GitLab

This fixes another cause of crashing without a valid parent proxy in
RenderFrameProxy::CreateFrameProxy.  This occurred when a renderer
crashed, and another renderer added a child frame, which triggered a
new proxy for that frame to be created for the crashed process.  The
crashed process was recreated just to create the proxy, and the proxy
creation crashed because its parent proxy didn't exist.

This CL fixes InitRenderFrameProxy to not recreate a process just to
create proxies.  The process should only come back if it ever needs to
host a RenderFrame, and all the proxies should already be created
then.

BUG=476846

Review URL: https://codereview.chromium.org/1138413002

Cr-Commit-Position: refs/heads/master@{#329972}

Clang's static analyzer emits the warning "-Wobjc-designated-initializers" for
CustomFrameView, because CustomFrameView overrides two designated initializers,
but the methods don't call super. The methods throw exceptions, which Clang's
static analyzer isn't smart enough to catch. Declaring the methods with
NS_UNAVAILABLE suppresses the clang warning.

BUG=

Review URL: https://codereview.chromium.org/1101923002

Cr-Commit-Position: refs/heads/master@{#329971}

Make Mandoline always run the mojo:window_manager.
(afaik we don't want to specify other command line apps)
Make Mojo Runner run the command line app (fallback on WM).
(The Android impl wasn't respecting the command line app)
Consolidate on Context::RunCommandLineApplication helper.

BUG=486220
TEST=Android mojo shell and apptest runner can run non-WM apps (eg. apptests).
R=sky@chromium.org

Review URL: https://codereview.chromium.org/1134713003

Cr-Commit-Position: refs/heads/master@{#329970}

Don't open the Collected Cookies dialog if the related web contents is null because the tab was already closed.

BUG=434569

Review URL: https://codereview.chromium.org/1017683003

Cr-Commit-Position: refs/heads/master@{#329969}

Review URL: https://codereview.chromium.org/1138313003

Cr-Commit-Position: refs/heads/master@{#329968}

BUG=465453

Review URL: https://codereview.chromium.org/1128823008

Cr-Commit-Position: refs/heads/master@{#329967}

SchemeIsCryptographic more appropriately reflects
the intent (that a cryptographically secure
transport was used - e.g. https:// and wss:// schemes),
whereas SchemeIsSecure also considers situations where
the scheme is locally trusted (e.g. filesystem URLs)

BUG=362214

Review URL: https://codereview.chromium.org/1131963003

Cr-Commit-Position: refs/heads/master@{#329966}

Per in-person discussions, as long as we have a fork of mojo shell we'll fork the interfaces.

I'll do third_party/mojo/src/mojo/public/java/application in a followup.

TBR=tsepez

Review URL: https://codereview.chromium.org/1139123006

Cr-Commit-Position: refs/heads/master@{#329965}

Previously, if a group target like 'mandoline:all' was specified as
an additional_compile_target that we wanted to build, the gn implementation
of 'analyze' would not handle it correctly (it could only handle dependencies
on files that existed in the build directory, like executables).

This patch modifies the MB implementation to also handle label-like targets
(the ninja mandoline:all target is the equivalent of the GN //mandonline:all
target); we do this by calling 'gn refs' twice, once looking for a list
of output paths to match against the compile targets and once looking for
a list of phony labels to match against the compile targets.

We also will match against just the target_name, so 'chrome_shell_apk' will
also match "//chrome/android:chrome_shell_apk'. This may result in
too many targets being rebuilt, but we can adjust that if need be.

This is somewhat inefficient, but the alternatives would be to either
1) force the user to specify the stamp files for a group, which would be
   ugly, or
2) force the user to specify all the compile targets in terms of GN labels,
   which would be different from how the recipes work w/ GYP and possibly
   require us to map binary names to targets outside of GN, which would
   be a maintenance headache.

R=scottmg@chromium.org,brettw@chromium.org
BUG=487035
CQ_EXTRA_TRYBOTS=tryserver.chromium.mac:mac_chromium_gn_rel;tryserver.chromium.win:win8_chromium_gn_rel

Review URL: https://codereview.chromium.org/1135743005

Cr-Commit-Position: refs/heads/master@{#329964}

This includes removing most direct references and uses from telemetry.

BUG=267773,476709

Review URL: https://codereview.chromium.org/1132993004

Cr-Commit-Position: refs/heads/master@{#329963}

Summary of changes available at:
https://chromium.googlesource.com/skia/+log/86a17e7..87a773c

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
TBR=robertphillips@google.com

Review URL: https://codereview.chromium.org/1136123008

Cr-Commit-Position: refs/heads/master@{#329962}

Instead of being optimistic about using the secure proxy (i.e. use it unless
the check fails), flip the logic so that the secure proxy is not used unless
the check succeeds.

BUG=466753

Review URL: https://codereview.chromium.org/1132083004

Cr-Commit-Position: refs/heads/master@{#329961}

This prevents layout tests from crashing with --disable-gpu

Review URL: https://codereview.chromium.org/1135943006

Cr-Commit-Position: refs/heads/master@{#329960}

Enable firewall hole punching for Chrome Apps by taking this feature out
from behind a flag.

BUG=478231

Review URL: https://codereview.chromium.org/1136843008

Cr-Commit-Position: refs/heads/master@{#329959}

Move ContentBrowserSanityChecker and WebContentsObserverSanityChecker from content/public/test to content/test

TEST=compiles
BUG=None

Review URL: https://codereview.chromium.org/1137083005

Cr-Commit-Position: refs/heads/master@{#329958}

This enables ownership of content shell support for testing
Web Bluetooth. An example of early support being added in
https://codereview.chromium.org/1132943002
content/shell/browser/layout_test/layout_test_bluetooth_adapter_provider.h/cc

BUG=420275

Review URL: https://codereview.chromium.org/1142523003

Cr-Commit-Position: refs/heads/master@{#329957}

Added a new helper function to determine whether the document has focus,
taking into account whether an iframe or webview element has focus
and whether the document is hidden.

This fixes issues where there were two ChromeVox active indicators showing,
one on the outer frame and one inside the inner frame or webview. It also
fixes issues with other speech-generating events being processed in both
frames at once.

BUG=484904

Review URL: https://codereview.chromium.org/1137403003

Cr-Commit-Position: refs/heads/master@{#329956}

This will allow us to keep sending reports over HTTPS until the server
can handle encrypted reports over HTTP.

BUG=461590

Review URL: https://codereview.chromium.org/1134283003

Cr-Commit-Position: refs/heads/master@{#329955}

As of KitKat, the standard Google IME always calls
deleteSurroundingText(1, 0) when the soft backspace button is pressed.
But Jellybean and below instead mixed in a backspace key event into the
IME stream, and the current Samsung keyboard still has this behavior.
If we only forward such key events to Blink during an ongoing
composition, then Blink will delete its string but the
BaseInputConnection won't update its state, leading to a desync
and strange bugs later.

Our sendKeyEvent method already had a workaround to call
deleteSurroundingText directly in such cases. sendKeyEvent is called
when Samsung keyboard is used with Chrome, but WebView instead goes
through the alternate Chrome.dispatchKeyEvent path.  For that reason we
currently have a Samsung+WebView specific desync bug.

This patch changes dispatchKeyEvent to call sendKeyEvent, like
https://codereview.chromium.org/759033002 did before it was reverted, to
extend the workaround to WebView.  This would have the unfortunate side
effect of causing a bug with backspace key-repeats on physical
keyboards, as the code there only applies backspace on keyup.  So this
patch changes it to keydown (which is better UX even aside from
key repeat) and also forwards the system-supplied key events directly
instead of throwing them out and generating artificial ones.

BUG=483514

Review URL: https://codereview.chromium.org/1137673005

Cr-Commit-Position: refs/heads/master@{#329954}

The tail end of libusb_submit_transfer assumes that the transfer object
is still valid. It may in fact have already been freed by the transfer
completion callback if that call occurs on a different thread. As a
simply workaround (since the libusb code will be going away soon) this
patch simply posts the transfer deletion back to the thread where
libusb_submit_transfer is called so that they are serialized.

BUG=487712

Review URL: https://codereview.chromium.org/1126203008

Cr-Commit-Position: refs/heads/master@{#329953}

BUG=None

Review URL: https://codereview.chromium.org/1124193005

Cr-Commit-Position: refs/heads/master@{#329952}

See https://codereview.chromium.org/1131553008/

Cr-Commit-Position: refs/heads/master@{#329951}

The application of special storage policy is split out into a new class,
QuotaPolicyCookieStore, in content/browser/net.

BUG=467596
TEST=No visible impact.

Review URL: https://codereview.chromium.org/1016643004

Cr-Commit-Position: refs/heads/master@{#329950}

R=lcwu@chromium.org,servolk@chromium.org,damienv@chromium.org
BUG=internal b/19868692

Review URL: https://codereview.chromium.org/1125033003

Cr-Commit-Position: refs/heads/master@{#329949}

  chrome.contextMenus.

We should not increment |num_items| if we're not adding the item to the
  menu model. We have a test for this already:
  ExtensionContextMenuModelTest::ExtensionItemTest, since that is looking
  at |num_items| to verify
  (by ExtensionContextMenuModelTest::CountExtensionItems()).

BUG=463467
Test=Add 6 browser_action items from an extension that has "contextMenus" permission:
for (var i = 0; i < chrome.contextMenus.ACTION_MENU_TOP_LEVEL_LIMIT; i++) {
  (function(idx) {
    chrome.contextMenus.create({ contexts: [ "browser_action" ], title: "entry " + idx }, function() {
      console.debug("created " + idx, chrome.runtime.lastError);
    })
  })(i);
}
Expect all 6 (ACTION_MENU_TOP_LEVEL_LIMIT) items to show up when you right
click the extension icon in browser toolbar.

Review URL: https://codereview.chromium.org/1137733002

Cr-Commit-Position: refs/heads/master@{#329948}

Android AAPT 21+ will automatically move attributes to a version
of the XML file with -vXX resource qualifier. So we no longer need
our generate_v14_compatible_resources.py script.

But AAPT only drops the attributes, so if we don't use our script,
we need to put left/right in addition to start/end in our layout xmls.

The plan is, disabling our script for Android support library and
keep using for ourselves, until we can drop the script entirely
when we discontinue support for JB.

TBR=cjhopman@chromium.org
BUG=487391

Review URL: https://codereview.chromium.org/1136953009

Cr-Commit-Position: refs/heads/master@{#329947}

This changes implementation of Directory::NextId() to no longer depend on
"next_id" counter stored in Sync DB. Even though there was a special precaution
to avoid potential collision of local IDs by fast forwarding next_id by a large
number on every save there are still a couple scenarios leading the the collision
between a local ID generated on a client and originator_client_item_id coming
from the server (which is basically a local ID of an item previously committed
to the server.

The new implementation of Directory::NextId() is based on base::GenerateGUID.
The following changes accompany the transition to the new ID format:
1) The next_id field is no longer needed in Directory::PersistedKernelInfo
2) The code that persists next_id is no longer needed.
3) Forcing the directory to save by marking it with KERNEL_SHARE_INFO_DIRTY
    flag right after loading it is no longer necessary. That was done primarily
    to increment next_id by a large value as a way to attempt this problem.
4) Fixing #3 revealed a bug in Directory::IncrementTransactionVersion that
    doesn't mark the directory with KERNEL_SHARE_INFO_DIRTY when incrementing
    a datatype's transaction version. That caused a number of bookmark tests
    to fail due to the transaction version check. Fixed it.
5) Changing local IDs to GUIDs had a subtle effect on some tests. For the
    types that don't support ordering, sorting of entries in ParentChildIndex
    defaults to ID comparison. Since generation of local IDs base on next_id
    followed a predictable pattern, some tests relied on order of items returned
    by ParentChildIndex. Switching to GUIDs disrupted the expected order
    (of entries that are unordered in their nature). To preserve the existing
    ordering pattern for unordered entries, I changed ParentChildIndex to
    default to META_HANDLE comparison.
6) Another test that failed due to the GUID transition was
    SyncableDirectoryTest.BookmarkTagTest. The test relied on a predefined,
    repeatable local ID which was no longer the case with the GUID generation.
    I found the test to be redundant. There is another tests that verifies
    the same functionality - unique bookmark tag generation algorithm.

For now the change leaves next_id in the database even though the field is
no longer used. I should probably file another bug to track its removal.

BUG=362467

Review URL: https://codereview.chromium.org/1136953013

Cr-Commit-Position: refs/heads/master@{#329946}

The use_pinch_zoom_scrollbars flag is dead.

Review URL: https://codereview.chromium.org/1136843009

Cr-Commit-Position: refs/heads/master@{#329945}

Add a test to confirm that data type selections are set when SCF is
closed rather than immediately when the selection is made.

BUG=480604

Review URL: https://codereview.chromium.org/1140593002

Cr-Commit-Position: refs/heads/master@{#329944}

Summary of changes available at:
https://chromium.googlesource.com/chromium/blink/+log/be1e6ee..842d065

TBR=jsbell@chromium.org,timvolodine@chromium.org

Review URL: https://codereview.chromium.org/1126263006

Cr-Commit-Position: refs/heads/master@{#329943}

Improve quality of software rendering path by replacing call to SkCanvasVideoRenderer::ConvertVideoFrameToRGB in video_decoder_shim with a shader and render-to-texture.

Questions about this CL:

* gpu::gles2::GLESInterface
Would it be correct to send a base::WeakPtr<GLESInterface> to the YUVConv
class constructor to use for everything in this instance of the class?
ATM, it just sends a GLESInterface* to methods that need it (which is
probably wrong).  It was not clear to me whether the GLESInterface * could
change between the ::Initialize() method

* OpenGL state
YUVConv::convert() changes texture bindings and viewport.  I currently read
the old viewport so that I can restore it - is this necessary?  I don't save
former texture bindings either - is this necessary?

* OpenGL 'getters'
When compiled debug, why do all output variables to GL functions need to
contain either 0 or -1 before calling the function?  In my case GetIntegerv
and GetShaderiv.

* Errors
It is possible that shader compile could fail.  If it fails, an error string
can be retrieved.  What should I do in this case?
Same with opengl errors.  I thought it would be good to check for them in
debug builds, but what do I do if I catch one?

* Texture formats
I need a single-component texture format for uploading the Y,U,V components.
On desktop I would use GL_RED, but unextended GLES does not have GL_RED, you
have to use GL_LUMINANCE.  The current code uses GL_RED, and it works, but do
I need to check for GLES extension for GL_RED or can I assume it is there?

* Texture sizes
Even with the description in video_frame.h the distinction between coded_size
visible_size and natural_size was not entirely clear to me.  It seemed the
output texture size in VideoDecoderShim::OnOutputComplete followed "coded_size"
so that is what I used as well.  It was my intention to always decode the
entire image to the user's buffer, not some subregion - so is my code correct?

* FBO's
I see that DrawBuffers is DrawBuffersEXT, and that the default GLES
implementation does not support it.  Is it not necessary in this environment
to call DrawBuffers?

* ::SendPictures
If there is a format error yuv_converter will return false.  How can an error
be handled here?

* testing
I have tested with the video_decoder example program (using VPx) and our
application which uses h264.  I believe both use I420 format.  Do you have an
easy way to test other formats?  Is this class part of a unit test?

BUG=483183,477737
R=bbudge@chromium.org
TEST=Run pepper video_decoder example in ppapi/examples/video_decode

Review URL: https://codereview.chromium.org/1111653004

Cr-Commit-Position: refs/heads/master@{#329942}

BUG=none
R=ganesh@chromium.org,pkasting@chromium.org

Review URL: https://codereview.chromium.org/1141073002

Cr-Commit-Position: refs/heads/master@{#329941}

Because Android's API doesn't let us know when a user has
deleted a shortcut from the home screen, record that the
shortcut still exists whenever the shortcut is launched.
This applies to both regular shortcuts and web app launches.

BUG=460229

Review URL: https://codereview.chromium.org/1129293009

Cr-Commit-Position: refs/heads/master@{#329940}

Useful for mojo developers.

BUG=486809

Review URL: https://codereview.chromium.org/1128903005

Cr-Commit-Position: refs/heads/master@{#329939}

BUG=486916
TEST=No more NC_AUTO_SUBFRAME renderer kills.

Review URL: https://codereview.chromium.org/1124313003

Cr-Commit-Position: refs/heads/master@{#329938}

"Recently closed" is always hidden and is first on the chopping block.

R=estade@chromium.org
BUG=329637

Review URL: https://codereview.chromium.org/1130753007

Cr-Commit-Position: refs/heads/master@{#329937}

This is particularly important on ChromeOS with freon stack.
SwapBuffers ack is sent when the buffers actually swapped, not when the
call to swap them was made. This changed the scheduler behavior and delayed
beginning of the next frame, so to regain it we allowed multiple in flight
frames. The effect this has on latency info is that two frames arriving in
rapid succession would override incorrect latency info making it look like
things ultimately happened earlier than they did.

This patch associates latency info with a particular frame's callback
thus avoiding it being overridden.

BUG=485302

Review URL: https://codereview.chromium.org/1128323006

Cr-Commit-Position: refs/heads/master@{#329936}

This CL saves ~50 minutes from the average Android Cycle times.

Note to perf sheriff, this may result a minor shift in the result for the android tests. It is unlikely since the hardware is typical.

R=sullivan@chromium.org
BUG=466101,471921

Review URL: https://codereview.chromium.org/1134633004

Cr-Commit-Position: refs/heads/master@{#329935}

BUG=

Review URL: https://codereview.chromium.org/1128233007

Cr-Commit-Position: refs/heads/master@{#329934}

Accidentally inverted a check during refactoring in
https://crrev.com/1137513002

Add back the '!' character that I deleted, and add unit tests to
guard against future regressions.

BUG=486675

Review URL: https://codereview.chromium.org/1138373002

Cr-Commit-Position: refs/heads/master@{#329933}