Hi, This was recently brought to my attention.
I would point out that neither of these are actually bugs. First, the unconditional fsync() was intentional behavior. At worst, it's a NOP as the file will likely be flushed momentarily anyway. Secondly, the gdbm_close() function is meant to approximate the behavior of the dbm_close() function, which is specified in the standards as a void function. Hence, there's no reason to check return values.
