Link &Child1, const Link &Child2);
|
|
|
|
///\brief Recursively define an ordering between children of a node
|
|
static bool linkOrderLess(const Link &A, const Link &B) {
|
|
const order LinkOrder = cmpLinks(A, B, /*IgnoreInheritance=*/false);
|
|
if (LinkOrder != order::equal)
|
|
return LinkOrder < 0;
|
|
|
|
revng_log(CmpLog,
|
|
"No order between " << A.first->ID << " and " << A.first->ID
|
|
<< ", must recur");
|
|
// In case the two nodes are equivalent, explore the whole subtree
|
|
// TODO: cache the result of this comparison?
|
|
const auto [SubtreeOrder, _, __] = exploreAndCompare(A, B);
|
|
revng_assert(SubtreeOrder != order::equal);
|
|
return SubtreeOrder == order::less;
|
|
}
|
|
|
|
static std::tuple<order, EdgeList, EdgeList>
|
|
exploreAndCompare(const Link &Child1, const Link &Child2) {
|
|
if (Child1.first->ID == Child2.first->ID)
|
|
return { order::equal, { Child1 }, { Child2 } };
|
|
|
|
EdgeList VisitStack1{ Child1 }, VisitStack2{ Child2 };
|
|
EdgeList NextToVisit1, NextToVisit2;
|
|
size_t CurIdx = 0;
|
|
do {
|
|
// Append the newly found nodes to the visit stack of each subtree
|
|
size_t NextSize = NextToVisit1.size();
|
|
revng_assert(NextSize == NextToVisit2.size());
|
|
|
|
if (NextSize > 0) {
|
|
size_t PrevSize = VisitStack1.size();
|
|
VisitStack1.reserve(PrevSize + NextSize);
|
|
VisitStack2.reserve(PrevSize + NextSize);
|
|
VisitStack1.insert(VisitStack1.end(),
|
|
std::make_move_iterator(NextToVisit1.begin()),
|
|
std::make_move_iterator(NextToVisit1.end()));
|
|
VisitStack2.insert(VisitStack2.end(),
|
|
std::make_move_iterator(NextToVisit2.begin()),
|
|
std::make_move_iterator(NextToVisit2.end()));
|
|
NextToVisit1.clear();
|
|
NextToVisit2.clear();
|
|
}
|
|
|
|
// Perform bfs on new nodes
|
|
for (; CurIdx < VisitStack1.size(); CurIdx++) {
|
|
const Link &L1 = VisitStack1[CurIdx];
|
|
const Link &L2 = VisitStack2[CurIdx];
|
|
const auto &[Node1, Edge1] = L1;
|
|
const auto &[Node2, Edge2] = L2;
|
|
revng_log(CmpLog, "Comparing " << Node1->ID << " with " << Node2->ID);
|
|
|
|
if (Node1->ID == Node2->ID)
|
|
continue;
|
|
|
|
// Return if the links are different
|
|
const order LinkOrder = cmpLinks(L1, L2, /*IgnoreInheritance=*/true);
|
|
if (LinkOrder != order::equal)
|
|
return { LinkOrder, VisitStack1, VisitStack2 };
|
|
|
|
revng_log(CmpLog, "Could not tell the difference");
|
|
|
|
if (not isPointerEdge(L1)) {
|
|
// Enqueue the successors of the current nodes
|
|
revng_assert(not isPointerEdge(L2));
|
|
NextToVisit1.reserve(NextToVisit1.size() + Node1->Successors.size());
|
|
NextToVisit2.reserve(NextToVisit2.size() + Node2->Successors.size());
|
|
llvm::copy(Node1->Successors, std::back_inserter(NextToVisit1));
|
|
llvm::copy(Node2->Successors, std::back_inserter(NextToVisit2));
|
|
|
|
// Sort the newly enqueued nodes
|
|
size_t NChildren = Node1->Successors.size();
|
|
revng_assert(NChildren == Node2->Successors.size());
|
|
|
|
std::sort(NextToVisit1.end() - NChildren,
|
|
NextToVisit1.end(),
|
|
linkOrderLess);
|
|
std::sort(NextToVisit2.end() - NChildren,
|
|
NextToVisit2.end(),
|
|
linkOrderLess);
|
|
}
|
|
}
|
|
} while (NextToVisit1.size() > 0);
|
|
|
|
return { order::equal, VisitStack1, VisitStack2 };
|
|
}
|
|
|
|
///\brief Check if two subtrees are equivalent, saving the visited nodes in the
|
|
/// order in which they were compared.
|
|
static std::tuple<bool, EdgeList, EdgeList>
|
|
areEquivSubtrees(const Link &Child1, const Link &Child2) {
|
|
auto [Result, Visited1, Visited2] = exploreAndCompare(Child1, Child2);
|
|
bool AreSubtreesEqual = Result == order::equal;
|
|
|
|
return { AreSubtreesEqual, Visited1, Visited2 };
|
|
}
|
|
|
|
///\brief Visit the two subtrees of \a Child1 and \a Child2. If they are
|
|
/// equivalent, merge each node with the one it has been compared to.
|
|
///
|
|
///\return true if the two nodes were merged, and merged subtree
|
|
///\param TS the graph in which the comparison should be performed
|
|
///\param Child1 the root of the first subtree
|
|
///\param Child2 the root of the second subtree, will be collapsed if
|
|
/// equivalent to the subtree of \a Child1
|
|
static std::tuple<bool, EdgeList, std::set<LTSN *>>
|
|
mergeIfTopologicallyEq(LayoutTypeSystem &TS,
|
|
const Link &Child1,
|
|
const Link &Child2) {
|
|
if (Child1.first == Child2.first) {
|
|
revng_log(CmpLog, "Same Node!");
|
|
return { false, {}, {} };
|
|
}
|
|
|
|
auto [AreEquiv, Subtree1, Subtree2] = areEquivSubtrees(Child1, Child2);
|
|
if (not AreEquiv) {
|
|
revng_log(CmpLog, "Different!");
|
|
return { false, {}, {} };
|
|
}
|
|
|
|
revng_log(CmpLog, "Equivalent!");
|
|
|
|
// Create a map between nodes to merge and the corresponding merge
|
|
// destination, in order to:
|
|
// 1. avoid duplicates in merging list
|
|
// 2. check that a node is never merged into two separate nodes
|
|
// 3. handle the case in which the merge destination has to be merged itself
|
|
std::map</*to merge*/ LTSN *, /*to keep*/ LTSN *> MergeMap;
|
|
for (const auto &[Link1, Link2] : llvm::zip(Subtree1, Subtree2)) {
|
|
auto *NodeToKeep = Link1.first;
|
|
auto *NodeToMerge = Link2.first;
|
|
|
|
const auto &[_, Inserted] = MergeMap.insert({ NodeToMerge, NodeToKeep });
|
|
revng_assert(Inserted or MergeMap.at(NodeToMerge) == NodeToKeep);
|
|
}
|
|
|
|
// Redirect chains of nodes that have to be merged together
|
|
llvm::SmallPtrSet<LTSN *, 8> Subtree1MergedNodes;
|
|
for (auto &[NodeToMerge, NodeToKeep] : MergeMap) {
|
|
if (NodeToKeep == NodeToMerge or Subtree1MergedNodes.contains(NodeToMerge))
|
|
continue;
|
|
|
|
auto MapEntry = MergeMap.find(NodeToKeep);
|
|
llvm::SmallPtrSet<LTSN *, 8> MergeChain;
|
|
|
|
// Find chains of nodes to merge
|
|
while (MapEntry != MergeMap.end()) {
|
|
Subtree1MergedNodes.insert(MapEntry->first);
|
|
const auto &[_, Inserted] = MergeChain.insert(NodeToKeep);
|
|
// Avoid loops
|
|
if (not Inserted)
|
|
break;
|
|
|
|
NodeToKeep = MapEntry->second;
|
|
|
|
// Go to next node of the chain
|
|
MapEntry = MergeMap.find(NodeToKeep);
|
|
}
|
|
|
|
// Update the merge destination of all the nodes of the chain
|
|
for (auto *N : MergeChain)
|
|
MergeMap.at(N) = NodeToKeep;
|
|
}
|
|
|
|
// Execute merge
|
|
std::set<LTSN *> ErasedNodes;
|
|
for (auto &[NodeToMerge, NodeToKeep] : MergeMap) {
|
|
if (NodeToKeep == NodeToMerge)
|
|
continue;
|
|
|
|
// TODO: light merge
|
|
TS.mergeNodes({ NodeToKeep, NodeToMerge });
|
|
ErasedNodes.insert(NodeToMerge);
|
|
}
|
|
|
|
// Remove merged nodes from subtree1
|
|
if (Subtree1MergedNodes.size() > 0) {
|
|
for (auto It = Subtree1.begin(); It != Subtree1.end();) {
|
|
if (Subtree1MergedNodes.contains(It->first))
|
|
It = Subtree1.erase(It);
|
|
else
|
|
++It;
|
|
}
|
|
}
|
|
|
|
return { true, Subtree1, ErasedNodes };
|
|
}
|
|
|
|
///\brief Remove conflicting edges and collapse single children after merging.
|
|
static bool
|
|
postProcessMerge(LayoutTypeSystem &TS, const EdgeList &MergedSubtree) {
|
|
bool Modified = false;
|
|
|
|
// Merging nodes together might have created conflicting edges, i.e.
|
|
// instance-offset-0 edges that connect two nodes with an already
|
|
// existing inheritance edges: remove them.
|
|
for (auto &E : MergedSubtree) {
|
|
// Materialize predecessors to avoid iterator invalidation
|
|
llvm::SmallVector<LTSN *, 8> PredNodes;
|
|
for (auto &PredLink : E.first->Predecessors)
|
|
PredNodes.push_back(PredLink.first);
|
|
|
|
// Remove conflicts from predecessors
|
|
for (auto &Pred : PredNodes)
|
|
Modified |= RemoveConflictingEdges::removeConflicts(TS, Pred);
|
|
|
|
// Remove conflict from node
|
|
Modified |= RemoveConflictingEdges::removeConflicts(TS, E.first);
|
|
}
|
|
|
|
// Merging nodes and removing conflicts might have created situations in
|
|
// which a node has a single collapsible child: collapse it into its parent.
|
|
LTSN *SubtreeRoot = MergedSubtree.begin()->first;
|
|
for (auto &N : post_order(NonPointerFilterT(SubtreeRoot)))
|
|
Modified |= CollapseSingleChild::collapseSingle(TS, N);
|
|
|
|
return Modified;
|
|
}
|
|
|
|
static auto getSuccEdgesToChild(LTSN *Parent, LTSN *Child) {
|
|
auto &Succ = Parent->Successors;
|
|
return llvm::iterator_range(Succ.lower_bound({ Child, nullptr }),
|
|
Succ.upper_bound({ std::next(Child), nullptr }));
|
|
}
|
|
|
|
bool DeduplicateUnionFields::runOnTypeSystem(LayoutTypeSystem &TS) {
|
|
bool TypeSystemChanged = false;
|
|
if (VerifyLog.isEnabled())
|
|
revng_assert(TS.verifyDAG() and TS.verifyInheritanceTree());
|
|
|
|
if (Log.isEnabled())
|
|
TS.dumpDotOnFile("before-deduplicate-union-fields.dot");
|
|
|
|
llvm::SmallPtrSet<LTSN *, 16> VisitedUnions;
|
|
|
|
for (LTSN *Root : llvm::nodes(&TS)) {
|
|
revng_assert(Root != nullptr);
|
|
if (not isRoot(Root))
|
|
continue;
|
|
|
|
llvm::SmallVector<LTSN *, 8> PostOrderFromRoot;
|
|
for (LTSN *UnionNode : post_order(NonPointerFilterT(Root))) {
|
|
if (UnionNode->InterferingInfo != AllChildrenAreInterfering
|
|
or VisitedUnions.contains(UnionNode))
|
|
continue;
|
|
|
|
revng_log(Log, "****** Union Node found: " << UnionNode->ID);
|
|
VisitedUnions.insert(UnionNode);
|
|
PostOrderFromRoot.push_back(UnionNode);
|
|
}
|
|
|
|
// Visit all Union nodes in post-order. The post-order needs to be cached
|
|
// because children can be merged during traversal, which would invalidate
|
|
// iterators in llvm::post_order if we use it vanilla.
|
|
for (LTSN *UnionNode : PostOrderFromRoot) {
|
|
revng_log(Log,
|
|
"****** Try to dedup children of UnionNode with ID: "
|
|
<< UnionNode->ID);
|
|
|
|
// Since a node can be connected to the parent union by more than one
|
|
// edge, we keep track of the **nodes** that we have to visit and the
|
|
// **edges** we visited. In this way, when comparing subtrees, we consider
|
|
// all the edges incoming from the parent node, so that, if we merge
|
|
// two nodes, we don't have to update other links in the worklist.
|
|
llvm::SmallSetVector<LTSN *, 8> UnionChildrenToCompare;
|
|
llvm::SmallSet<LTSN *, 8> OriginalUnionChildren;
|
|
llvm::SmallSet<LTSN *, 8> AnalyzedNodesNotMerged;
|
|
|
|
// We keep a separate list of successors since we might need to re-enqueue
|
|
// some of them.
|
|
for (const Link &L : UnionNode->Successors) {
|
|
UnionChildrenToCompare.insert(L.first);
|
|
OriginalUnionChildren.insert(L.first);
|
|
}
|
|
|
|
bool UnionNodeChanged = false;
|
|
while (UnionChildrenToCompare.size() > 0) {
|
|
LTSN *CurChild = UnionChildrenToCompare.pop_back_val();
|
|
|
|
// The CurChild can be connected to UnionNode with more than one edge
|
|
// (inheritance and instance at offset 0), so consider them all when
|
|
// comparing CurChild with the AnalyzedNotMerged.
|
|
// TODO: turn this into an iterator range
|
|
bool UnionChildrenMerged = false;
|
|
auto CurChildEdges = getSuccEdgesToChild(UnionNode, CurChild);
|
|
for (auto &CurLink : CurChildEdges) {
|
|
revng_assert(isInheritanceEdge(CurLink) or isInstanceEdge(CurLink));
|
|
|
|
// We want to compare CurChild with all the other nodes that we have
|
|
// looked at in previous iterations, and try to merge it with one of
|
|
// them.
|
|
for (LTSN *NotMergedNode : AnalyzedNodesNotMerged) {
|
|
auto MergedEdges = getSuccEdgesToChild(UnionNode, NotMergedNode);
|
|
|
|
bool AnalyzedNotMergedInvalidated = false;
|
|
for (const Link &NotMergedLink : MergedEdges) {
|
|
revng_assert(isInheritanceEdge(NotMergedLink)
|
|
or isInstanceEdge(NotMergedLink));
|
|
|
|
auto [IsMerged,
|
|
Preserved,
|
|
Erased] = mergeIfTopologicallyEq(TS,
|
|
NotMergedLink,
|
|
CurLink);
|
|
if (not IsMerged)
|
|
continue;
|
|
|
|
// If we merged something, there should be at least one preserved
|
|
// node and one erase one
|
|
revng_assert(not Preserved.empty());
|
|
revng_assert(not Erased.empty());
|
|
|
|
TypeSystemChanged = true;
|
|
UnionNodeChanged = true;
|
|
revng_log(Log, "Merged!");
|
|
|
|
// The following call coul remove stuff from Preserved and add it
|
|
// to Erased.
|
|
// BUT:
|
|
// - postProcessMerge only calls
|
|
// - RemoveConflictingEdges::removeConflicts only removes
|
|
// edges,
|
|
// - CollapseSingleChild::collapseSingle only removes nodes
|
|
// with if these two are safe we're good
|
|
// - RemoveConflictingEdges::removeConflicts only removes edges,
|
|
// not nodes, so it cannot change Preserved nor Erases, hence
|
|
// it's safe
|
|
// - CollapseSingleChild::collapseSingle only removes nodes with
|
|
// exactly one parent, so it cannot remove nodes that were not
|
|
// originally children of the union, because if they were they
|
|
// would have had more than one incoming edge so they wouldn't
|
|
// be removed.
|
|
postProcessMerge(TS, Preserved);
|
|
|
|
for (auto &ErasedNode : Erased) {
|
|
// The ErasedNode has been deleted while merging, so we never
|
|
// want it to be processed again.
|
|
bool Erased = UnionChildrenToCompare.remove(ErasedNode);
|
|
UnionChildrenMerged |= Erased;
|
|
Erased = AnalyzedNodesNotMerged.erase(ErasedNode);
|
|
AnalyzedNotMergedInvalidated |= Erased;
|
|
}
|
|
|
|
UnionChildrenMerged |= Erased.contains(CurChild);
|
|
|
|
// This should always be true, since whenever we merge we are at
|
|
// least erasing CurChild, merging it with NotMergedNode.
|
|
revng_assert(UnionChildrenMerged);
|
|
|
|
for (auto &[PreservedNode, _] : Preserved) {
|
|
// The PreservedNode is preserved (not erased) by merge, but
|
|
// the merge process might have changed it.
|
|
// So, if it'a an original children of the UnionNode, we need
|
|
// to re-process it, hence we add it to UnionChildreToCompare,
|
|
// and remove it from AnalyzedNodesNotMerged.
|
|
if (OriginalUnionChildren.count(PreservedNode)) {
|
|
UnionChildrenToCompare.insert(PreservedNode);
|
|
bool Erased = AnalyzedNodesNotMerged.erase(PreservedNode);
|
|
AnalyzedNotMergedInvalidated |= Erased;
|
|
}
|
|
}
|
|
|
|
// This should always be true, since whenever we merge we are at
|
|
// least preserving the NotMergedNode, which might have been
|
|
// changed by the merge.
|
|
revng_assert(AnalyzedNotMergedInvalidated);
|
|
|
|
// We have merged the NotMergedNode into CurChild, we have to
|
|
// brake out of all the loops looking at CurChild and at
|
|
// AnalyzedNodesNotMerged, since both of these might have
|
|
// changed.
|
|
// looking at the next node in UnionChildrenToCompare.
|
|
break;
|
|
}
|
|
|
|
if (AnalyzedNotMergedInvalidated) {
|
|
// If we just merged CurChild into NotMergedNode we have
|
|
// invalidated the AnalyzedNodesNotMerged iterators.
|
|
// So we have to exit this loop and re-start iterating on
|
|
// UnionChildrenToCompare.
|
|
break;
|
|
}
|
|
}
|
|
|
|
// If the children of the UnionNode have been changed by the merge,
|
|
// the CurChildEdges iterator ranges have been invalidated. So we
|
|
// have to break out of this loop as well.
|
|
if (UnionChildrenMerged)
|
|
break;
|
|
}
|
|
|
|
// If we haven't merged CurChild with anything we can mark it as
|
|
// analyzed and not merged.
|
|
if (not UnionChildrenMerged) {
|
|
AnalyzedNodesNotMerged.insert(CurChild);
|
|
revng_log(Log, "Child " << CurChild->ID << " not merged");
|
|
}
|
|
}
|
|
|
|
// Collapse the union node if we are left with only one member
|
|
if (UnionNodeChanged) {
|
|
CollapseSingleChild::collapseSingle(TS, UnionNode);
|
|
RemoveConflictingEdges::removeConflicts(TS, UnionNode);
|
|
}
|
|
}
|
|
}
|
|
|
|
if (Log.isEnabled())
|
|
TS.dumpDotOnFile("after-deduplicate-union-fields.dot");
|
|
if (VerifyLog.isEnabled()) {
|
|
revng_assert(TS.verifyInheritanceDAG());
|
|
revng_assert(TS.verifyInheritanceTree());
|
|
revng_assert(TS.verifyConflicts());
|
|
}
|
|
|
|
return TypeSystemChanged;
|
|
}
|
|
|
|
} // end namespace dla
|