With APZ, we always layerize the first scrollable element of the page,
if the page itself is not scrollable. These additional layers can cause
fuzzy reftest failures in two ways: differences in blending, and by
disabling sub-pixel test anti-aliasing. The latter is only a problem
with unaccelerated drawing, because we don't support component alpha
layers with BasicLayers. (We also don't support them with
BasicCompositor, but "Reftest unaccelerated" only tests BasicLayers).
Prior to this patch, we failed to honor:
* margin-left on elements in inline layout with 0 width and 0 height
* margin-right on elements in inline layout with 0 width
I think that was because the code in CanPlaceFrame to discard both
margins when the width was 0 was running after the left-margin was
applied, unless the later code in PlaceFrame (checking both width 0 and
height 0) un-applied that left margin.
The assertion count change in test_value_computation.html is due to 2
additional "bad width" assertions (I presume from honoring large
margins that were previously ignored).
The change to 538935-1-ref.html is to match an improvement in rendering
of the margins in the test, where both sides of the margin are now
honored.
The change to layout/reftests/text-overflow/marker-basic-ref.html is to
keep the reference (which uses margins) rendering the same way following
the changes to margin handling.
The new behavior (in the reftests added in layout/reftests/inline/)
matches at least Chromium; I didn't check any other browsers.
The dynamic toolbar enlarges the browser slightly, which causes extra fuzzing
to be required on dynamic-clipPath-01.svg and scroll-rounding layout tests,
but also makes translate-pattern-1.sg pass.
Previously we snapped the results of nsDisplayItem::GetBounds and
nsDisplayItem::GetOpaqueRegion internally. By tracking which display items were
inside transforms, we disabled snapping quite conservatively whenever an ancestor
had a transform, which is undesirable.
With this patch, we don't snap inside GetBounds or GetOpaqueRegion, but just return
a boolean flag indicating whether the item will draw with snapping or not. This flag
is conservative so that "true" means we will snap (if the graphics context has a transform
that allows snapping), but "false" means we might or might not snap (so it's always safe
to return false).
FrameLayerBuilder takes over responsibility for snapping item bounds. When it converts
display item bounds to layer pixel coordinates, it checks the snap flag returned from
the display item and checks whether the transform when we draw into the layer will be
a known scale (the ContainerParameters scale factors) plus integer translation. If both
are true, we snap the item bounds when converting to layer pixel coordinates. With
this approach, we can snap item bounds even when the items have ancestors with active
transforms.
