TL;DR: is it possible with pytest to create an instance of an object and then run a battery of tests on the same instance ?
I have a custom C API, that uses posix threads and does operations on serial ports. My API is an implementation of a custom/proprietary protocol with a master/slave messaging. With simple test programs in C (or in Python), it works correctly.
But, with pytest, when testing sending and reception of messages, pytest repeatdly open/close the COM port and create/delete the background thread, as each unit test is written in the following manner:
def test_send_message_XXX(serial1, serial2):
master = my_api.Master(serial1)
slave = my_api.Slave(serial2)
message = my_api.SomeCustomMessage()
master.send_message(message)
## tempo to be sure the message was sent
time.sleep(0.5)
rec = slave.receive()
slave.close()
## master is closed when the object is destroyed
assert rec == message.data
The problem is that, most of the time, when I run the tests, I have either a pytest crash (ending with a segfault), or some random cases where the COM port cannot be opened, or is opened but read throws a bad file descriptor (running ls /dev shows the serial devices, and some other "lower level" tests are running correctly).
When I use this API in a "production" (simulated) context, these errors never shows up.
Is there a way to initialize master and slave only once, and then run all the tests on the same instances ?
You can use Fixtures or set-up/tear-down fixtures