In the definition of function sortIG of kpoints.py and in the call to it in __init_espresso of __kpoints.py
In the comparison of t(table) & t(spglib) of _get_irreps_table, I set a default tolerance 1e-6, rather than 1e-8, which has been problematic in a particular QE calculation.
Found 2 bugs:
Also, I added the example of Bi HOTI.