//===- ThreadSafetyUtil.h ---------------------------------------*- C++ -*-===//
 
//
 
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 
// See https://llvm.org/LICENSE.txt for license information.
 
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
//
 
//===----------------------------------------------------------------------===//
 
//
 
// This file defines some basic utility classes for use by ThreadSafetyTIL.h
 
//
 
//===----------------------------------------------------------------------===//
 
 
 
#ifndef LLVM_CLANG_ANALYSIS_ANALYSES_THREADSAFETYUTIL_H
 
#define LLVM_CLANG_ANALYSIS_ANALYSES_THREADSAFETYUTIL_H
 
 
 
#include "clang/AST/Decl.h"
 
#include "clang/Basic/LLVM.h"
 
#include "llvm/ADT/StringRef.h"
 
#include "llvm/ADT/iterator_range.h"
 
#include "llvm/Support/Allocator.h"
 
#include <cassert>
 
#include <cstddef>
 
#include <cstring>
 
#include <iterator>
 
#include <ostream>
 
#include <string>
 
#include <vector>
 
 
 
namespace clang {
 
 
 
class Expr;
 
 
 
namespace threadSafety {
 
namespace til {
 
 
 
// Simple wrapper class to abstract away from the details of memory management.
 
// SExprs are allocated in pools, and deallocated all at once.
 
class MemRegionRef {
 
private:
 
  union AlignmentType {
 
    double d;
 
    void *p;
 
    long double dd;
 
    long long ii;
 
  };
 
 
 
public:
 
  MemRegionRef() = default;
 
  MemRegionRef(llvm::BumpPtrAllocator *A) : Allocator(A) {}
 
 
 
  void *allocate(size_t Sz) {
 
    return Allocator->Allocate(Sz, alignof(AlignmentType));
 
  }
 
 
 
  template <typename T> T *allocateT() { return Allocator->Allocate<T>(); }
 
 
 
  template <typename T> T *allocateT(size_t NumElems) {
 
    return Allocator->Allocate<T>(NumElems);
 
  }
 
 
 
private:
 
  llvm::BumpPtrAllocator *Allocator = nullptr;
 
};
 
 
 
} // namespace til
 
} // namespace threadSafety
 
 
 
} // namespace clang
 
 
 
inline void *operator new(size_t Sz,
 
                          clang::threadSafety::til::MemRegionRef &R) {
 
  return R.allocate(Sz);
 
}
 
 
 
namespace clang {
 
namespace threadSafety {
 
 
 
std::string getSourceLiteralString(const Expr *CE);
 
 
 
namespace til {
 
 
 
// A simple fixed size array class that does not manage its own memory,
 
// suitable for use with bump pointer allocation.
 
template <class T> class SimpleArray {
 
public:
 
  SimpleArray() = default;
 
  SimpleArray(T *Dat, size_t Cp, size_t Sz = 0)
 
      : Data(Dat), Size(Sz), Capacity(Cp) {}
 
  SimpleArray(MemRegionRef A, size_t Cp)
 
      : Data(Cp == 0 ? nullptr : A.allocateT<T>(Cp)), Capacity(Cp) {}
 
  SimpleArray(const SimpleArray<T> &A) = delete;
 
 
 
  SimpleArray(SimpleArray<T> &&A)
 
      : Data(A.Data), Size(A.Size), Capacity(A.Capacity) {
 
    A.Data = nullptr;
 
    A.Size = 0;
 
    A.Capacity = 0;
 
  }
 
 
 
  SimpleArray &operator=(SimpleArray &&RHS) {
 
    if (this != &RHS) {
 
      Data = RHS.Data;
 
      Size = RHS.Size;
 
      Capacity = RHS.Capacity;
 
 
 
      RHS.Data = nullptr;
 
      RHS.Size = RHS.Capacity = 0;
 
    }
 
    return *this;
 
  }
 
 
 
  // Reserve space for at least Ncp items, reallocating if necessary.
 
  void reserve(size_t Ncp, MemRegionRef A) {
 
    if (Ncp <= Capacity)
 
      return;
 
    T *Odata = Data;
 
    Data = A.allocateT<T>(Ncp);
 
    Capacity = Ncp;
 
    memcpy(Data, Odata, sizeof(T) * Size);
 
  }
 
 
 
  // Reserve space for at least N more items.
 
  void reserveCheck(size_t N, MemRegionRef A) {
 
    if (Capacity == 0)
 
      reserve(u_max(InitialCapacity, N), A);
 
    else if (Size + N < Capacity)
 
      reserve(u_max(Size + N, Capacity * 2), A);
 
  }
 
 
 
  using iterator = T *;
 
  using const_iterator = const T *;
 
  using reverse_iterator = std::reverse_iterator<iterator>;
 
  using const_reverse_iterator = std::reverse_iterator<const_iterator>;
 
 
 
  size_t size() const { return Size; }
 
  size_t capacity() const { return Capacity; }
 
 
 
  T &operator[](unsigned i) {
 
    assert(i < Size && "Array index out of bounds.");
 
    return Data[i];
 
  }
 
 
 
  const T &operator[](unsigned i) const {
 
    assert(i < Size && "Array index out of bounds.");
 
    return Data[i];
 
  }
 
 
 
  T &back() {
 
    assert(Size && "No elements in the array.");
 
    return Data[Size - 1];
 
  }
 
 
 
  const T &back() const {
 
    assert(Size && "No elements in the array.");
 
    return Data[Size - 1];
 
  }
 
 
 
  iterator begin() { return Data; }
 
  iterator end() { return Data + Size; }
 
 
 
  const_iterator begin() const { return Data; }
 
  const_iterator end() const { return Data + Size; }
 
 
 
  const_iterator cbegin() const { return Data; }
 
  const_iterator cend() const { return Data + Size; }
 
 
 
  reverse_iterator rbegin() { return reverse_iterator(end()); }
 
  reverse_iterator rend() { return reverse_iterator(begin()); }
 
 
 
  const_reverse_iterator rbegin() const {
 
    return const_reverse_iterator(end());
 
  }
 
 
 
  const_reverse_iterator rend() const {
 
    return const_reverse_iterator(begin());
 
  }
 
 
 
  void push_back(const T &Elem) {
 
    assert(Size < Capacity);
 
    Data[Size++] = Elem;
 
  }
 
 
 
  // drop last n elements from array
 
  void drop(unsigned n = 0) {
 
    assert(Size > n);
 
    Size -= n;
 
  }
 
 
 
  void setValues(unsigned Sz, const T& C) {
 
    assert(Sz <= Capacity);
 
    Size = Sz;
 
    for (unsigned i = 0; i < Sz; ++i) {
 
      Data[i] = C;
 
    }
 
  }
 
 
 
  template <class Iter> unsigned append(Iter I, Iter E) {
 
    size_t Osz = Size;
 
    size_t J = Osz;
 
    for (; J < Capacity && I != E; ++J, ++I)
 
      Data[J] = *I;
 
    Size = J;
 
    return J - Osz;
 
  }
 
 
 
  llvm::iterator_range<reverse_iterator> reverse() {
 
    return llvm::reverse(*this);
 
  }
 
 
 
  llvm::iterator_range<const_reverse_iterator> reverse() const {
 
    return llvm::reverse(*this);
 
  }
 
 
 
private:
 
  // std::max is annoying here, because it requires a reference,
 
  // thus forcing InitialCapacity to be initialized outside the .h file.
 
  size_t u_max(size_t i, size_t j) { return (i < j) ? j : i; }
 
 
 
  static const size_t InitialCapacity = 4;
 
 
 
  T *Data = nullptr;
 
  size_t Size = 0;
 
  size_t Capacity = 0;
 
};
 
 
 
}  // namespace til
 
 
 
// A copy on write vector.
 
// The vector can be in one of three states:
 
// * invalid -- no operations are permitted.
 
// * read-only -- read operations are permitted.
 
// * writable -- read and write operations are permitted.
 
// The init(), destroy(), and makeWritable() methods will change state.
 
template<typename T>
 
class CopyOnWriteVector {
 
  class VectorData {
 
  public:
 
    unsigned NumRefs = 1;
 
    std::vector<T> Vect;
 
 
 
    VectorData() = default;
 
    VectorData(const VectorData &VD) : Vect(VD.Vect) {}
 
  };
 
 
 
public:
 
  CopyOnWriteVector() = default;
 
  CopyOnWriteVector(CopyOnWriteVector &&V) : Data(V.Data) { V.Data = nullptr; }
 
 
 
  CopyOnWriteVector &operator=(CopyOnWriteVector &&V) {
 
    destroy();
 
    Data = V.Data;
 
    V.Data = nullptr;
 
    return *this;
 
  }
 
 
 
  // No copy constructor or copy assignment.  Use clone() with move assignment.
 
  CopyOnWriteVector(const CopyOnWriteVector &) = delete;
 
  CopyOnWriteVector &operator=(const CopyOnWriteVector &) = delete;
 
 
 
  ~CopyOnWriteVector() { destroy(); }
 
 
 
  // Returns true if this holds a valid vector.
 
  bool valid() const  { return Data; }
 
 
 
  // Returns true if this vector is writable.
 
  bool writable() const { return Data && Data->NumRefs == 1; }
 
 
 
  // If this vector is not valid, initialize it to a valid vector.
 
  void init() {
 
    if (!Data) {
 
      Data = new VectorData();
 
    }
 
  }
 
 
 
  // Destroy this vector; thus making it invalid.
 
  void destroy() {
 
    if (!Data)
 
      return;
 
    if (Data->NumRefs <= 1)
 
      delete Data;
 
    else
 
      --Data->NumRefs;
 
    Data = nullptr;
 
  }
 
 
 
  // Make this vector writable, creating a copy if needed.
 
  void makeWritable() {
 
    if (!Data) {
 
      Data = new VectorData();
 
      return;
 
    }
 
    if (Data->NumRefs == 1)
 
      return;   // already writeable.
 
    --Data->NumRefs;
 
    Data = new VectorData(*Data);
 
  }
 
 
 
  // Create a lazy copy of this vector.
 
  CopyOnWriteVector clone() { return CopyOnWriteVector(Data); }
 
 
 
  using const_iterator = typename std::vector<T>::const_iterator;
 
 
 
  const std::vector<T> &elements() const { return Data->Vect; }
 
 
 
  const_iterator begin() const { return elements().cbegin(); }
 
  const_iterator end() const { return elements().cend(); }
 
 
 
  const T& operator[](unsigned i) const { return elements()[i]; }
 
 
 
  unsigned size() const { return Data ? elements().size() : 0; }
 
 
 
  // Return true if V and this vector refer to the same data.
 
  bool sameAs(const CopyOnWriteVector &V) const { return Data == V.Data; }
 
 
 
  // Clear vector.  The vector must be writable.
 
  void clear() {
 
    assert(writable() && "Vector is not writable!");
 
    Data->Vect.clear();
 
  }
 
 
 
  // Push a new element onto the end.  The vector must be writable.
 
  void push_back(const T &Elem) {
 
    assert(writable() && "Vector is not writable!");
 
    Data->Vect.push_back(Elem);
 
  }
 
 
 
  // Gets a mutable reference to the element at index(i).
 
  // The vector must be writable.
 
  T& elem(unsigned i) {
 
    assert(writable() && "Vector is not writable!");
 
    return Data->Vect[i];
 
  }
 
 
 
  // Drops elements from the back until the vector has size i.
 
  void downsize(unsigned i) {
 
    assert(writable() && "Vector is not writable!");
 
    Data->Vect.erase(Data->Vect.begin() + i, Data->Vect.end());
 
  }
 
 
 
private:
 
  CopyOnWriteVector(VectorData *D) : Data(D) {
 
    if (!Data)
 
      return;
 
    ++Data->NumRefs;
 
  }
 
 
 
  VectorData *Data = nullptr;
 
};
 
 
 
inline std::ostream& operator<<(std::ostream& ss, const StringRef str) {
 
  return ss.write(str.data(), str.size());
 
}
 
 
 
} // namespace threadSafety
 
} // namespace clang
 
 
 
#endif // LLVM_CLANG_ANALYSIS_ANALYSES_THREADSAFETYUTIL_H