Towards a Benchmark Set for Program Repair Based on Partial Fixes