A skipped step is not a pass
Why a test run with an unexecuted step must fail, and how quietly reporting unknowns as passes is the exact failure a test tool exists to prevent.
Every test executive has to answer one uncomfortable question: what is the verdict of a step that never ran? A sequence aborts at step 7 of 20. Steps 8 through 20 produced no measurements, judged no limits, exercised nothing. What goes in the report?
The tempting answer is to leave them blank, or grey, or 'skipped' — and then to summarise the run by the steps that did execute. Six passed, one failed, the rest didn't happen. On a busy bench, three weeks later, that summary line is all anyone reads. And a summary that says 'passed 6 of 7 executed' has a way of becoming 'mostly passing' in a slide, and 'passing' in a decision.
NOKTRA Sequencer takes the strict position: if a step never ran, the run failed. Not because the unit under test did anything wrong, but because the run does not know whether it did. A test report is a claim about what was verified. A skipped step is precisely the absence of verification, and reporting an absence as anything adjacent to success is how bad units ship with clean paperwork.
This sounds obvious written down. It is not the default in practice, because strictness is annoying. An operator who aborts a run to fix a cable does not want a red FAIL over something that 'wasn't a real failure'. The pressure to soften the verdict is constant and reasonable-sounding, and it must lose every time, because the report outlives the context. The paper does not remember that the abort was innocent.
There is a second design consequence. If unexecuted means failed, then teardown must still run — power-down, discharge, disconnect — even in a failing run, and the report must show that it ran. Failure handling stops being an edge case and becomes part of the sequence contract: what aborts, what is skipped, what always executes.
The verdict logic itself stays deliberately dumb: pass and fail come from a pure function of the rule and the measurement. No clock, no I/O, no state. Two runs over the same readings cannot disagree. All of the judgement lives in one place a reviewer can read, and none of it lives in the moment when someone is tired and wants the run to be over.