Bug 203355 – EXIF orientation is ignored for some CSS images

• Component: CSS
• Status: RESOLVED FIXED

[Simon Fraser (smfr)]

Created attachment 381775 [details]
Testcase

We ignore EXIF orientation when hitting the rendering codepaths that go through GraphicsContext::drawTiledImage(), so fail to respect EXIF orientation for CSS images in tiled background-image, border-image etc.

[Radar WebKit Bug Importer]

<rdar://problem/56568303>

[Said Abou-Hallawa]

Created attachment 387973 [details]
Patch

[Said Abou-Hallawa]

Created attachment 388093 [details]
Patch

[Said Abou-Hallawa]

(In reply to Said Abou-Hallawa from comment #3)
> Created attachment 388093 [details]
> Patch

In this patch the EXIF orientation is respected with css background images and css border images.

[Said Abou-Hallawa]

Created attachment 388097 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 388097 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=388097&action=review

> Source/WebCore/platform/graphics/BitmapImage.cpp:162
> +    FloatRect rect = { IntPoint::zero(), size() };

It's a bit weird to initialize the location of a FloatRect with IntPoint::zero

Is size() computed respecting orientation? I.e. do we know this code won't distort the image?

[Said Abou-Hallawa]

Comment on attachment 388097 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=388097&action=review

>> Source/WebCore/platform/graphics/BitmapImage.cpp:162
>> +    FloatRect rect = { IntPoint::zero(), size() };
> 
> It's a bit weird to initialize the location of a FloatRect with IntPoint::zero
> 
> Is size() computed respecting orientation? I.e. do we know this code won't distort the image?

I will replace the IntPoint::zero() by FloatPoint().

Yes size() is computed respecting orientation since it takes an ImageOrientation argument which has a default value 'FromImage'. The definition of it in Image.h is

virtual FloatSize size(ImageOrientation = ImageOrientation::FromImage) const = 0;

[Said Abou-Hallawa]

Created attachment 388116 [details]
Patch

[Said Abou-Hallawa]

Created attachment 388248 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 388248 [details]
Patch

Clearing flags on attachment: 388248

Committed r254841: <https://trac.webkit.org/changeset/254841>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Jonathan Bedard]

Test was landed broken:
https://results.webkit.org/?suite=layout-tests&test=scrollingcoordinator%2Fmac%2Ffixed-scrolled-body.html

[Jonathan Bedard]

Oops, wrong tests: https://results.webkit.org/?suite=layout-tests&suite=layout-tests&test=fast%2Fimages%2Fexif-orientation-border-image.html&test=fast%2Fimages%2Fexif-orientation-background-image-no-repeat.html

[Said Abou-Hallawa]

(In reply to Jonathan Bedard from comment #13)
> Oops, wrong tests:
> https://results.webkit.org/?suite=layout-tests&suite=layout-
> tests&test=fast%2Fimages%2Fexif-orientation-border-image.
> html&test=fast%2Fimages%2Fexif-orientation-background-image-no-repeat.html

Yes. The non CG ports have to skip these tests or to respect the EXIF image orientation when drawing the native image for CSS background and border.