Incorrect Box2f metadata

[markdjwilliams]

The following code demonstrates an issue where writing or reading a Box2f attribute into an EXR's metadata fails:

import OpenEXR
import numpy as np

metadata = {
    "test_box2i": ((0, 0), (1, 1)),
    "test_box2f": ((0.0, 0.0), (1.0, 1.0)),
}

image = OpenEXR.Part(
    header={
        "compression": OpenEXR.ZIP_COMPRESSION,
        "type": OpenEXR.scanlineimage,
        "test_box2f": ((0.0, 0.0), (1.0, 1.0)),
        "test_box2i": ((0, 0), (1, 1)),        
    },
    channels={
        "R": np.zeros((128, 128), dtype=np.float32),
        "G": np.zeros((128, 128), dtype=np.float32),
        "B": np.zeros((128, 128), dtype=np.float32),
    },
)

f = OpenEXR.File([image])
f.write("metadata.exr")

The relevant lines of output from exrheader metadata.exr show that the box2i is written correctly, but the value for the box2f is wrong:

test_box2f (type box2f): (3.40282e+38 3.40282e+38) - (-3.40282e+38 -3.40282e+38)
test_box2i (type box2i): (0 0) - (1 1)

[cary-ilm]

Thanks for the catch, there was a glitch in the test as well as the code. Fixed in #1944.