AIKIDO-2026-194918

TL;DR

OpenEXR's scanline reader decodes ZIP-compressed chunks for channels that use vertical subsampling. When a channel sets ySampling greater than one, the decoder tests sample alignment using the absolute file row but computes the destination compact-buffer row from the chunk-local row, so later sampled rows are written to the wrong offsets and the final sampled row is never written. An application that reads a crafted subsampled EXR into a compact FrameBuffer then exposes stale bytes from its own output buffer and processes pixels from the wrong file rows. The fix aligns the destination row calculation with the absolute file row.