GitLab

Removes a bunch of dead code that was relevant years ago before Chromium
even had a compositor, including a capture from the "backing store" and
a separate "render thread" for scaling and YUV conversion. Also,
gutted-out and simplified the unit tests.

In addition, BUILD files were modified so that only the platforms that
have implementation build/link the relevant modules.

BUG=159234

Review-Url: https://codereview.chromium.org/2605973002
Cr-Commit-Position: refs/heads/master@{#441288}

When double tapping on Forward navigation button the following happens:
- loadRequest: is called for the first Forward Tap (WKNavigation #1)
- loadRequest: is called for the second Forward Tap (WKNavigation #2)
- didStartProvisionalNavigation: is called for WKNavigation #1
- didFailProvisionalNavigation: is called for WKNavigation #1

CRWWebController used to record WKNavigation #1 as latest navigation
in didStartProvisionalNavigation: overriding WKNavigation #2.

Later inside didFailProvisionalNavigation: CRWWebController cancelled
the load because WKNavigation #1 was determined as latest.

This CL replaces _latestWKNavigation object with CRWWKNavigationStates
object which allows better tracking of latest WKNavigation object.

BUG=677327

Review-Url: https://codereview.chromium.org/2601083003
Cr-Commit-Position: refs/heads/master@{#441287}

Layout tests and blink perf tests will run with SlimmingPaintInvalidation
enabled.

BUG=646176

Review-Url: https://codereview.chromium.org/2595323003
Cr-Commit-Position: refs/heads/master@{#441286}

Note also the change made to CustomFrameView. Seems odd that this issue exists there after all these years. Maybe I should be making that change somewhere else?

R=sky@chromium.org

Review-Url: https://codereview.chromium.org/2613613002
Cr-Commit-Position: refs/heads/master@{#441285}

This is a re-land of half of the code in
https://codereview.chromium.org/2605743002
which was reverted because it changed layout test color.

TBR=avi,esprehn
BUG=634102

Review-Url: https://codereview.chromium.org/2615523002
Cr-Commit-Position: refs/heads/master@{#441284}

https://chromium.googlesource.com/external/github.com/catapult-project/catapult.git/+log/4ffd54d31cd5..a067dd2f2b17

$ git log 4ffd54d31..a067dd2f2 --date=short --no-merges --format='%ad %ae %s'
2017-01-03 jbudorick [systrace] Stop the browser after each test.

Documentation for the AutoRoller is here:
https://skia.googlesource.com/buildbot/+/master/autoroll/README.md

If the roll is causing failures, see:
http://www.chromium.org/developers/tree-sheriffs/sheriff-details-chromium#TOC-Failures-due-to-DEPS-rolls

CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel
TBR=catapult-sheriff@chromium.org

Review-Url: https://codereview.chromium.org/2611893002
Cr-Commit-Position: refs/heads/master@{#441283}

Reported crashes suggest a situation where we may double-free a
BrowserWindowController. Stacks show a UAF while trying to release a
reference to |self| held by an ObjC closure. The free occurs on an
autorelease triggered by a closure.

Most likely these are happening on the same closure. That is
__NSFireDelayedPerform "free()s" and its __delayedPerformCleanup
immediately tries to access to free again.

One hypothetical way this could occur would be if AppKit invoked
-[BrowserWindowController windowWillClose:] multiple times. Each time
would post a delayed autorelease closure. When the second is executed,
it would create the situation described.

Protect against this with an early exit in windowWillClose: if it's
already been called.

BUG=671213

Review-Url: https://codereview.chromium.org/2597023002
Cr-Commit-Position: refs/heads/master@{#441282}

We currently allow AC3/EAC3 codec id variants lower-case OTIs,
i.e. mp4a.a5 and mp4a.a6 in most places (e.g. stream_parser_factory.cc
and mime_util_internal.cc). We also need to support those lower-case
variants in StringToAudioCodec, since otherwise they won't get
recognized when an app adds a source buffer with that codec id and it's
checked in SourceBufferState::Init.

BUG=internal b/33554560

Review-Url: https://codereview.chromium.org/2587413004
Cr-Commit-Position: refs/heads/master@{#441281}

This fix memory corruption crash when app is launched from shelf
shortcut.

TEST=Manually, crash is no longer observed.
BUG=676392

Review-Url: https://codereview.chromium.org/2612453002
Cr-Commit-Position: refs/heads/master@{#441280}

BUG=675987
TEST=realtimeanalyser-basic.html

Review-Url: https://codereview.chromium.org/2598683002
Cr-Commit-Position: refs/heads/master@{#441279}

No need to hardcode to vector as that exposes more implementation
details that needed.  It can easily support all types.

BUG=

Review-Url: https://codereview.chromium.org/2616503002
Cr-Commit-Position: refs/heads/master@{#441278}

Add AudioScheduledSourceNode as a base class of schedulable source nodes to capture
the common features of these source nodes.

Intent: https://groups.google.com/a/chromium.org/d/msg/blink-dev/Vr6XJNfVAK4/RU3AP19WAQAJ

BUG=661207
TEST=audio-scheduled-source-basic.html

Review-Url: https://codereview.chromium.org/2471353004
Cr-Commit-Position: refs/heads/master@{#441277}

This CL:
* Cleans up the network-proxy HTML and JS files
* Adds a 'Save' button for configuring manual proxy configurations
  since a partial configuration might be invalid.
* Prevents updates from overwriting in-progress manual config changes.
* Fixes the mechanism for saving other proxy configurations also.
* Fixes logic determining when fields are editable.

BUG=677357
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:closure_compilation

Review-Url: https://codereview.chromium.org/2605373002
Cr-Commit-Position: refs/heads/master@{#441276}

In Chrome OS launcher's start page, we recreate the app tile views
if we need to display different numbers of apps compared with the
previous time (e.g., install a new app or uninstall an existing app).
We need to make sure the number of the displayed apps is properly
updated. Otherwise, it might cause browser crash because of retrieving
a stale pointer in StartPageView::StartPageTilesContainer::GetTileItemView().

This simple CL is a temporary fix that meant to merge back to M56. It
will be reverted later on Tot and a better but more complicated fix
will be landed.

BUG=675150

Review-Url: https://codereview.chromium.org/2605463003
Cr-Commit-Position: refs/heads/master@{#441275}

This also includes:
 - removal of the old base_test_runner + test_dispatcher code, since
   the linker tests were the last remaining users.
 - gn changes to make a wrapper script for the linker tests s.t. they
   can be run with <output directory>/bin/run_chromium_linker_test_apk

Review-Url: https://codereview.chromium.org/2605793002
Cr-Commit-Position: refs/heads/master@{#441274}

https://pdfium.googlesource.com/pdfium.git/+log/72f50f2021cd..d654d7854951

$ git log 72f50f202..d654d7854 --date=short --no-merges --format='%ad %ae %s'
2017-01-03 tsepez remove some CFX_ArrayTemplate in xfa_ffdocview.h

Documentation for the AutoRoller is here:
https://skia.googlesource.com/buildbot/+/master/autoroll/README.md

If the roll is causing failures, see:
http://www.chromium.org/developers/tree-sheriffs/sheriff-details-chromium#TOC-Failures-due-to-DEPS-rolls

TBR=dsinclair@chromium.org

Review-Url: https://codereview.chromium.org/2616523003
Cr-Commit-Position: refs/heads/master@{#441273}

Don't enable VrShell experiment in non-channel builds, since it
turns off overlay support for video playback on Android.  Enabling
tests is blocked on supporting overlay video properly anyway, so
nothing depends on this default.

BUG=678018

Review-Url: https://codereview.chromium.org/2613543002
Cr-Commit-Position: refs/heads/master@{#441272}

Moved writing-mode to be generated in ComputedStyleBase, which uses the
external type platform/text/WritingMode. Also added functionality to
add 'get' to the front of getters that have names that will conflict
with type names once the Blink rename has happened (and all methods
start with a capital letter).

BUG=628043

Review-Url: https://codereview.chromium.org/2609133002
Cr-Commit-Position: refs/heads/master@{#441271}

- Code clean up, refactoring and reordering.
- Add/update the comment. Some of them are irrelevant or out-of-date.

BUG=676386

Review-Url: https://codereview.chromium.org/2590823007
Cr-Commit-Position: refs/heads/master@{#441270}

CompositorFrameSinkHolder lives until resources are released or Buffer is
destroyed.

CompositorFrameSinkHolder no longer holds references to itself in the field
release_callback_.

BUG=675004

Review-Url: https://codereview.chromium.org/2584953002
Cr-Commit-Position: refs/heads/master@{#441269}

Fix for cbiesinger comment:
https://codereview.chromium.org/2568743005#msg16

Tnx to ikilpatrick for making me put these asserts for all code paths,
this bug got caught early.

BUG=635619

[ng_negmargin]

Review-Url: https://codereview.chromium.org/2605093002
Cr-Commit-Position: refs/heads/master@{#441268}

Failures on Windows ATI and NVIDIA:
https://build.chromium.org/p/tryserver.chromium.win/builders/win_optional_gpu_tests_rel/builds/6268
https://build.chromium.org/p/tryserver.chromium.win/builders/win_optional_gpu_tests_rel/builds/6116

R=kbr@chromium.org,jmadill@chromium.org
BUG=676848
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel;master.tryserver.chromium.android:android_optional_gpu_tests_rel

Review-Url: https://codereview.chromium.org/2614543002
Cr-Commit-Position: refs/heads/master@{#441267}

When downloading a dangerous download, Chrome should show an infobar in current tab.
However, the java side only checks the filename for dangerous download.
As a result, in some cases it may close a blank tab even if the download is dangerous.
That causes the download to be removed and cause a crash when returning to the caller.

BUG=667215

Review-Url: https://codereview.chromium.org/2609113004
Cr-Commit-Position: refs/heads/master@{#441266}

BUG=675325,579646

Review-Url: https://codereview.chromium.org/2608383002
Cr-Commit-Position: refs/heads/master@{#441265}

Don't enable SurfaceView overlays for video if the metadata says
that the video is rotated, since SurfaceView won't rotate it.
Also switch out of overlay mode if the metadata changes.

This also changes the 'exit full screen' logic to check if an
overlay is in use to decide if it needs to turn them off.

BUG=669081

Review-Url: https://codereview.chromium.org/2611783002
Cr-Commit-Position: refs/heads/master@{#441264}

BUG=612271
TEST=browser_tests --gtest_filter="CrSettingsPeoplePageSetupPinDialogTest*"
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:closure_compilation

Review-Url: https://codereview.chromium.org/2376293005
Cr-Commit-Position: refs/heads/master@{#441263}

BUG=none

Review-Url: https://codereview.chromium.org/2608003002
Cr-Commit-Position: refs/heads/master@{#441262}

Reland [Mojo Video Capture] Decouple VideoCaptureController from VideoCaptureBufferPool and VideoCaptureDeviceClient

PatchSet 1 is the state as reviewed and landed previously, see
https://codereview.chromium.org/2551193002/

This state caused a regression where video capture would be stuck in a loop of
old frames on Android when moving the browser to the background and bringing it
back to the foreground. This was caused by a new buffer pool instance being
created instead of the existing one being reused, while the renderer side would
still be pointing to buffers of the previous instance.

PatchSet 2 applies a fix for this issue.

Below is the CL description of the original CL:

Decouple VideoCaptureController from VideoCaptureBufferPool and
VideoCaptureDeviceClient. This is needed in order to prepare
VideoCaptureController for working with both the (legacy) in-process video
capture and the new video_capture Mojo service.

List of changes:
* Introduce an abstraction FrameBufferPool for the VideoCaptureController to
  talk to the VideoCaptureBufferPool.
* Move VideoFrameReceiverOnIOThread to separate file.
* Move factory logic for creating and connecting
  VideoCaptureBufferPoolImpl and VideoCaptureDeviceClient out of
  VideoCaptureController and into VideoCaptureManager as well as the tests.
* Update OWNERS to include new file and include chfremer as owner.
* In VideoCaptureDeviceClient: Rename |frame_format| to |foramt|.

This CL is part of the Mojo Video Capture work. For the bigger picture,
see [1] CL1.9.8

BUG=584797
TEST=
  content_unittests,
  video_capture_unittests,
  Apprtc loopback on Debug,
  Desktop Capture Example extension on Release

[1] https://docs.google.com/a/chromium.org/document/d/1Qw7rw1AJy0QHXjha36jZNiEuxsxWslJ_X-zpOhijvI8/edit?usp=sharing

Review-Url: https://codereview.chromium.org/2597833002
Cr-Commit-Position: refs/heads/master@{#441261}

BUG=None

Review-Url: https://codereview.chromium.org/2586543003
Cr-Commit-Position: refs/heads/master@{#441260}

Specifically, log the ResourceScheduler throttling a request, and starting
a throttled requests.

This CL  also includes modifying some aspects of URLRequest logging:
* Log priority and url at URLRequest creation time (so if the ResourceScheduler
  blocks it, there will be some identifier for the request).
* Remove priority from URLRequest::StartJob logging (redundant with logging
  in the constructor and SetPriority method).
* Always log SetPriority (it's relevant to the ResourceScheduler even if there's
  no job attached to the URLRequest).

BUG=None
R=mmenke@chromium.org
R=csharrison@chromium.org
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:closure_compilation

Review-Url: https://codereview.chromium.org/2579933002
Cr-Commit-Position: refs/heads/master@{#441259}

This test also enables WebViewPointerLockInteractiveTests on OOPIF-based
<webview>.

BUG=582562

Review-Url: https://codereview.chromium.org/2591783003
Cr-Commit-Position: refs/heads/master@{#441258}

BUG=

Review-Url: https://codereview.chromium.org/2616523002
Cr-Commit-Position: refs/heads/master@{#441257}

We wait for the compositor before declaring the test complete.

BUG=664857

Review-Url: https://codereview.chromium.org/2607853003
Cr-Commit-Position: refs/heads/master@{#441256}

BUG=674258,674317
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2

Review-Url: https://codereview.chromium.org/2608543002
Cr-Commit-Position: refs/heads/master@{#441255}

With a Windows-drawn titlebar we need to make sure we're reporting the
frame color that Windows uses so that elements like the incognito icon
will contrast correctly. This patch changes theme_service_win to not
assume custom titlebar mode is on when calculating colors.

R=estade@chromium.org
BUG=676566

Review-Url: https://codereview.chromium.org/2605443002
Cr-Commit-Position: refs/heads/master@{#441254}

In macOS, when the user right clicks on the media router action while it's in
the toolbar overflow menu, it gets popped out onto the toolbar to show its
context menu. This change makes the action go back into the overflow menu once
the context menu is closed, which will be consistent with the behavior of
extension actions.

BUG=670585

Review-Url: https://codereview.chromium.org/2579613002
Cr-Commit-Position: refs/heads/master@{#441253}

Fixes flakiness caused by a subtle distinction in document fullscreen
state change: The tabCapture.onStatusChanged API can sometimes report a
fullscreen state change before the document itself is aware the change
occurred. This is because the tabCapture API's listener reports on the
browser's fullscreen state, whereas the document is only aware of its
own internal fullscreen state (which is updated only after the browser
notifies it of a change).

The flakiness was due to the C++ side of the test being notified that
the page had entered fullscreen before the page itself was aware of it.
This caused the simulated mouse click (to toggle fullscreen mode) to be
made early, and then both the C++ and JS sides of the test became
out-of-sync and froze, each waiting on the other to do something.

This change fixes the issue by always using the document's fullscreen
change listeners, to ensure the out-of-sync condition cannot occur.
This change also "freshens up" the JavaScript code in this test, using
newer ES6 syntax.

BUG=675851

Review-Url: https://codereview.chromium.org/2604133002
Cr-Commit-Position: refs/heads/master@{#441252}

If a selection crossed between shadow roots, it wouldn't be detected
and a new blank property would be added on mouse up.

The last patch about this was missing a vital exclamation point.

BUG=654163

Review-Url: https://codereview.chromium.org/2609243002
Cr-Commit-Position: refs/heads/master@{#441251}

Due to a quirk in VC++ it is easy to accidentally prevent a const global
array from being placed in the read-only data segment. This change
removes some should-be-harmless 'const' keywords to work around this
quirk and move ~960 bytes to the .rdata (read-only) data segment in
chrome.dll.

VC++ bug is filed here:
https://connect.microsoft.com/VisualStudio/feedback/details/3117602

Other instances of this quirk are also being worked around.

BUG=677351

Review-Url: https://codereview.chromium.org/2605393002
Cr-Commit-Position: refs/heads/master@{#441250}

Move uniform management to the base fragment program, rather than
repeating them in every sub-class.

BUG=667966
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_trusty_blink_rel

Review-Url: https://codereview.chromium.org/2601343002
Cr-Commit-Position: refs/heads/master@{#441249}