FrameLayerManager::RecomputeItemsVisibility() was being called on every
call to FrameLayerBuilder::DrawPaintedLayer(), each time for the region
to be painted by that paint call. This is inefficient when progressive
paint is enabled. Change it so that we compute the visibility of all the
layer's items within the total region to be painted, but only on the
first paint after the display list has been modified.
I confirmed that the patch fixes the original testcase (attachment
8634600). I also confirmed that with the whole patch,
layout/style/test/test_descriptor_syntax_errors.html passes, but with
the new tests but not the code change, it reports 12 failures.
The MobileViewportManager ("MVM") is responsible for setting the CSS viewport on
any of the following events:
- a page is painted for the first time (on the before-first-paint event)
- a meta-viewport tag is added (on the DOMMetaAdded event)
- the full-zoom is changed (on the FullZoomChanged event)
- if the window is resized (ResizeReflow gets called as part of normal layout
processing, and this will pick up a new CSS viewport from MVM)
If the CSS viewport changes or if it is the initial paint, the MVM additionally
calls SetResolutionAndScaleTo on the presShell to update the displayed zoom.
The APZ code in AsyncPanZoomController::NotifyLayersUpdated already has
corresponding code to accept this updated zoom when the CSS viewport changes.
The MobileViewportManager ("MVM") is responsible for setting the CSS viewport on
any of the following events:
- a page is painted for the first time (on the before-first-paint event)
- a meta-viewport tag is added (on the DOMMetaAdded event)
- the full-zoom is changed (on the FullZoomChanged event)
- if the window is resized (ResizeReflow gets called as part of normal layout
processing, and this will pick up a new CSS viewport from MVM)
If the CSS viewport changes or if it is the initial paint, the MVM additionally
calls SetResolutionAndScaleTo on the presShell to update the displayed zoom.
The APZ code in AsyncPanZoomController::NotifyLayersUpdated already has
corresponding code to accept this updated zoom when the CSS viewport changes.
There is a common pattern on the web where a click listener is registered on a
container element high up in the DOM tree, and based on the target of the click
events, it performs the appropriate action. In such cases, our existing fluffing
code was not getting activated anywhere inside the container, because the entire
container was considered clickable. However, this is not user-friendly because
often the actual targets inside the container are small and hard to hit. Also,
the fluffing code will often take the container element itself as the target,
even if the user actually hit something inside the container.
This patch changes this behaviour so when an event hits inside a clickable
container, fluffing still occurs, but is restricted to DOM descendants of the
container. This allows fluffing to work in the above scenarios, and since the
events will bubble up to the container, the listeners on the container are
guaranteed to still trigger.
There is a common pattern on the web where a click listener is registered on a
container element high up in the DOM tree, and based on the target of the click
events, it performs the appropriate action. In such cases, our existing fluffing
code was not getting activated anywhere inside the container, because the entire
container was considered clickable. However, this is not user-friendly because
often the actual targets inside the container are small and hard to hit. Also,
the fluffing code will often take the container element itself as the target,
even if the user actually hit something inside the container.
This patch changes this behaviour so when an event hits inside a clickable
container, fluffing still occurs, but is restricted to DOM descendants of the
container. This allows fluffing to work in the above scenarios, and since the
events will bubble up to the container, the listeners on the container are
guaranteed to still trigger.
* * *
Bug 1131451 part 2a - Remove hack for rtl-in-vertical-mode from ReflowAbsoluteFrame. r=dholbert
* * *
Bug 1131451 part 2b - Mark relative-overconstrained tests that now pass in vertical mode with rtl. r=dholbert
* * *
Bug 1131451 part 2c - Mark vertical border-collapse bevel tests that now pass. r=dholbert
* * *
Bug 1131451 part 2d - Remove partial rtl-in-vertical support from nsBidiPresUtils now that logical-coordinate classes handle it better. r=dholbert
* * *
Bug 1131451 part 2e - Remove hack for float positioning in vertical mode with dir=rtl. r=dholbert
* * *
Bug 1131451 part 2f - Mark vertical-mode float-in-rtl reftests that are now passing. r=dholbert
* * *
Bug 1131451 part 2g - Compute both dimensions of containerSize in nsFlexContainerFrame::DoLayout. r=dholbert
* * *
Bug 1131451 part 2h - Mark flexbox writing-mode tests that are now passing. r=dholbert
