the use of threading.lock() in the performance example

[alexandrecuer]

I can see that the performance example has been reworked
https://github.com/pymodbus-dev/pymodbus/blob/dev/examples/v2.5.3/performance.py

you do not use anymore the threading.lock() in the test, even if the variable is still in the code

pymodbus/examples/v2.5.3/performance.py

Line 34 in ab83cc8

_thread_lock = Lock()

I guess it is because the test covers modbus tcp ?

If testing in modbus rtu,the lock would be necessary ?

Sorry if I am misunderstanding things...

best

Alex

[janiversen]

I am not aware that is has been reworked, that is basically the reason why it is in the v2.5.3 directory. When reworked it will move up to the other updated examples.

In version 3.x we do not use locks in tests.

[alexandrecuer]

Thanks
Can i just ask why it was decided not to use locks ?
In practise in rtu mode, we need to assure at if a thread wants to send a second request, it will wait that the previous one, sent by another thread, has been answered. So yu need a lock or a queue, unless you work on modbus tcp which accepts multiple connections.

[janiversen]

The code was rewritten (not the example) and I am sure when the example gets rewritten it will be without a lock(). That is not the same as to say applications never need to use locks, just that the performance example do not need it.

[alexandrecuer]

by the way, there is a lock in the transaction manager, where the request is processed, if I am correct....

pymodbus/pymodbus/transaction.py

Line 63 in 94fa6b1

self._transaction_lock = RLock()

pymodbus/pymodbus/transaction.py

Line 124 in 94fa6b1

with self._transaction_lock:

used in the base client

pymodbus/pymodbus/client/base.py

Line 134 in 94fa6b1

self.transaction = DictTransactionManager(self, **kwargs)

I need more time to understand things, asyncio is not so easy to decrypt for me

[janiversen]

I was not talking about what we do internally in the code, but what was needed to make performance.py work.