The idea is to convert tests with subtests to either use piglit_run_selected_subtests or call piglit_register_subtests at the start of the test. Either approach enumerates all all subtests to stdout in JSON format before running them. The python framework can use this to deduce which subtest crashed.