It turns out that calling HasBorder() is especially expensive for themed frames since we call into the theme engine to compute the border, so avoiding it is a nice win.
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.
When rendering just the current Selection (Print - Selection) then don't create display items
for table-related frames unless the frame itself is part of the selection, and always ask
descendant frames to build display lists [in case they are selected].
This is the first of two patches to honor inflation during intrinsic
width calculation (which we need to do to make some form controls
inflate correctly).
Add an extra change hint, UpdateOverflow, that can be used to specify that
a frame's overflow areas may have changed and that they need to be recalculated.
When a transform on a frame changes, instead of marking it for reflow, set this
hint instead.
There is an added virtual function on nsIFrame, UpdateOverflow, which is called
recursively on a frame when the corresponding hint is set, to allow it to
update its overflow areas.
Outer table frames act as CSS2.1 table wrapper boxes. We used to lay them out
without taking their margins into the account, which meant that their width was
always equal to the available width. This breaks horizontal positioning of
absolutely positioned kids of a table frame.
The main purpose of this patch is to apply the margins of tables to their outer
frame, instead of the inner frame. This means that the inner table frame will
always have a zero margin, which means that a lot of the stuff which used to
rely on the fact that table margins are applied to the inner frame need to
change.
In particular, in order to get the computed margins of a table, we used to query
the inner table frame, and this patch corrects that. Also, when shrink wrapping
tables, we used to not take the margins of the inner table frame into account,
which is fixed by this patch too. nsBlockReflowState::
ComputeReplacedBlockOffsetsForFloats also needed to be changed to read the
margin values from the outer frame too.
Also, as part of this patch, we start to respect the CSS2.1 margin model for
captions on all sides. This means that in particular, the top/bottom margins on
the top-outside and bottom-outside captions will not be collapsed with the
top/bottom margins of the table, and that the margins of the caption element
contribute to the width and height of the outer table frame. The
427129-table-caption reftest has been modified to match this new behavior.
Another side effect of this bug is fixing bug 87277, and the reftests for that
bug are marked as passing in this patch.
