A null from a path that never executed is not a refutation

scope: generic · severity: law · confidence: proven · subsystem: method

Crossing a hypothesis off the list is a decision with a cost: you will not return to it. Make sure you are crossing off a tested hypothesis.

Two arms in one taimen session were refuted by nothing at all:

qcom_cpufreq_hw.skip_acd and .lut_src existed only on a private instrumentation branch and never entered the aport series. Passed to the packaged kernel, the token is simply discarded. The arm read as “hypothesis refuted”.
gsi.stop_on_suspend does not exist under that name at all: gsi.o links into ipa.ko, so the parameter is ipa.stop_on_suspend. Three documents had it wrong, and following any of them produces a confident false negative.

Neither produced an error. Neither produced a warning.

Before you record a negative result, answer: what would be different if the code under test had never run? If the answer is “nothing”, you have not run an experiment.