I filed these two bugs against multi-range read (MRR) in MariaDB 10.0.12 which affected us and could use some review of the patch for correctness, along with development of a proper test case:

MDEV-6878: Use of uninitialized saved_primary_key in Mrr_ordered_index_reader::resume_read()

MDEV-6879: Dereference of NULL primary_file->table in DsMrr_impl::get_disk_sweep_mrr_cost()

Could someone please take a look at them?

Regards,

Jeremy