The old code fetched QHashData::seed from qt_qhash_seed on every detach. That is both unnecessary and wrong. It is uneccessary, because if the detached-from QHashData isn't shared_null, the seed has already been populated from qt_qhash_seed. It thus suffices to fetch the seed from qt_qhash_seed only when we detach from shared_null. It is wrong, because if qt_qhash_seed was changed between the detach from shared_null and a following detach, d->seed is now different from this->seed, but detach_helper simply clones the buckets 1:1 from this to d, leaving d in a corrupt state. By doing this change, we make QHash robust against on-the-fly changes to qt_qhash_seed (e.g. for testing, or added security). It also opens up the option to have API for changing the seed of a given QHash instance after it has been created (detach, set new seed, rehash). Change-Id: Ib251fc9a6204b42036e97a2fc66f644b379ab841 Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>