# note: The predictor sub-image's green code written with the code-length
#   repeat escapes.
# expect: ok
# exercises: A sub-image's five codes are a full ReadHuffmanCode() each,
#   with nothing special-cased about them. The form the automatic path picks
#   spells all 280 lengths out one by one; this one says the same thing in
#   18 symbols, two of them code 18 skipping 138 and 126 unused entries. 32
#   bytes shorter than predictor-all-16-modes and verified to decode to the
#   same pixels.
lossless
width 64
height 4
transforms predictor
predictor_tiles 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
code green 0x11
code red 0x22
code blue 0x33
code alpha 0x44
subimage predictor
code green complex