ElliotAdams0161 · 17 Jan 2026, 09:08 UTC
Replace the blanket success at the documented-query boundary with explicit response cases, then let the real parser interpret them; start with a valid reply, an unsuccessful query and missing required data rather than a pretend model of the whole cell.
23 pointsChanged the fake boundary. Missing data now reaches the parser, which still turns it into ready.
8 pointsWhat happens after one valid ready reading followed by that missing reply?
19 pointsScreen keeps ready. No age change, no failure shown. Found the bug we'd been hiding.
-4 pointsKeep that sequence as a regression case and require the parser to report unavailable data without manufacturing readiness; also test startup without a previous value, so retaining an old good result is not the only failure you catch.
12 pointsParser fixed. Missing-field, failed-query and startup-empty cases show unavailable; good-then-failed shows the retained reading's age and failure. Misspelled fake calls now fail too.
19 pointsGood catch; keep the test result labelled as the software result when this is demonstrated, because a passed test of missing data is not a statement that the fixture holding your reference housing is ready.
15 pointsThis discussion is closed to new replies after six months without activity. Last activity: .