This makes the color matrix and component transfer Moz2D filters generate an infinite output, which is then cropped to the primitive's filter primitive subregion by a subsequent crop filter node. This still gives us different behavior than other browser when the primitive subregion is overridden using the x/y/width/height attributes - other browsers either ignore those completely (IE) or only let them crop the default subregion (which is defined to be the same as the input subregion) and not enlargen it - but I'll fix that in a separate bug.
This introduces some infrastructure to start tracking the program state more carefully. Currently
it only tracks the program because UseProgram was showing up the most in profiles, but more
state could also be added as needed.
--HG--
extra : rebase_source : 289ab867ef63032f7604a6cbe6d0230fd0a9411a
DrawTargets that support region clipping are able to clip to regions effectively. Other DrawTargets go through expensive complex clip paths. When invalidating regions we have code that draws in multiple steps when using Direct2D to avoid this expense, we should base the decision to do this simply on whether the DrawTarget supports fast region clipping.
Calling AcquireSync on textures created with
D3D11_RESOURCE_MISC_SHARED_KEYEDMUTEX appears to be really slow, this really
bites us with the number of textures involved when tiling.
I've also tested using ID3D11Query to check when work is completed, this also
appears to be pretty slow.
This instead just uses a single texture with a keyedmutex, and makes sure we
draw to it last and lock it first. It's pretty hacky, but seems to work really
well so far.
--HG--
extra : rebase_source : bf5f56751f993e507fcd1e5c386fb1fc9a3f73af
For proper hit-testing we need to deal with situations where the initial touch
input falls on a dispatch-to-content region of an APZC, and the main thread
later tells us that it actually hit another APZC. This patch allows input blocks
to have an "unconfirmed" APZC target for the initially-found target that can
then be replaced with a different target. Input blocks without a confirmed
target are not ready for processing.
Previously we output of APZ hit testing was an APZC instance and an "in
overscroll" flag. Now we need to also distinguish between the hit region and the
dispatch-to-content region for an APZC, so this patch introduces an enum that
makes this more explicit.
Until bug 1082594 is fixed, the area occupied by inactive scrollable subframes
don't get added to their layer's dispatch-to-content region. Therefore the APZ
code doesn't know what area corresponds to the subframe, even though it knows
there is a subframe because of the scrollinfo layer. This change inserts a small
workaround to deal with the scrollinfo layers specially, so that the composition
bounds from the scrollinfo layer is used to obtain the area. Once bug 1082594
is fixed this can probably be removed.
When the event-regions are provided by layout, use them to populate the APZ-side
hit testing regions. The existing code is still used when event-regions are not
provided by layout.
This simply swaps out the existing nsIntRegion used in APZ hit-testing for
an EventRegion object. The dispatch-to-content region inside the EventRegions
is kept empty, and there are no functional changes here.
