Why jerry_get_value_from_error release error/non-error value in different way?

[lygstate]

trafficstars

https://github.com/jerryscript-project/jerryscript/blob/ba06d492a340bee7d82ff9595851cd278b7bd572/jerry-core/api/jerry.c#L1625-L1642

Seems not consistence

[lygstate]

For implement napi_value to object

inline jerry_value_t AS_JERRY_OBJECT(napi_value nvalue) {
  jerry_value_t val = (jerry_value_t)(uintptr_t) nvalue;
  if (jerry_value_is_error(val)) {
    jerry_value_t err_val = jerry_get_value_from_error(val, false);
    jerry_release_value(err_val);
    return err_val;
  }
  return val;
}

[rerobika]

It's simply about internal reference count maganement. The if the value is not error, and should be released we just don't put reference for it. If the value should not be released, it means it is needed later, so it kept alive with ecma_copy_value .

[galpeter]

Side note on the following code:

    jerry_value_t err_val = jerry_get_value_from_error(val, false);
    jerry_release_value(err_val);
    return err_val;

the release_value here will release the error object which you are returning. This seems incorrect as the err_val at the time of return would not have a reference (it was released)

[lygstate]

Side note on the following code:

    jerry_value_t err_val = jerry_get_value_from_error(val, false);
    jerry_release_value(err_val);
    return err_val;

the release_value here will release the error object which you are returning. This seems incorrect as the err_val at the time of return would not have a reference (it was released)

Yeap, I am confusing about the error object, and didn't got the clue how to using it correctly without memory leak.