idiff fails on similar EXRs with different dataWindows about oiio HOT 5 CLOSED

Sorry, Soren, I'm not proud of the fact that this one fell through the cracks.

It looks like those files are no longer on Dropbox, so I can't tell for sure, but by any chance was the larger one BLACK in the top and bottom stripes that weren't present in the image with the smaller data window?

By design, idiff assumes that if the two images don't contain the same number of pixels, they can't possibly "match", since they don't contain the same data.

I can imagine two alternate behaviors for idiff (which could be used in combination):

(1) Allowing a black pixel in one image to "match" a nonexistent pixel in the other image.

(2) Having a mode in which only pixels within the "display" windows of the two images are compared, i.e. they may differ outside the display region and that's ok.

Would either or both of those be more helpful to you in the real world?

from oiio.

I would vote for option 1
On 22/04/2013 11:59 PM, "Larry Gritz" [email protected] wrote:

Sorry, Soren, I'm not proud of the fact that this one fell through the
cracks.

It looks like those files are no longer on Dropbox, so I can't tell for
sure, but by any chance was the larger one BLACK in the top and bottom
stripes that weren't present in the image with the smaller data window?

By design, idiff assumes that if the two images don't contain the same
number of pixels, they can't possibly "match", since they don't contain the
same data.

I can imagine two alternate behaviors for idiff (which could be used in
combination):

(1) Allowing a black pixel in one image to "match" a nonexistent pixel in
the other image.

(2) Having a mode in which only pixels within the "display" windows of the
two images are compared, i.e. they may differ outside the display region
and that's ok.

Would either or both of those be more helpful to you in the real world?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/189#issuecomment-16829325
.

from oiio.

As far as I can tell, although idiff fails comparisons of images with different data window sizes, oiiotool already exhibits the desired behavior:

oiiotool a.exr b.exr --diff

At least as of some recent oiiotool/ImageBufAlgo overhauls.

from oiio.

As I said, the requested behavior was already the case for recent oiiotool versions, and pull request #586 fixes it fully for idiff as well.

from oiio.

Committed the fix to master.

from oiio.