One role per agent does not work

[Mofo50C]

Describe the bug The backend option --one-role-per-agent does nothing.

To Reproduce Steps to reproduce the behavior:

1. Open GUI
2. Go to settings
3. Paste in --one-role-per-agent for backend arguments
4. Run verify on protocol
5. Attacks still show agents assuming multiple roles

Expected behavior Use this protocol and verify the claim(I, Commit, R, Kir) https://gist.github.com/Mofo50C/31246287024580ab24283c559f7887fd

Scyther version used:

• backend version: v1.2-25-g2a698fa
• GUI version is compiled from master, but it doesn't show which version

Platform information (please complete the following information):

• OS: Ubuntu 20.04 running on WSL2 on Windows 10
• Running python 3.8 and wxPython 4.2.1

[cascremers]

Please add a picture here of the attacks. (When I try this, none of the generated attacks contain an agent that executes multiple roles.)

[Mofo50C]

Hi there, this is what i get for I's claim on commit R

[Mofo50C]

I've tried it with option "--one-role-per-agent"; in the actual protocol and --one-role-per-agent in the backend arguments. With 5 runs and best attack.

[cascremers]

In your image, there are two agents, and they each perform a different role. This is matches the requirements for the "one role per agent" switch.

Perhaps your concern is that Alice (when executing the I role), upon receiving the message, assumes that Alice is also performing the R role -- but this is only a variable instantiation, and does not imply Alice plays both roles. If the protocol has no explicit checks to exclude this behavior, then this is typically possible.

In any case, I think this is not a bug report, but rather a misunderstanding of the "one role per agent" restriction, which is met here.

[Mofo50C]

Oh I see, after playing around with the settings, i came across the following attack that is what i thought would be shown with the "one role per agent" argument.