Our UR5e fixture display accepts every fake response

Harriet_Bowen · 10 Feb 2026, 22:17 UTC

Closed
HA
Harriet_Bowen
Offline mock says true to everything. Even invented calls. How do I get a test worth trusting?

20 replies

OM
OmarChen1187
Replying to Harriet_Bowen

Constrain the substitute to your supported interface first, including method names and arguments. Then give it representative successful and failed responses rather than one friendly default.

5points
HA
Harriet_Bowen
Replying to OmarChen1187

Strict calls added. It catches bad names now. Still need the response cases.

19points
MA
MariaBowen
Replying to Harriet_Bowen

What does the operator see when the query succeeds but reports no housing present? That needs to be distinguishable from a failed query. Both might prevent an ordinary next step, but they need different explanations.

20points
IA
Ian_Bailey
Replying to Harriet_Bowen

Build those as separate examples from the documented interface your wrapper uses. Also try missing fields and an unfamiliar state value. Don't let the substitute invent a neat answer that the actual wrapper never receives.

3points
JA
JaneBaker0510
Replying to MariaBowen

Please don't call all those cases empty. Empty is a perfectly understandable physical claim.

19points
IS
IsaacCarter1042
Replying to Harriet_Bowen

Does the existing code check the response contents, or just whether it received something?

9points
HA
Harriet_Bowen
Replying to IsaacCarter1042

Isaac, just whether it received something. Our nonempty failure object becomes housing present. Lovely.

9points
RA
RaviBell0650
Replying to Harriet_Bowen

So the fake hid a parser bug too.

9points
EL
EllaAli0207
Replying to JaneBaker0510

Who changes the misleading screen text, the developer or the operator trainer?

12points
TH
ThomasAli0190
Replying to EllaAli0207

On a station I reviewed, the programmer chose the labels alone and the operator read unavailable as an empty fixture; I'd have both look at the actual cases together.

19points
CL
ClaraAdams0089
Replying to MariaBowen

And show the fixture identity. A correct state for the wrong fixture is still wrong.

16points
HA
Harriet_Bowen
Replying to EllaAli0207

Trainer and developer paired up, Ella. They have failed query, valid empty, valid present and wrong-identity examples.

17points
ER
EricBowen
Replying to Harriet_Bowen

Did they try a failed query after a valid present result? I would like to know whether the previous result stays visible and how they explain its age to the operator.

17points
IA
Ian_Bailey
Replying to EricBowen

Eric, that's worth a sequence test rather than four separate screenshots. Start with a valid reading, fail the next one, then recover. Check both identity and displayed validity at each transition.

25points
OM
OmarChen1187
Replying to Harriet_Bowen

Also keep the bad-method test. A screen fix should not loosen the interface substitute again.

18points
JA
JaneBaker0510
Replying to Harriet_Bowen

What words did the trainer actually choose?

19points
HA
Harriet_Bowen
Replying to JaneBaker0510

Valid readings say present or empty. Failed or mismatched readings say status unavailable, with the last valid reading separately labelled.

14points
MA
MariaBowen
Replying to Harriet_Bowen

That sounds clearer. Can the trainer distinguish those cases without the developer announcing which response is about to arrive? Otherwise you are testing whether they can repeat the explanation.

-2points
HA
Harriet_Bowen
Replying to MariaBowen

Yes. Mixed sequence passed without hints, including failure after present and later recovery. Parser tests pass too. Thanks, Maria.

5points
RA
RaviBell0650
Replying to Harriet_Bowen

A less agreeable mock earned its keep.

21points

Discussion closed

This discussion is closed to new replies after six months without activity. Last activity: .