GitLab

Cherry-pick of http://crrev.com/26274003.

> Use total page scale in the LayerScrollOffsetDelegate.
>
> In order for the LayerScrollOffsetDelegate to see the most up to
> date pageScale we need to use the total page scale rather than
> just the value that was last synced with the main thread.
>
> BUG=None
> TEST=cc_unittests

BUG: 10806014
Change-Id: I6d0c7dad789bf799e3122e19dc70d522974cfb42

Cherry-pick of uncommitted https://codereview.chromium.org/24676002/ PS2

Depends on Change-Id: Ica4cdeaa35b971b75ef1e664cd54f3fc7f95507a in
third_party/WebKit, which is also a fork from upstream Chromium, so this
is also a fork.

Conflicts:
	content/renderer/web_preferences.cc

BUG: 10750978

Change-Id: I8cab4b5ac209c55c2ac5a498cd4dc72f6e0a4229

If the first CookieManager API called by the app was
SetAcceptFileSchemeCookies or AllowFileSchemeCookies, the app would
crash with a segfault as the CookieMonster has not yet been created. Add
the call to ensure that it exists (creating the temporary one if needed)
before accessing it.

Bug: 11141036
Change-Id: I7a8a5960b159d34adfbd9aa00adf05896ea207d0

Merge "Cherry-pick http://crrev.com/25233002 "Make Android WebView filtering depend on scrolling status."" into klp-dev

This fixes an image quality regression which affects to some degree
almost every software WebView using an image, by introducing the same
workaround as classic WebView.

BUG: 10706494
Change-Id: I758257c5046b167780e5fcd07918f1855f08760c

Bug 11070366

Taken from
https://src.chromium.org/viewvc/chrome?revision=227511&view=revision and
straight-forward reolve (new setting was added amonst other settings
that didn't exist back in M30)

Original description:

Plumb the flag to blink and the compositor.
Disable by default in Android webview for now.

Depends on blink CL: https://src.chromium.org/viewvc/blink?view=rev&revision=158956

BUG=303945

Review URL: https://codereview.chromium.org/25904004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@227511 0039d316-1c4b-4281-b951-d872f2087c98

Conflicts:
	cc/base/switches.cc
	cc/base/switches.h
	cc/trees/layer_tree_settings.cc
	content/renderer/web_preferences.cc
	webkit/common/webpreferences.cc
	webkit/common/webpreferences.h

Change-Id: I140594e9b280f408c7b237ee5f17c569911f5b85

Bug:10910989

This CL fixes the shouldoverrideurlloading by
1. communicating the shouldoverrideurlloading IPC over a
control IPC to prevent any freezes when webcontents is not ready.
It also sends the opener_id as routing_id if a popup is involved.

2. Not using WebContentsObserver for onPageStarted. Using WebContentsObserver
causes stale URLs and out of order onPageStarteds.

This change also removes the now unneeded change in
content/render_view_host, which breaks upstream tests.

Change-Id: I3175418b0d30f2ca5e20ed840af7d1eb041f112a

Cherry-pick of Chromium crrev.com/r227614

BUG: 9103603

PictureImageLayerImpl ignores PictureLayerImpl::ideal_content_scale_ and
conditionally sets tile scale to 1.f. This makes the comparison in
PictureLayerImpl::AppendQuads to check for incomplete tiles incorrect.

Fix by overriding PictureImageLayerImpl::CalculateContentsScale to
always set ideal_content_scale_ to 1.f.

Fix for internal bug b/9103603.

TEST=PictureImageLayerImplTest.IgnoreIdealContentScale

Change-Id: Ib73dedd91e60d16a8610b4e6b8e560bfa4e12966

Cherry-pick of Chromium crrev.com/227508

Bug: 11065663

Original description:

The AwQuotaManagerBridge needs to be usable from any thread, not just
the UI thread. Make it refcounted and have its methods post to the UI
thread if not already there. Since none of the methods are synchronous
there's no need to wait for the posted messages to be handled.

BUG=305205
R=benm@chromium.org

Review URL: https://codereview.chromium.org/26477002

Change-Id: I11bc54b2dac83af8aedd5b28b33ba6e24a3d33d4
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@227508 0039d316-1c4b-4281-b951-d872f2087c98

This commit was generated by merge_to_master.py.

Change-Id: I0ee9cadd8d40aa5fea7450dadb93244110642752

This commit was generated by merge_to_master.py.

Change-Id: Icbada855096943f29f59aa9bfe136f26dd055024

Bug: 11041218

This reverts commit daf6d950 which
is the second piece of 0a8f47fb.
Reverted for the same reason as explained in
  f4dddd43d60c5820a173ba337a00ea490a5e1a93

Bug: 11041218

The existing approach can cause two issues with applications:
The stale urls (the WebContentsObserver seems never really supported
didStartLoading for Android correctly) in onPageStarted, and out of order
onPageStarted's. Reverting.

This reverts commit 0a8f47fb.

Change-Id: I3b81ce8243bec26164a2fdf01569ef824c480cb4

Cherry-pick of Chromium crrev.com/r227261

BUG: 11086057

Original description:

If shrinksStandaloneImagesToFit is true, then it breaks single image
documents in wrap contents mode.

Change-Id: Ie52f6eba207d210f3fbc8a80d7d449272f5047e0

BUG: 11001058

Webview does not use ProactiveBeginFrame, so cannot depend on BeginFrame
for activation.

Note that this patch as is cannot land in upstream Chromium since all
other platforms do use ProactiveBeginFrame.

Forking since Chromium trunk looks very different so fixes there cannot
be cherry-picked now.

Change-Id: I6a6e2e6f9dcfc7b29d0af2e0f04e2e042610380e

Cherry pick of https://codereview.chromium.org/25453008/

Bug: 7882891

Original Description:
[Android WebView] Disable WebRTC.

Android WebView compiles out WebRTC, so we should disable runtime
interfaces too.

Change-Id: I72da32a95af8feb395a144a64a267061a6179fd0

Cherry pick of https://chromiumcodereview.appspot.com/25878003/

Bug: 10825362

Original Description:
[Android WebView] Add an API to control favicon loading.

Change-Id: If752842bbff8f5b07719d84d69cad8bb4642bd45

Refactor the CookieManager to allow it to be used without starting
Chromium. If Chromium has not been started when the cookie manager is
first used, a temporary cookie monster will be created and used to
service requests (with its own client/backend threads). If the app
subsequently triggers regular Chromium startup (e.g. by creating a
WebView) then the temporary one will be shut down cleanly before the
real instance is created.

Bug: 10969703
Change-Id: I903630cbdbccc61958e5e8b9179c0a3d09594a8b

Cherry-pick of Chromium crrev.com/r226947

BUG: 11063091

Original description:

The instance of AwContentsIoThreadClient is uninstalled when AwContents
(along with WebContents and RenderViewHost) is destroyed. But RenderView
can still be alive and send more requests. It may also be possible for
AwContentsIoThreadClient to be uninstalled after a request has started.
So null check all usage of AwContentsIoThreadClient::FromID to make sure
it returns a real object.

Change-Id: I17e7b7a8cccb6e833a8a176a236437688891dd16

Merge "Cherry-pick: [Android] Prevent re-entrance into ContentVideoView.exitFullscreen" into klp-dev

Cherry-pick of Chromium crrev.com/r226859

BUG: 11057493

Original description:

For Android WebView, if app decides to call exitFullscreen from
onDestroyContentVideoView, then re-entering into exitFullscreen and
causing an infinite recursion.

Fix by moving the onDestroyContentVideoView to after removeSurfaceView.
The next time destroyContentVideoView in, onDestroyContentVideoView will
not be called since mVideoSurfaceView is already null.

Change-Id: I5c4de70d1fe34a82e3393ed7d9164555040f5919

Cherry-pick of Chromium crrev.com/r226844

BUG: 11048885

Original description:

This is to reduce the number of gpu memory buffer allocations which
can consume multiple file descriptors.

Change-Id: Ifa525d26a02dc4ba6af0821d8f4f911c2ab71c17

Bug 10932261

This allows the AwSettings (java only) instance to be created in advance
of AwContents, thereby allowing the AwSettings to be used early on when
the chromium stack has not be fully started.
As a result AwSettings is now able to live as a 'dumb' Java container of
settings, prior to the chromium stack being initialized at all.

Change-Id: I4059a93a7ad132043998fe2eab318a3b066c882e

Bug 10932261

Originally from http://crrev.com/25092005/ PS9, but modified to land
in klp-dev.

Original Description:
[Android] Android code should not assume main thread == UI thread.

Although typically the case, the main thread of an android app is not
necessarily the thread in which the View is hosted (the Android "UI"
thread). Update base.ThreadUtils to be able to specify a non-main thread
UI thread, and update places in the code that make this assumption to
use the refactored ThreadUtils class.

Change-Id: Ie2967fedafcece90f4e65b52a84bdd21f6b745cd

Cherry-pick of Chromium crrev.com/r226676

BUG: 10951963

Original description:

We can never assume that calling View invalidate will lead to a draw,
since view can be invisible, off screen, clearView, etc so draw will
be optimized out. Therefore must always ensure that the fallback tick
matches what the compositor expects, ie if compositor expects a
BeginFrame, then we must always have a pending tick.

One case where this does not hold true is if a DrawGL clip does not
cover the entire visible area, we call invalidate, set block_invalides
to true, but does not schedule a fallback tick. Until the next draw
comes, webview remains in a live lock situation that no more fallback or
invalidates will be called.

Fix by not setting block_invalidates in this case so subsequent fallback
ticks can still be scheduled.

Change-Id: I469ea26094e5e8f0c00f4c94663d776d59b6d266

Cherry-pick of Chromium crrev.com/r226643

Conflicts:
	android_webview/browser/in_process_view_renderer.cc

BUG: 10606165

Original Description:

In OnMemoryPressure, drop tiles on invisible browser views.

Not reusing base::MemoryPressureListener because Android
WebView requires onTrimMemory to be called synchronously.

Change-Id: I5aaca024392ef964baac0bc53f554ea59b8dac41

Cherry-pick of Chromium crrev.com/r226104

BUG: 10932261

Original description:

In Android, a short lived EGL context is created on the UI thread during
start up to collect GPU info. One side effect is that if there is an
existing EGL context current, then it is released. This is a problem for
Android WebView but may affect Chrome as well.

This patch saves and restores the existing current context if there is
one.

Change-Id: I4efec482833a6e06e81d22711935ac2b937756ce

Cherry-pick of Chromium crrev.com/226442

BUG: 10606165

Original description:

NICE_TO_HAVE will pre-render the soon-bin, which is based on both
velocity and distance. Previously, EVERYTHING will also pre-render the
eventually bin. NICE_TO_HAVE is a better fit for the Android WebView
memory management model where we assume a webview will not go berserk
with memory.

Note that this can still fill up on fling.

Change-Id: Iba36767172c89ac59d0209769940cf8c33f6571f