C++：将运算符作为参数传递会导致错误 "expected an identifier"

Question

我使用 bazel 0.20.0 和 VS2015 为 C++ 编译了 Tensorflow

我在 VS2019 中创建了一个简单的 C++-Project 并尝试构建它但出现以下问题：

...\tensorflow\core\platform\default\logging.h 中受影响的代码部分：

// Helper functions for CHECK_OP macro.
// The (int, int) specialization works around the issue that the compiler
// will not instantiate the template version of the function on values of
// unnamed enum type - see comment below.
// The (size_t, int) and (int, size_t) specialization are to handle unsigned
// comparison errors while still being thorough with the comparison.
#define TF_DEFINE_CHECK_OP_IMPL(name, op)                                 \
  template <typename T1, typename T2>                                     \
  inline string* name##Impl(const T1& v1, const T2& v2,                   \
                            const char* exprtext) {                       \
    if (TF_PREDICT_TRUE(v1 op v2))                                        \
      return NULL;                                                        \
    else                                                                  \
      return ::tensorflow::internal::MakeCheckOpString(v1, v2, exprtext); \
  }                                                                       \
  inline string* name##Impl(int v1, int v2, const char* exprtext) {       \
    return name##Impl<int, int>(v1, v2, exprtext);                        \
  }                                                                       \
  inline string* name##Impl(const size_t v1, const int v2,                \
                            const char* exprtext) {                       \
    if (TF_PREDICT_FALSE(v2 < 0)) {                                       \
      return ::tensorflow::internal::MakeCheckOpString(v1, v2, exprtext); \
    }                                                                     \
    const size_t uval = (size_t)((unsigned)v1);                           \
    return name##Impl<size_t, size_t>(uval, v2, exprtext);                \
  }                                                                       \
  inline string* name##Impl(const int v1, const size_t v2,                \
                            const char* exprtext) {                       \
    if (TF_PREDICT_FALSE(v2 >= std::numeric_limits<int>::max())) {        \
      return ::tensorflow::internal::MakeCheckOpString(v1, v2, exprtext); \
    }                                                                     \
    const size_t uval = (size_t)((unsigned)v2);                           \
    return name##Impl<size_t, size_t>(v1, uval, exprtext);                \
  }

// We use the full name Check_EQ, Check_NE, etc. in case the file including
// base/logging.h provides its own #defines for the simpler names EQ, NE, etc.
// This happens if, for example, those are used as token names in a
// yacc grammar.
TF_DEFINE_CHECK_OP_IMPL(Check_EQ, ==)  // Compilation error with CHECK_EQ(NULL, x)?
TF_DEFINE_CHECK_OP_IMPL(Check_NE, !=)  // Use CHECK(x == NULL) instead.
TF_DEFINE_CHECK_OP_IMPL(Check_LE, <=)
TF_DEFINE_CHECK_OP_IMPL(Check_LT, <)
TF_DEFINE_CHECK_OP_IMPL(Check_GE, >=)
TF_DEFINE_CHECK_OP_IMPL(Check_GT, >)
#undef TF_DEFINE_CHECK_OP_IMPL

导致以下错误： “需要一个标识符”在行

TF_DEFINE_CHECK_OP_IMPL(Check_EQ, ==)  // Compilation error with CHECK_EQ(NULL, x)?
TF_DEFINE_CHECK_OP_IMPL(Check_NE, !=)  // Use CHECK(x == NULL) instead.
TF_DEFINE_CHECK_OP_IMPL(Check_LE, <=)
TF_DEFINE_CHECK_OP_IMPL(Check_LT, <)
TF_DEFINE_CHECK_OP_IMPL(Check_GE, >=)
TF_DEFINE_CHECK_OP_IMPL(Check_GT, >)

我没有看到问题。我是 Visual Studio 和 C++-Noob 但这些行应该有效。

我尝试了以下解决方案： Passing operator as a parameter

其中 #define TF_DEFINE_CHECK_OP_IMPL(name, op) 替换为 #define TF_DEFINE_CHECK_OP_IMPL(name, std::function<bool(bool,bool)> op)

但这没有用，我不认为我希望它成为模板。

有什么建议吗？

Answer 1

尝试update bazel 版本，因为 MS2015 可能兼容 bazel 0.20，但 MS2019 不兼容。

Checkout compatibility table

C++：将运算符作为参数传递会导致错误 "expected an identifier"

C++: Passing Operator as Parameter leads to Error "expected an identifier"

c++

tensorflow

visual-studio-2015

compilation

parameter-passing