Can someone provide a brief clarification about build reproducibility in general?<p>The stated aim is that when you compile the same source, environment, and instructions the end result is bit identical.<p>There is, however; hardware specific optimizations that will naturally negate this stated aim, and I don't see how there's any way to avoid throwing out the baby with the bathwater.<p>I understand why having a reproducible build is needed on a lot of fronts, but the stated requirements don't seem to be in line with the realities.<p>At its most basic, there is hardware, where the hardware may advertise features it doesn't have, or doesn't perform the same instructions in the same way, and other nuances that break determinism as a property, and that naturally taints the entire stack since computers rely heavily on emergent design.<p>This is often hidden in layers of abstraction and/or may be separated into pieces that are architecture dependent vs independent (freestanding), but it remains there.<p>Most if not all of the beneficial properties of reproducible builds rely on the environment being limited to a deterministic scope, and the reality is manufacturers ensure these things remain in a stochastic scope.