Add SE050 test

[sosthene-nitrokey]

This PR adds a test that is introduced in the admin app with Nitrokey/admin-app#5

Note that it takes around 30seconds to complete, it's not crashing.

[szszszsz]

I see that this is generally a caller for the in-firmware test, and interpreter of the results gathered there.

1. Why the control of the tests is placed in the firmware instead of the test calling code?
   Was it easier that way? Or is this a quick POC?
2. I think pytest would be better to use here, but its done already.

[sosthene-nitrokey]

Why the control of the tests is placed in the firmware instead of the test calling code?

What do you mean by "the control of the test"?

I think pytest would be better to use here, but its done already.

The goal here is to get info from beta testers that the SE050 driver works as expected with NK3 that are already deployed. If doing so through pytest we likely wouldn't get feedback since users are unlikely to use it.

[szszszsz]

What do you mean by "the control of the test"?

You are essentially calling a fixed test routine in the firmware, and waiting for the results. Instead you could construct test cases on the calling side, and use a proper direct communication channel to se050.

If doing so through pytest we likely wouldn't get feedback since users are unlikely to use it.

That explains things. I do not think though running pytest itself is that much trouble for a beta tester, but that's my opinion.

[sosthene-nitrokey]

You are essentially calling a fixed test routine in the firmware, and waiting for the results. Instead you could construct test cases on the calling side, and use a proper direct communication channel to se050.

I wanted to test the actual APDU generation from the driver to test it, so all commands to the SE050 needs to be run from rust code. This is meant to test the driver in real conditions. Communicating to the SE050 directly from nitropy would make that much less useful, even if it were to test the T=1 over I2C part of the driver.

I do not think though running pytest itself is that much trouble for a beta tester, but that's my opinion.

I agree, but running pytest is much less discoverable while nk3 test is documented and used.

[daringer]

this does not interfere with anything else, should we merge this ?

[robin-nitrokey]

Running this with the se050-test-app feature disabled now gives me:

[4/5]	se050    	SE050                   	FAILURE  	object of type 'type' has no len()

[sosthene-nitrokey]

Running this with the se050-test-app feature disabled now gives me:

[4/5]	se050    	SE050                   	FAILURE  	object of type 'type' has no len()

Can you try again? This should be fixed I believe.

[robin-nitrokey]

Yes, fixed.

[robin-nitrokey]

We could also check for INVALID_COMMAND instead of whitelisting firmware versions.

[sosthene-nitrokey]

Right. We can also update admin-app to return it when the command is disabled instead of INVALID_LENGTH.

[sosthene-nitrokey]

Returning UnsupportedCommand seems to lead to pynitrokey not receiving the error.

[sosthene-nitrokey]

Do you know why that is there: https://github.com/Nitrokey/pynitrokey/blob/master/pynitrokey/nk3/admin_app.py#L72

[robin-nitrokey]

I think the idea was to make it easier to handle the case that the firmware does not (yet) support a command. _call always returns bytes, so None always indicates that the command is not supported. And checking for None is easier than catching the exception and checking the error code.

[sosthene-nitrokey]

Ok, so I can just skip if None.

[sosthene-nitrokey]

Relevant PR in the admin app: Nitrokey/admin-app#12