Our offline FR10 display likes every response it has ever been given

TobyAli0231 · 19 Aug 2025, 01:47 UTC

Closed
TO
TobyAli0231
The mock behind our sample display returns true for everything, so of course the screen looks healthy while the controller is away. A saved failed read has a completely different shape and our parser doesn't even get through it. I'd like to use captured responses properly. What would make a small offline test set useful without building a pretend controller around it?

12 replies

RA
RachelBarnes0609
Replying to TobyAli0231

Start with that failed read and a matching successful one. Same SDK version and method. Does the screen show failure clearly, or does the parser exception leave the old value there?

18 points
TO
TobyAli0231
Replying to RachelBarnes0609

Old value stays there. It looks current because the little update indicator belongs to the display timer, not the read.

10 points
LU
LucaChen1135
Replying to TobyAli0231

Then test the indicator too; successfully repainting an old value shouldn't give it a new read time.

11 points
PA
PavelBennett0701
Replying to TobyAli0231

Our test fixture once reshaped all saved responses into a common format before handing them to the parser. Very tidy, and it removed the exact awkward response we wanted to test. I'd keep your captures as they arrived, with notes beside them instead

4 points
RA
RachelBarnes0609
Replying to PavelBennett0701

Pavel, was that in the test loader or the application? I can imagine missing it if both use the same helper.

7 points
TO
TobyAli0231
Replying to LucaChen1135

Luca, I've added last successful read time to the proposed display change. Pavel, keeping the raw captures separately; the test will feed them into the same parsing entry point as the real read.

15 points
PA
PavelBennett0701
Replying to RachelBarnes0609

Rachel, the loader. Application parser never saw the original structure. We removed that conversion and found the bug immediately.

7 points
BR
BrunoAdams0104
Replying to TobyAli0231

Include the documented failure forms as well as the captures you happen to have. I would label their source and version separately, because an example constructed from documentation is useful but is not an observed controller response.

6 points
LU
LucaChen1135
Replying to BrunoAdams0104

And have an unexpected response fail visibly, since the small set will never contain every possible bad input.

1 points
RA
RachelBarnes0609
Replying to TobyAli0231

Have you got the stale-value case passing yet? That would be a useful first result to show the people waiting for this screen.

20 points
TO
TobyAli0231
Replying to RachelBarnes0609

Yes. Failed and malformed test responses now show unavailable, with any retained value clearly marked last known and its original read time unchanged. I also found one consumer treating a nonempty failure response as true, which the old mock had rather helpfully concealed.

15 points
RA
RachelBarnes0609
Replying to TobyAli0231

That is worth putting in the handover example. Does reconnect get its own test next?

3 points

Discussion closed

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