GitLab

http://whatwg.org/html#texttrackcue

TEST=LayoutTests/media/track/opera/interfaces/TextTrackCue/track.html

Review URL: https://codereview.chromium.org/356903006

git-svn-id: svn://svn.chromium.org/blink/trunk@177168 bbb929c8-8fbe-4397-9dbb-9b2b20218538

https://codereview.chromium.org/60013004 started this but must have missed these.

Review URL: https://codereview.chromium.org/359023002

git-svn-id: svn://svn.chromium.org/blink/trunk@177167 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This fixes a WinXP bot failure after https://codereview.chromium.org/332633006.
This case was presumably missed because WinXP is not in the CQ bot set.

TBR=sigbjornf@opera.com
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/355313002

git-svn-id: svn://svn.chromium.org/blink/trunk@177166 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This CL reduces the number of hit tests in trivial tap scenarios from 15
to 9.  Later CLs will push GestureEventWithHitTestResults up to
WebViewImpl, down to synthetic mouse events, and attempt to share hit
test results across some event types (further reducing from 9 down to 2
or 3).

GestureEvents fall into two categories that are handled very differently
today for hit-testing purposes: scroll events and non-scroll events.
Scroll events do per-frame hit-tests and propagate across EventHandler
instances from the outer-most frame to the inner-most (because they may
do some work for scroll bubbling at each frame).  Non-scroll events do a
global hit-test and jump directly to the inner most frame.  Make this
confusing distinction more explicit.

Then consolidate non-scroll GestureEvent hit-tests in blink core into a
single method: EventHandler::targetGestureEvent which returns a new
GestureEventWithHitTestResults type.  Pass these "targeted events"
around blink instead of doing repeated hit tests.

Eliminate duplicate frame tree walk on most gesture hit tests.
Previously we'd do an initial hit test to find the right subframe and
EventHandler instance, then use EventHandler::hitTestResultAtPoint to
hit test from the main frame again all the way down.

This CL should not have any noticeable web-exposed behavior changes, but
it paves the road for future optimizations which will.

Mutate HitTestResult to make it reflect the results of touch adjustment
(effectively converting a rect-based test result to a point-based result
somewhere inside that rect).  This makes it possible to use the same
code for gestures that do and don't do adjustment, avoids passing around
extra data, and makes bugs with using unadjusted points less likely.
Update touch- adjustment entry points to take a HitTestResult instead of
a point (which we'd then need to hit-test against again).  Touch
adjustment tests still need to do hit tests here though.

Remove unused 'baseEventType' hack (synthetic mouse events have carried
a 'from touch' bit for awhile now that probably supercedes whatever this
was being used for).

Fix a bug with resize handle behavior in iframes not being consistent
with other resize handle behavior.

Includes a trivial tweak to FROM_HERE macro to be explicit about
namespace usage since it can be used both within and outside of WebCore
namespace contexts.  This makes eclipse happy.

Fix WebViewImpl::handleInputEvent trace logging of event name.

BUG=381728

Review URL: https://codereview.chromium.org/338543003

git-svn-id: svn://svn.chromium.org/blink/trunk@177164 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=177154

TBR=jchaffraix@chromium.org

Review URL: https://codereview.chromium.org/338983005

git-svn-id: svn://svn.chromium.org/blink/trunk@177162 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=389728

Review URL: https://codereview.chromium.org/360583002

git-svn-id: svn://svn.chromium.org/blink/trunk@177161 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=381876
TBR=tkent for web

Review URL: https://codereview.chromium.org/355193003

git-svn-id: svn://svn.chromium.org/blink/trunk@177160 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This is needed to build the generated bindings.

R=brettw@chromium.org

Review URL: https://codereview.chromium.org/359693002

git-svn-id: svn://svn.chromium.org/blink/trunk@177159 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This lets us remove code elsewhere that did the same thing, but potentially incorrectly and with unnecessary knowledge of squashing.

BUG=384520

Review URL: https://codereview.chromium.org/354223002

git-svn-id: svn://svn.chromium.org/blink/trunk@177158 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Instead use the same paths as other compositing updates.


BUG=384520

Review URL: https://codereview.chromium.org/358153003

git-svn-id: svn://svn.chromium.org/blink/trunk@177157 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This code doesn't many any sense. I also added a RELEASE_ASSERT that the field
already had this value and every test passed. That means this code doesn't do
anything.

R=eseidel@chromium.org

Review URL: https://codereview.chromium.org/345463010

git-svn-id: svn://svn.chromium.org/blink/trunk@177156 bbb929c8-8fbe-4397-9dbb-9b2b20218538

There's no reason we can't use the clip cache when there's an enclosing
pagination layer. None of the rest of the code in RenderLayerClipper cares
about paginations layers.

R=eseidel@chromium.org

Review URL: https://codereview.chromium.org/355293002

git-svn-id: svn://svn.chromium.org/blink/trunk@177155 bbb929c8-8fbe-4397-9dbb-9b2b20218538

We forgot to push LayoutState for inlines which meant that
we didn't invalidate correctly relatively positioned inlines
as the offset was wrong. This was visible in
fast/repaint/inline-relative-positioned.html where we invalidated
at the wrong position (and ultimately for the wrong reasons).

Review URL: https://codereview.chromium.org/358033003

git-svn-id: svn://svn.chromium.org/blink/trunk@177154 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=177136

BUG=379147
TBR=senorblanco@chromium.org

Review URL: https://codereview.chromium.org/359073002

git-svn-id: svn://svn.chromium.org/blink/trunk@177153 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch fixes a use after free during preferred logical width
calculation when we have autosizing enabled. This is a speculative
local fix to see if more of these regressions pop up; the final fix
will need to address crbug.com/389648 as well.

BUG=385268

Review URL: https://codereview.chromium.org/358163002

git-svn-id: svn://svn.chromium.org/blink/trunk@177152 bbb929c8-8fbe-4397-9dbb-9b2b20218538

R=eseidel@chromium.org
NOTRY=true

Review URL: https://codereview.chromium.org/355213003

git-svn-id: svn://svn.chromium.org/blink/trunk@177150 bbb929c8-8fbe-4397-9dbb-9b2b20218538

All entrypoints in SVGTextQuery start with check if the text content
element has any boxes. If it doesn't, a value representing "empty" is
returned. In all cases this value is the same as the default value of the
corresponding query-data. SVGTextQuery::executeQuery itself has no
problems handling this case (and will return 'false' as expected.)
Removing these early-outs means a few more cycles is needed in these
"degenerate" cases, although that seems like a small (and hopefully
reasonable) price to pay for less code (and duplication/exit-paths).

Review URL: https://codereview.chromium.org/358033005

git-svn-id: svn://svn.chromium.org/blink/trunk@177149 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Perform an intersection of the involved ranges to reject fragments early,
while still avoiding calls to mapStartEndPositionsIntoFragmentCoordinates.
This means the "cache" added in https://codereview.chromium.org/349223003
can be dropped, since we'll now never even reach it. This gives a moderate
speed-up - a few percent - on the original TC compared to using the cache.

BUG=267504

Review URL: https://codereview.chromium.org/359783002

git-svn-id: svn://svn.chromium.org/blink/trunk@177148 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Jim.

R=abarth@chromium.org
BUG=389681

Review URL: https://codereview.chromium.org/356273002

git-svn-id: svn://svn.chromium.org/blink/trunk@177147 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Adam says this is no longer used.

BUG=389681
R=abarth@chromium.org, abarth, dpranke

Review URL: https://codereview.chromium.org/360553005

git-svn-id: svn://svn.chromium.org/blink/trunk@177146 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This testing framework is horrible.  When you have a syntax error
you have to go digging in the console to fine it, it doesn't notice.

BUG=389681
R=abarth@chromium.org, abarth

Review URL: https://codereview.chromium.org/359813002

git-svn-id: svn://svn.chromium.org/blink/trunk@177145 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This CL deletes RenderLayer::updateLayerPositionsAfterDocumentScroll and
unwinds all the complexity it created. Now that we're always passing
IsOverflowScroll, we can remove all the other flags for deciding when to clear
the clip cache. Also, now we always start the walk at RenderLayers that would
trigger HasSeenAncestorWithOverflowClip, so we can just assume that flag is
set.  That means there are no flags left!

Also, finally delete RenderLayerClipper::clearClipRects. Now everyone clears
all the descendant clip rects too.

Review URL: https://codereview.chromium.org/331653004

git-svn-id: svn://svn.chromium.org/blink/trunk@177144 bbb929c8-8fbe-4397-9dbb-9b2b20218538

If a replaced element has a percentage height specified then its height should
compute to 'auto' when its containing block does not have a height 'specified
explicitly'. We were taking this to mean when its containing block's specified
height value is 'auto' - in fact it means when the containing block's computed
height is auto. This brings us into line with IE and Firefox.

http://lists.w3.org/Archives/Public/www-style/2014Jun/0079.html

BUG=385877

Review URL: https://codereview.chromium.org/347053007

git-svn-id: svn://svn.chromium.org/blink/trunk@177143 bbb929c8-8fbe-4397-9dbb-9b2b20218538

On linux use_webaudio_ffmpeg is true and when it is platform needs to
depend on //third_party/ffmpeg. This target is set up in the GN build now
(at least for this platform), so it's safe to re-enable.

R=brettw@chromium.org

Review URL: https://codereview.chromium.org/353943003

git-svn-id: svn://svn.chromium.org/blink/trunk@177142 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The server version has won. Any features missing
from the server version should be added there.

BUG=389681
R=abarth@chromium.org, abarth, dpranke

Review URL: https://codereview.chromium.org/351873005

git-svn-id: svn://svn.chromium.org/blink/trunk@177141 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Some folks still probably run GOM locally, but
it's not worth the burden on the code to keep this around.
If we want this we should re-build it to be web only.

R=abarth@chromium.org, abarth, dpranke
BUG=389681

Review URL: https://codereview.chromium.org/355193005

git-svn-id: svn://svn.chromium.org/blink/trunk@177140 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=177119

BUG=381683
TBR=jkarlin@chromium.org

Review URL: https://codereview.chromium.org/339613004

git-svn-id: svn://svn.chromium.org/blink/trunk@177139 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=177023

TBR=fmalita@chromium.org

Review URL: https://codereview.chromium.org/353213002

git-svn-id: svn://svn.chromium.org/blink/trunk@177138 bbb929c8-8fbe-4397-9dbb-9b2b20218538

It's dead.  The auto rebaseline bot won.

BUG=389681
R=abarth@chromium.org, abarth, dpranke

Review URL: https://codereview.chromium.org/358173003

git-svn-id: svn://svn.chromium.org/blink/trunk@177137 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Skia change: https://skia.googlesource.com/skia/+/bccac6cbf128147587bfc2bc1bdeb46c08175bca
Rolled into Chrome at r278948.

TBR=fmalita
BUG=379147

Review URL: https://codereview.chromium.org/355203004

git-svn-id: svn://svn.chromium.org/blink/trunk@177136 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Don't include trailing space in linewidth calculations when deciding
if a float can fit on a line.

Merge http://trac.webkit.org/changeset/148622

Review URL: https://codereview.chromium.org/343663004

git-svn-id: svn://svn.chromium.org/blink/trunk@177135 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Not sure why the tests changed from expecting '' to undefined,
but I doubt it matters.  Useful to be able to start from
all tests passing.

TBR=ojan@chromium.org
BUG=389681

Review URL: https://codereview.chromium.org/353183003

git-svn-id: svn://svn.chromium.org/blink/trunk@177134 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This allows us to remove a redundant squashing-related property in RenderLayer,
and is also a step towards not having non-compositing code know about how the layers
are positioned & drawn.

BUG=384520

Review URL: https://codereview.chromium.org/358913003

git-svn-id: svn://svn.chromium.org/blink/trunk@177133 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=386944

Review URL: https://codereview.chromium.org/356183002

git-svn-id: svn://svn.chromium.org/blink/trunk@177132 bbb929c8-8fbe-4397-9dbb-9b2b20218538

It was a race between the event arrived into timeline and the script that was evaluated in the page.

BUG=316150

Review URL: https://codereview.chromium.org/351873003

git-svn-id: svn://svn.chromium.org/blink/trunk@177131 bbb929c8-8fbe-4397-9dbb-9b2b20218538

When a TextIterator is created for a range that contains a replaced
element then one ',' (comma) will surely be emitted. This way the
length of the created TextIterator will be at least 1. In spite of
this, the WebCore::PlainTextRange::createRangeFor method assumes that
the length of the iterator only can be one if that one character is a
newline. The patch fixes this wrong assumption.

R=tkent@chromium.org
BUG=385630

Review URL: https://codereview.chromium.org/342483002

git-svn-id: svn://svn.chromium.org/blink/trunk@177130 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This comment is about ASSERTing the consistency of state that no long exists.
Also, remove m_canSkipRepaintRectsUpdateOnScroll because it's just
renderer()->isTableCell().

R=esprehn@chromium.org

Review URL: https://codereview.chromium.org/356123004

git-svn-id: svn://svn.chromium.org/blink/trunk@177129 bbb929c8-8fbe-4397-9dbb-9b2b20218538

We would eventually like to implement position: sticky, but the current
implementation isn't designed in a way that integrates well with the existing
scrolling and compositing system. For example, position: sticky relies upon
updateLayerPositionsAfterDocumentScroll to function correctly, but that
function has no other purpose and can otherwise be removed. Similarly,
position: sticky doesn't work at all with composited overflow scrolling, which
is now the default mechanism for driving scrolling in the engine.

Once we've got our scrolling and compositing house in order, we should return
to position: sticky and implement the feature in a way that integrates well
with the rest of the engine. For now, however, this CL removes our current
implementation so we can focus on improving our implementation of the scrolling
features we've already shipped.

R=eseidel@chromium.org, esprehn@chromium.org
BUG=389638, 231752

Review URL: https://codereview.chromium.org/346603007

git-svn-id: svn://svn.chromium.org/blink/trunk@177128 bbb929c8-8fbe-4397-9dbb-9b2b20218538

When emulation is unavailable (on device), only show the message about it.
When emulation is available, but without responsive design mode, move
toggle button to the OverridesView.

BUG=none

Review URL: https://codereview.chromium.org/358973003

git-svn-id: svn://svn.chromium.org/blink/trunk@177127 bbb929c8-8fbe-4397-9dbb-9b2b20218538

[TypeChecking=Nullable] on a nullable attribute meant that the C++ getter
wasn't called with a "bool& isNull" argument, instead the C++ value in a
boolean context was used to determine nullness.

This was only used for attributes where the C++ value was a pointer, and
made perfect sense for such attributes. So much sense, in fact, that it
makes sense to let that be the default behavior for such attributes.

Review URL: https://codereview.chromium.org/352643005

git-svn-id: svn://svn.chromium.org/blink/trunk@177126 bbb929c8-8fbe-4397-9dbb-9b2b20218538