each iframe.
The old code that attempted to do this during layout wasn't able to work
for multiple reasons: it couldn't know where the iframe was going to be
on the page (because of nested iframes), and at the time it was building
the display list for a fragment it couldn't know where that fragment was
going to be in page coordinates.
This patch rewrites that code so that both the sizes and positions of
iframes are determined by the compositor. Layout layerizes all iframes
and marks the iframe layers with the appropriate pipeline and subpage
IDs so that the compositor can place them correctly. This approach is
similar in spirit to Gecko's `RefLayer` infrastructure. The logic that
determines when it is time to take the screenshot for reftests has been
significantly revamped to deal with this change in delegation of
responsibility.
Additionally, this code removes the infrastructure that sends layout
data back to the layout task to be destroyed, since it is now all
thread-safe and can be destroyed on the script task.
The failing tests now fail because of a pre-existing bug related to
intrinsic heights and borders on inline replaced elements. They happened
to pass before because we never rendered the iframes at all, which meant
they never had a chance to draw the red border the tests expect to not
render!
Closes#7377.
Now that NativeDisplay can be shared between the compositor and the
paint task, we can move the LayerBuffer cache to the compositor. This
allows surfaces to be potentially reused between different paint tasks
and will eventually allow OpenGL contexts to be preserved between
instances of GL rasterization.
The basic idea is it's safe to output an image for reftest by testing:
- That the compositor doesn't have any animations active.
- That the compositor is not waiting on any outstanding paint messages to arrive.
- That the script tasks are "idle" and therefore won't cause reflow.
- This currently means page loaded, onload fired, reftest-wait not active, first reflow triggered.
- It could easily be expanded to handle pending timers etc.
- That the "epoch" that the layout tasks have last laid out after script went idle, is reflected by the compositor in all visible layers for that pipeline.
Some debugging reveals that the send_back_unused_buffers() quite often
sends empty vectors back to the paint task. This still incurs an
communication overhead though. Instead check that the there actually are
buffers to send back.
Instead of cloning pipelines and storing them once per layer, store
them globally in the compositor and access them via id. This trades
lots of unnecessary duplication for a HashMap lookup.
