Enlarging a mmapped region
When enlarging a mmapped region by giving the current pointer as first argument and mmapping on the next byte, if the MAP_FIXED flag was not set, I observed that the 'wish' to keep the same pointer is never fulfilled on my system, meaning that the system can never enlarge the mmapped region without overwriting other things, so the function that enlarges columns always ends up munmapping and mmapping again. If MAP_FIXED is set, the pointer is always kept and everything that was in the new enlarged region is overwritten no matter what, which sounds like a terrible idea (see this discussion http://stackoverflow.com/questions/3521303/is-there-really-no-mremap-in-darwin)
Should we keep trying to directly enlarge anyway (but it could be costly for a result that seems constantly negative)?