ResourceManager needs to be closed to clear instrument buffer

[bilderbuchi]

I observe weird problems running pytest tests with pyvisa-sim backed fixture objects, in that replies remain in a read buffer despite the objects + resource managers being destroyed:

>>> import pyvisa
>>> rm = pyvisa.ResourceManager(visa_library='@sim')
>>> instr = rm.open_resource('ASRL2::INSTR')
>>> instr.write('*IDN?')
7
>>> instr.read()
'SCPI,MOCK,VERSION_1.0\n'  # OK, that's what I expect
>>> instr.write('*IDN?')  # we don't fetch that response from the instrument
7
>>> id(instr)
140011221241280
>>> del instr  # Kill the instrument (or instr.close())
>>> del rm  # (or rm.close())
>>> rm = pyvisa.ResourceManager(visa_library='@sim')
>>> instr = rm.open_resource('ASRL2::INSTR')
>>> instr.read()  # Why is this response in this different instrument object?
'SCPI,MOCK,VERSION_1.0\n'
>>> id(instr)
140011221241472  # the IDs of the instr objects are different, so it's not somehow the same

To fix/avoid this, one needs to call rm.close(). instr.close() is neither sufficient nor necessary.

Matthieu said:

The fact that closing the instrument does not solve the issue still says that something is off. Instruments do not share outgoing buffers between connections and we should mimic that.

Originally posted by @bilderbuchi in #68 (comment)