I've warned before that offline identities don't prove a real interface supplies them. Now my FR5 status-display fake returns true everywhere while the cell is reserved for commissioning. How do I test response parsing meaningfully without claiming physical simulation?
Use explicit methods and arguments matching the documented SDK your application supports, then return identified captured responses where available. Separate invented malformed cases from real captures. Test what the display means after each response, not just whether a call returned something agreeable.
I've selected one displayed-status method and tied the fake to its documented signature. Saved success and failure replies will exercise the parser; constructed bad shapes are labelled separately.
Include a successful read followed by failure with the old display state present. Also make an unknown method or wrong argument fail loudly. An accommodating fake can leave interface mistakes untouched even when its canned results look authentic.
The strict fake caught a wrong argument in our caller. The retained-value test also fails, so fixing the call doesn't finish the display work. Real interface confirmation remains a commissioning task.
Keep both failures as regression cases when repaired. One protects the call contract, the other protects the person reading the screen; a passing parser assertion alone would not cover both.