autorotate - issue 1162

[Hassan-Mallah]

Backend checklist

• Formatted my code by running isort . && black . in app/backend
• Added tests for any new code or added a regression test if fixing a bug
• All tests pass
• Run the backend locally and it works
• Added migrations if there are any database changes, rebased onto develop if necessary for linear migration history

Frontend checklist

• Formatted my code with yarn format && yarn lint --fix
• There are no warnings from yarn lint
• There are no console warnings when running the app
• Added any new components to storybook
• Added tests where relevant
• All tests pass
• Clicked around my changes running locally and it works
• Checked Desktop, Mobile and Tablet screen sizes

[aapeliv]

Are you able to come up with a simple test for this? If it seems very tricky (might be), then no problem

[Hassan-Mallah]

Hey,

Is this what we need? :)
I wasn't able to recreate the error on Mozilla, but i applied autorotate before removing EXIF data.
cheers

[Hassan-Mallah]

Are you able to come up with a simple test for this? If it seems very tricky (might be), then no problem

it looks like the test should be here:
def test_strips_exif(client_with_secrets):
i'll check after work :)

[aapeliv]

The way to go would be to create a small test image (maybe a 3x3 image) with some given EXIF rotation info, and then creating a new test to make sure the image was rotated correctly, etc. I'm just not sure how annoying it is to create such a test image!

[Hassan-Mallah]

hmmm, that's not clear from the issue :)
so we need to autorotate the image? the error is that we rotated the image, so we shouldn't rotate it, right?

[aapeliv]

If I understand correctly; it's that a lot of software stores the image's orientation instead of actual 'moving' the pixels. So when you for example upload a portrait oriented photo from an iPhone it probably assumes the viewer (browser) will rotate it, but we strip that data away. So we need to rotate it ourselves to truly move the pixels in order for it to display correctly in the browser after exif stripping.

So you'd need to make a test image that "displays" correctly in full viewers, but whose data was not 'moved' and just has a roatiton exif tag; and then making sure that image shows up correctly after we save it.

Does that make sense?

[Hassan-Mallah]

Oh!, ya, now it's clear. thnks for the explanation.

[aapeliv]

@Hassan-Mallah: I know you're busy for now, so I'll merge this so that the bug fix can be deployed. You can write a new PR with a test, or just leave it; it's not so important.