Fix fallout from r366811.
PR: 250442
How can provider be NULL there, if this function either called by the provider or by g_dev_taste() where you are checking for attach error?
This part is OK, I am just thinking whether we should print this error, since as I see you are not doing it for other classes.
Drop unnecessary chunk and add another one, fixing error handling in g_uzip(4).
I'd prefer to leave it for now, in case this triggers in some unexpected cases.