Bug 317712

When a view-transition-named element has a CSS transform (e.g. scale(2, 3)) applied, WebKit captures its snapshot at the element's CSS layout size rather than its visual size. The GPU then upscales the bitmap when compositing it through the ::view-transition-group transform, which causes a subtle but measurable rendering quality difference compared to natively rendered content at the same size. Chromium avoids this by capturing at visual resolution. This came up while working on the object-view-box WPT tests. The two failing tests (object-view-box-old-image.html and object-view-box-new-image.html) both have #target with transform: scale(2, 3), and their actual rendering requires fuzzy tolerance (maxDifference=85; totalPixels=600) to pass because of this quality difference. For comparison, content-with-transform-old-image.html covers the same scenario without object-view-box and passes pixel-perfectly even with the fuzzy annotation removed. Additionally, removing the object-view-box property from the two failing object-view-box WPT tests does not eliminate the blur, confirming that the rendering quality difference is caused by the CSS transform on the captured element, not by object-view-box itself. Proposed Patch: The fix could be in snapshotElementVisualOverflowClippedToViewport() in ViewTransition.cpp, accounting for the element's accumulated CSS transforms when computing the capture size.