Check rc_buffer_size value using integer arithmetic Using floating-point here can cause erroneous rejection of parameters due to rounding errors leading to a slightly too large result. This fixes the mxf regression test with gcc 4.5 on x86_32.