The first draft got this half right and half wrong. It correctly said the
existing histogram and clipping indicators are not FR-CULL-3's, but it
described them as adjacent — as though the work left were mostly focus
peaking and a relocation.
They are not adjacent. The histogram reads AdjustPass's 8-bit output and
counts clipping as r == 255, so it describes the frame the display is
about to show, after the entire develop chain. FR-CULL-3 asks for the
histogram of the sensor data, and gives its reason in the requirement
itself: a rendered image "systematically lies about what is recoverable in
the raw". A readout taken from the render cannot answer that question
however it is presented, which means two of the three bullets need a new
measurement rather than a new placement.
Found by the agent building focus peaking, who had to go looking at the
counters to find out.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
dtourolleandClaude Opus 5
8d5dc423ea