I haven't checked to be sure, but I suspect that the int64_value function is something that existed before integer values were added to the octave_value data type.  We should probably look at deprecating and removing all the functions that are defined with the INT_CONV_MEDTHOD macro in ov-base.cc short_value, ushort_value, int_value, uint_value, long_value, ulong_value, int64_value, uint64_value.

Consider the following oct file that just tries to parse an int64 argument and show it value:

DEFUN_DLD(show_int64, args, nargout,
          "-*- texinfo -*-\n\
@deftypefn {} {@var{out} =} ping_int64 (@var{in})\n\
@seealso{}\n\
@end deftypefn")
{
  int nargin = args.length ();

  if (nargin != 1)
    print_usage ();

  int64_t out = args(0).xint64_value ("IN should be a scalar int64 val");

  octave_stdout << out << std::endl;

  return ovl ();
}

Here is what I obtain when using this function:

octave:1> a = intmax ("int64")
a = 9223372036854775807
octave:2> show_int64 (a)
-9223372036854775808

See related discussion on the discourse forum:

https://octave.discourse.group/t/how-to-safely-use-int64-type/1096