Added a NULL check on TransInfo->data before sorting. At that point, the conversion could have been cancel if there's no selected elements (caused segfaults with PET). Another NULL check on TransInfo->data in postTrans fonction, since that is called on cancel on no selection too. Was missing a matrix multiplication when setting TransData->dist on meshes which caused PET on scaled meshes to be bad. (bug report 2395: http://projects.blender.org/tracker/?func=detail&atid=125&aid=2395&group_id=9 )