some form of isolation when dealing with untrusted data. As a result, this
document also outlines what issues we consider as TensorFlow security
vulnerabilities.

We recognize issues as vulnerabilities only when they occur in scenarios that we
outline as safe; issues that have a security impact only when TensorFlow is used
in a discouraged way (e.g. running untrusted models or checkpoints, data parsing

typedef struct TF_TransactionToken {
  void* token;
  TF_Filesystem* owner;
} TF_TransactionToken;

// The named union is needed here (as opposed to
// inside the `TF_Filesystem_Option_Value` struct)
// as MSVC does not recognize `typeof`.
typedef union TF_Filesystem_Option_Value_Union {
  int64_t int_val;
  double real_val;
  struct {
    char* buf;
    int buf_length;
  } buffer_val;
} TF_Filesystem_Option_Value_Union;

        `tf.keras.optimizer.legacy.XXX`. If you want to migrate to the new
        optimizer and find it does not support your optimizer, please file
        an issue in the Keras GitHub repo.
    *   **Error such as `Cannot recognize variable...`.** The new optimizer
        requires all optimizer variables to be created at the first
        `apply_gradients()` or `minimize()` call. If your workflow calls