[41]:... unit tests for sender and receiver errors

[Darrellbor]

PR for #41
Implemented

• TestReceiver_Bind
• TestReceiver_ReadFloats
• TestSender_Connect
• TestSender_WriteFloats

[Darrellbor]

@gavv @ortex PR ready for review

[coveralls]

Coverage: 85.822% (+8.7%) from 77.127% when pulling 8952369 on Darrellbor:41-unit_tests_for_sender_and_receiver_errors into ee4158d on roc-streaming:main.

[gavv]

Also @Asalle

[ortex]

Nice work!

a couple of tests could be added:

• for Sender.Connect, Receiver.Bind: add tests with an invalid endpoint that fails to convert to C endpoint, e.g. with invalid protocol Protocol(1)
• for Sender.Writefloat, Receiver.ReadFloats: add tests with zero-sized frame []float32{}

also, point for discussion (@gavv ):

Maybe add test Receiver_Closed (or something like that) and remove receiverClosedBefore scenario from all tests?
Because we need to check that each function call (except Сlose()) on a closed receiver return an error
e.g:

func TestReceiver_Closed(t *testing.T) {
	tests := []struct {
		name      string
		operation func(receiver *Receiver) error
	}{
		{
			name: "SetMulticastGroup after close",
			operation: func(receiver *Receiver) error {
				return receiver.SetMulticastGroup(SlotDefault, InterfaceAudioSource, "127.0.0.1")
			},
		},
		{
			name: "Bind after close",
			operation: func(receiver *Receiver) error {
				return receiver.Bind(SlotDefault, InterfaceAudioSource, nil)
			},
		},
		{
			name: "ReadFloats after close",
			operation: func(receiver *Receiver) error {
				recFloats := make([]float32, 2)
				return receiver.ReadFloats(recFloats)
			},
		},
	}
	for _, tt := range tests {
		ctx, err := OpenContext(ContextConfig{})
		require.NoError(t, err)
		require.NotNil(t, ctx)

		receiver, err := OpenReceiver(ctx, makeReceiverConfig())
		require.NoError(t, err)
		require.NotNil(t, receiver)

		err = receiver.Close()
		require.NoError(t, err)

		require.Equal(t, errors.New("receiver is closed"), tt.operation(receiver))

		err = ctx.Close()
		require.NoError(t, err)
	}
}

[gavv]

I agree with Andrey, separating this check will make the table tests more uniform, which is the point of tables. Ideally table tests should have minimal amount of conditions and if tests have rather different flow, they should belong to different tables / tests.

[Darrellbor]

@gavv @ortex
reviewed based on the comments and ready for re-review

[gavv]

Thanks!