On Thu, Dec 20, 2012 at 1:12 PM, Nicholas Clark <nick@ccl4.org> wrote: > SvLEN == 0 has a historical established meaning of "the core doesn't own > this buffer". If COW were disabled for all PVs with SvLEN == 0, does this > solve the issues in more than just this module? That is already taken into account in sv_setsv, so I guess something is happening somewhere else. I think SvCANCOW needs to be improved to handle this. Leon