Merge lp:~abreu-alexandre/oxide/content-script-injection-main-world into lp:~oxide-developers/oxide/oxide.trunk

[Alexandre Abreu]

Enable *under some circumstances* content scripts to be injected:

1. in the main js world,
2. with an access to the oxide messaging API.

The Oxide QML API itself is not changed to achieve that. This is achieved through some declarative header extensions made to greasemonkey like scripts support in Oxide and a specific world Id used to referrence the messages coming in and out of those scripts.

In order to achieve the features listed in 1 and 2, what would be needed is then:

- have a greasemonkey script,
- have the script obviously be declared as emulateGreasemonkey,
- use a specific context for the messaging ("oxide://main-world"),

As long as one has the proper context name the script is injected in the main world, but only scripts marked as gm scripts will have the additional oxide api,

Enable *under some circumstances* content scripts to be injected:

1. in the main js world,
2. with an access to the oxide messaging API.

The Oxide QML API itself is not changed to achieve that. This is achieved through some declarative header extensions made to greasemonkey like scripts support in Oxide and a specific world Id used to referrence the messages coming in and out of those scripts.

In order to achieve the features listed in 1 and 2, what would be needed is then:

- have a greasemonkey script,
- have the script obviously be declared as emulateGreasemonkey,
- use a specific context for the messaging ("oxide://main-world"),

As long as one has the proper context name the script is injected in the main world, but only scripts marked as gm scripts will have the additional oxide api,

[lp:~abreu-alexandre/oxide/content-script-injection-main-world]

417. By Chris Coulson on 2014-03-17

Merge network-callbacks branch. This adds the following:
- WebContextDelegateWorker, which is basically a worker script that exposes entry points
  for processing callbacks from Chromium's IO thread.
- WebContext.networkRequestDelegate, which allows applications to modify HTTP request URL's and
  modify HTTP headers, using WebContextDelegateWorker
- WebContext.storageAccessPermissionDelegate, which allows applicatons to override the default
  permissions for storage accesses, using WebContextDelegateWorker. This is currently only called
  for cookies, but will be expanded to local storage, appcache, indexeddb and webdb.
- WebContext.userAgentOverrideDelegate, which allows applications to override navigator.userAgent for
  specific URL's, using WebContextDelegateWorker. Technically, this could have been handled on the UI
  thread without WebContextDelegateWorker, but it's assumed that this will be used in conjunction
  with networkRequestDelegate for overriding the User-Agent header. Using WebContextDelegateWorker
  allows the application to keep all user-agent override data in one script engine.
- WebContext.cookiePolicy for setting the default cookie policy, and supporting third-party
  cookie blocking

418. By Chris Coulson on 2014-03-17

Add some missing compile-time asserts

419. By Chris Coulson on 2014-03-17

UserScript.onScriptWillBeDeleted isn't meant to be public

420. By Chris Coulson on 2014-03-17

Fix a typo

421. By Chris Coulson on 2014-03-17

WebContext.addUserScript should probably always take ownership of an unowned UserScript

422. By Chris Coulson on 2014-03-19

WebView.addMessageHandler and WebFrame.addMessageHandler should behave like WebContext.networkRequestDelegateWorker etc, and fail to add a handler that's already in use by another message target. Also, add a test to verify that deleting a ScriptMessageHandler removes it from its message target

423. By Chris Coulson on 2014-03-19

Update to Chromium 35.0.1897.2

424. By Chris Coulson on 2014-03-20

Omit the mojo bits for now - from http://src.chromium.org/viewvc/chrome?view=revision&revision=257587, it looks like it should be disabled anyway, but we still get a dependency on mojo_system. We need to install the shared lib for this to work

425. By Chris Coulson on 2014-03-20

Rename IOThreadDelegate to IOThreadGlobals. This makes more sense, and we have an IOThreadDelegate now in Qt glue, with a more appropriate use for that name

426. By Chris Coulson on 2014-03-20

Make IOThreadGlobals a proper singleton. Calling BrowserProcessMain::CreateIOThreadGlobals was always a bit weird

427. By Chris Coulson on 2014-03-20

Add URLRequestContextFactory

428. By Chris Coulson on 2014-03-20

Create URLRequestContext in URLRequestContextGetter

429. By Chris Coulson on 2014-03-20

Move common parts of URLRequestContext initialization in to URLRequestContextGetter

430. By Chris Coulson on 2014-03-20

Add a system URL request context

431. By Chris Coulson on 2014-03-20

Assert that InitializeRequestContext is only called once

432. By Chris Coulson on 2014-03-21

Use net::TransportSecurityPersister so that we have permanent storage for HSTS. Also, make this available in read-only mode to off-the-record contexts

433. By Chris Coulson on 2014-03-23

Add BrowserContext::Params

434. By Chris Coulson on 2014-03-23

Fix a double free on shutdown

435. By Chris Coulson on 2014-03-24

BrowserContext::GetCachePath() and BrowserContext::GetPath() no longer return an empty path for OTR contexts, and BrowserContext::IsOffTheRecord() always returns false for the main context, since r432

436. By Olivier Tilloy on 2014-03-24

Implement a dummy AccessTokenStore to prevent geolocation requests from crashing the renderer.

437. By Chris Coulson on 2014-03-24

Update exclude list for tarball

438. By Chris Coulson on 2014-03-25

This provides some improvements to the changes in r425-r431. It:
- Makes IOThreadGlobals::Data publically available as Globals, and renames
  IOThreadGlobals to IOThread. This provides a proper API split for accessing
  things that belong to the IO thread
- Reverts the change that made IOThread a singleton - we want to control when
  to destroy it
- Destroys the system URL request context before the IO thread has gone away
  (fixes LP: #1296668)
- Gets rid of URLRequestContextInitializer - this was overcomplicated. Now, we
  subclass URLRequestContextGetter instead
- Move the initialization for common paramters out of URLRequestContextGetter
  in to the constructor for URLRequestContext so that we don't need to rely
  on URLRequestContextGetter to create the request context. This was a bit
  weird anyway

439. By Olivier Tilloy on 2014-03-25

Implement the JavaScriptDialogManager interface.

440. By Chris Coulson on 2014-03-26

Update to Chromium 35.0.1908.4

441. By Chris Coulson on 2014-03-26

Fix some issues with WebPopupMenu:
- Ensure WebPopupMenu::RenderViewHostDeleted() only responds to the appropriate host (LP: #1257663)
- Make sure that Hide() coming from Chromium hides the popup
- Take in to account that WebPopupMenu deletion is asynchronous, so a previous one may still
  exist when a new one is created (assert that the old one is hidden)
- Move all of this stuff in to WebView while we're at it

442. By Chris Coulson on 2014-03-26

Delete WebFrameAdapterPrivate

443. By Chris Coulson on 2014-03-27

Remove footgun (webSecurityEnabled). It seems this is only for testing purposes in Chromium, and we have no need for it

444. By Chris Coulson on 2014-03-27

Remove the smooth scrolling pref - this is currently controlled in Chromium by a temporary command line switch. I suspect it will go away at some point

445. By Chris Coulson on 2014-03-28

The oxide::WebFrame destructor reenters virtual functions implemented by derived classes - add back a Destroy() method and make its destructor private

446. By Chris Coulson on 2014-03-28

When our RenderViewHost is swapped out, Chromium attaches the new RVH to the frame tree before notifying us of the swap. This means that the FrameTreeNode's corresponding to our WebFrame's have been deleted when we delete our frame tree, resulting in WebFrame::frame_tree_node_ being left dangling. Modify WebFrame to not store a pointer to its corresponding FrameTreeNode. This fixes a memory corruption bug

447. By Chris Coulson on 2014-03-28

Fix an invalid read in keyboard event handling

448. By Chris Coulson on 2014-03-31

Don't set GLX_CONTEXT_RESET_NOTIFICATION_STRATEGY_ARB and GLX_LOSE_CONTEXT_ON_RESET_ARB even if supported. Qt doesn't set those, and all GL contexts within a share group must be created with identical attributes

449. By Chris Coulson on 2014-04-01

Merge from lp:~chrisccoulson/oxide/window-opening

450. By Chris Coulson on 2014-04-01

Merge from lp:~mardy/oxide/lp1298264

451. By Chris Coulson on 2014-04-01

Add WebContext.sessionCookieMode

452. By Chris Coulson on 2014-04-01

Don't build blink debug symbols on native ARM builds, because we're still hitting OOM conditions despite throwing it every memory optimization that I can think of. We already turn them off for Chromium builds

453. By Chris Coulson on 2014-04-01

Change API version to 1.0

454. By Chris Coulson on 2014-04-01

When initializing a newly created WebView, ensure we sync the actual WebView size and visibility to its newly adopted WebContents

455. By Chris Coulson on 2014-04-02

Add code for creating new WebView's with the opener set (but disabled by default until we figure out a safe way of deciding what to do)

456. By Chris Coulson on 2014-04-02

For the navigation case, create WebContents after we have a new WebView

457. By Chris Coulson on 2014-04-02

Don't abort for unhandled dispositions, but do an early exit instead

458. By Olivier Tilloy on 2014-04-02

Send fake keydown and keyup events when composing text with an input method.

459. By Chris Coulson on 2014-04-02

Update to Chromium 35.0.1916.6

460. By Chris Coulson on 2014-04-02

Try to make sure we don't create multiple contexts with the same storage location when running the tests

461. By Chris Coulson on 2014-04-02

Lazy init WebContext (LP: #1297552 part 1)

462. By Chris Coulson on 2014-04-02

Expose the default WebContext with Oxide.defaultWebContext(), and get rid of all of the proxy properties from the Oxide singleton (LP: #1297552 part 2)

463. By Chris Coulson on 2014-04-02

Ignore changes to WebContext.sessionCookieMode when there is no data path

464. By Chris Coulson on 2014-04-03

Use qWarning in qt/core/glue

465. By Chris Coulson on 2014-04-03

Give UserScript.url a NOTIFY signal

466. By Chris Coulson on 2014-04-03

Change LoadEvent.error to LoadEvent.errorDomain, expose the raw error code as LoadEvent.errorCode and ensure we cover the full range of Chromium net errors

467. By Chris Coulson on 2014-04-03

Fix a race in tst_WebView_newViewRequested.qml

468. By Chris Coulson on 2014-04-03

Make use of net::IsCertificateError

469. By Chris Coulson on 2014-04-03

Don't duplicate the URL on WebFrame

470. By Alexandre Abreu on 2014-04-04

fixes

471. By Olivier Tilloy on 2014-04-04

Use the EGL vendor string to infer the form factor.
This will prevent application confinement from getting in the way.

472. By Chris Coulson on 2014-04-04

Dispatch WebView.navigationRequested for all window.open() calls or clicks on links with target="_blank". Applications should use this to deny requests to open a new view. Dispatch WebView.newViewRequested when a view should be created and shown. Applications that implement this are expected to create a new WebView here - it shouldn't be used for policy. Hence, NewViewRequest.url and NewViewRequest.userGesture are gone

473. By Chris Coulson on 2014-04-04

If we can't create new windows, coerce the navigation to a CURRENT_TAB navigation

474. By Chris Coulson on 2014-04-04

Fix a typo

475. By Chris Coulson on 2014-04-04

Add bug number

476. By Chris Coulson on 2014-04-04

Get rid of the hack that registers all resource context when the ResourceDispatcherHost is created - this is cruft left over from earlier days

477. By Chris Coulson on 2014-04-05

Build Chromium with toolkit_views: 0 and enable_plugins: 0

478. By Chris Coulson on 2014-04-05

Subframe navigations with a disposition other than CURRENT_TAB in a view that can't create new windows should be redirected to the main frame

479. By Chris Coulson on 2014-04-05

Make that last commit actually work

480. By Chris Coulson on 2014-04-07

Add an option for enabling proprietary codecs

481. By Chris Coulson on 2014-04-07

Build and ship the l10n pak files

482. By Chris Coulson on 2014-04-07

Add file picker support

483. By Chris Coulson on 2014-04-07

Add missing resources

484. By Chris Coulson on 2014-04-07

Only send left, middle and right mouse button events to the renderer. Allow extra mouse buttons to bubble up

485. By Chris Coulson on 2014-04-07

Unbreak mouse move events

486. By Chris Coulson on 2014-04-08

Add cursor support

487. By Chris Coulson on 2014-04-08

Use QQuickItem::setCursor

488. By Chris Coulson on 2014-04-08

Rework RenderViewItem resizing

489. By Chris Coulson on 2014-04-08

Make sure the locale pak files get installed

490. By Chris Coulson on 2014-04-08

Reimplement RenderSandboxHostLinux to run the SandboxIPCProcess in a proper child process (created with fork + exec), as opposed to just forking the browser process and running the helper without exec'ing. Chromium can do this because it fork's the browser process early on before any threads are created, whereas when we fork with Oxide, we already have a full QML application and many running threads. There is a risk of deadlock with this and concerns about memory consumption too, so it's safer to exec a new process image to run the sandbox IPC helper

491. By Chris Coulson on 2014-04-09

Update to Chromium 35.0.1916.27

492. By Chris Coulson on 2014-04-09

Ensure that the compositing configuration uses the GpuDataManager (although, we currently don't load any gpu info)

493. By Chris Coulson on 2014-04-09

Drop remove-broken-gcc-option.patch and add the fix to common.gypi

494. By Chris Coulson on 2014-04-09

Add a comment to oxide_sandbox_ipc_process.cc

495. By Chris Coulson on 2014-04-09

Delete unused patch for now

496. By Chris Coulson on 2014-04-09

Make force compositing mode work correctly

497. By Chris Coulson on 2014-04-10

In the non-compositing path, handle the case where Chromium gives us a backing store with an out of date size

498. By Chris Coulson on 2014-04-10

Get rid of RenderWidgetHostViewDelegatePrivate again

499. By Chris Coulson on 2014-04-10

Ensure we actually composite during resizes

500. By Chris Coulson on 2014-04-10

Fix a resize crash caused by sending more than one acknowledge for a buffer swap

501. By Chris Coulson on 2014-04-10

Don't unconditionally build with proprietary_codecs=1, as this causes us to advertise all codecs even when we don't build support for them

502. By Chris Coulson on 2014-04-11

Update Chromium to 36.0.1933.0

503. By Chris Coulson on 2014-04-11

Ensure WebPreferences.touchEnabled is not ignored

504. By Chris Coulson on 2014-04-11

The focus related methods on WebContentsView aren't called by content, but by the embedder for managing focus of views between tab changes. We don't need these, so assert that they aren't called

505. By Chris Coulson on 2014-04-11

WebView should be a focus scope

506. By Chris Coulson on 2014-04-15

Add experimental support for the Google Talk PPAPI plugin, required for Google Hangouts (hidden behind a build-time flag at the moment)

507. By Olivier Tilloy on 2014-04-15

Set no_javascript_access in ContentBrowserClient::CanCreateWindow(…).

508. By Olivier Tilloy on 2014-04-15

Improve error handling in form factor detection code.

509. By Chris Coulson on 2014-04-15

GpuChildThread is not thread safe, so store the global pointer in TLS. Provide a static function for accessing a thread safe MessageLoopProxy to dispatch tasks to the GPU thread

510. By Chris Coulson on 2014-04-15

Optimise the texture resource loading a bit, so that we batch loads in a single task

511. By Chris Coulson on 2014-04-15

Ensure we have a current GL context when deleting textures

512. By Chris Coulson on 2014-04-15

Make sure we use the correct MailboxManager to lookup the textures we get from Chromium's compositor

513. By Chris Coulson on 2014-04-16

Make ENABLE_PLUGINS the default and hide the Google Talk plugin support behind an environment variable

514. By Chris Coulson on 2014-04-16

Also make ENABLE_OXIDEQMLSCENE the default

515. By Chris Coulson on 2014-04-16

Add OXIDE_ALLOW_SANDBOX_DEBUGGING to make renderers dumpable

516. By Chris Coulson on 2014-04-16

Don't call prctl(PR_SET_NO_NEW_PRIVS) after forking the sandbox IPC process, as this prevents the suid sandbox from gaining enough privileges to map inode numbers to PID's on behalf of the zygote

517. By Chris Coulson on 2014-04-16

Move constants that are only used in one place from oxide_constants.h

518. By Chris Coulson on 2014-04-16

Turn the environment variable for GTalk in to a command line switch, so that the sandbox doesn't remove it

519. By Chris Coulson on 2014-04-16

Make sure we only enable options defined in the environment when they are set to "1"

520. By Chris Coulson on 2014-04-16

Update to Chromium 36.0.1941.0

521. By Chris Coulson on 2014-04-17

Add support for fullscreen API

522. By Chris Coulson on 2014-04-17

Make oxide::WebView::{Get,Set}URL and oxide::WebView::{,Set}IsFullscreen work correctly before Init is called (moving the url part from WebViewAdapter

523. By Chris Coulson on 2014-04-17

Fix a crash on Google Plus when frames are detached before we've done the idle script inject

524. By Chris Coulson on 2014-04-24

Merge the rendering-updates branch so that we can update Chromium. This removes support for the old 2D drawing mode and async compositing mode, and uses the software compositing and mailbox compositing modes (as an interim solution)

525. By Chris Coulson on 2014-04-24

Add back the ENABLE_COMPOSITING guard in updatePolish

526. By Chris Coulson on 2014-04-24

Don't hold pointers to pixel data or texture handles between frames, as those might go stale

527. By Chris Coulson on 2014-04-24

GpuUtils::GetAcceleratedFrameHandle is racy - the GPU thread task can run before we've incremented the reference count on the main thread, meaning that we can return an already deleted object. Enesure the main thread has a reference before dispatching any task to the GPU thread, and always return an object with a reference count already added

528. By Chris Coulson on 2014-04-24

Don't disable viewport in software compositing mode

529. By Chris Coulson on 2014-04-24

Get rid of BrowserProcessMain::flags and just do the form factor customizations in ContentMainDelegate

530. By Chris Coulson on 2014-04-24

Don't change compositing preferences now that compositing is always used

531. By Chris Coulson on 2014-04-24

Apply some Android-style customizations when on tablet and phone form factors. In particular, setShrinksViewportContentToFit prevents sites from being zoomed out sufficiently to not fit the display width

532. By Chris Coulson on 2014-04-24

Drop the form factor check in UpdateCursor, as it's not needed

533. By Chris Coulson on 2014-04-24

Set a wait cursor when loading

534. By Chris Coulson on 2014-04-24

We don't plan to implement InitAsPopup and InitAsFullscreen, as we have no use for popup RWHV's

535. By Chris Coulson on 2014-04-25

Deprecate WebPreferences.touchEnabled and WebPreferences.databasesEnabled, as both of these are runtime-wide settings in blink

536. By Chris Coulson on 2014-04-25

When fetching the texture ID, we wait on a sync point in the GL command stream. We block the Qt render thread until this has been fetched, which works with minimal overhead on the desktop. However, when running fairly heavy webgl demo's on mobile we sometimes have to wait in excess of 25ms, which means that the frame rate for the whole application falls through the floor. Mitigate against this in a couple of ways:
- Don't post a task to the GPU thread event queue to add a sync point callback. Instead, implement a task observer that processes these after every task
- Don't dispatch SwapAcceleratedFrame until we have a texture ID. Unfortunately, this does mean posting one extra task to the UI thread per frame. Not sure if there's another way to do that

537. By Chris Coulson on 2014-04-25

Update Chromium to 36.0.1951.5

538. By Chris Coulson on 2014-04-28

Fix the comment - BrowserProcessMain isn't reference counted anymore

539. By Chris Coulson on 2014-04-28

Add a couple of extra options that are used on Chrome for Android

540. By Olivier Tilloy on 2014-04-30

Fix a safety check in debug builds.

541. By Chris Coulson on 2014-05-01

Merge ninja-build branch. We now build the Chromium parts with ninja rather than make

542. By Chris Coulson on 2014-05-01

Unbreak checkouts when the release_deps URL changes

543. By Chris Coulson on 2014-05-01

Update to Chromium 36.0.1964.2

544. By Alexandre Abreu on 2014-05-02

merge trunk

545. By Alexandre Abreu on 2014-05-02

merge trunk

546. By Chris Coulson on 2014-05-02

linux_use_tcmalloc is deprecated

547. By Olivier Tilloy on 2014-05-07

Provide a Qt implementation of the default system LocationProvider.

548. By Chris Coulson on 2014-05-07

Add WebView.geolocationPermissionRequested API

549. By Chris Coulson on 2014-05-07

Update to Chromium 36.0.1976.2

550. By Chris Coulson on 2014-05-19

Update Chromium to 36.0.1985.8

551. By Chris Coulson on 2014-05-19

Omit WebContentsViewAura and RenderWidgetHostViewAura from the build, and save on a patch

552. By Chris Coulson on 2014-05-19

Make sure that unhandled keyboard events are passed to parent items. Thanks to Arthur Mello for starting this work

553. By Chris Coulson on 2014-05-19

Remove a superfluous include and member function prototype

554. By Chris Coulson on 2014-05-19

Stop using an internal blink header for keycodes, and add some missing codes

555. By Chris Coulson on 2014-05-19

We shouldn't generate keypress events with unmodified text on the web side for control characters. This makes Oxide behave like Chrome

556. By Chris Coulson on 2014-05-19

Add a compile assert that WebKeyboardEvent::text and WebKeyboardEvent::unmodifiedText are the same size, as we assume that they are

557. By Chris Coulson on 2014-05-19

Use ui::VKEY_PROCESSKEY in sendFakeCompositionKeyEvent

558. By Chris Coulson on 2014-05-19

Make sure we honour skip_in_browser in WebView::HandleKeyboardEvent

559. By Chris Coulson on 2014-05-19

Stop using blink::WebInputEvent::KeyDown. Aura and Android both use RawKeyDown + Char

560. By Olivier Tilloy on 2014-05-21

Add a loadHtml slot to the QML webview.

561. By Chris Coulson on 2014-05-21

Update Chromium to 36.0.1985.18

562. By Chris Coulson on 2014-05-22

Make BrowserContext::SetCookiePolicy work the same as BrowserContext::SetUserAgent

563. By Chris Coulson on 2014-05-22

Bump version to 1.1

564. By Chris Coulson on 2014-05-22

Don't abort in debug mode on key events

565. By Chris Coulson on 2014-05-22

Mouse, key and IM events are accepted by default in QtQuick, and we don't need to explicitly accept focus events

566. By Chris Coulson on 2014-05-22

Make sure we set the correct state on ignored mouse events

567. By Chris Coulson on 2014-05-22

HandleMouseEvent expects a pre-accepted mouse event

568. By Alexandre Abreu on 2014-05-23

fix win8 exclusion

569. By Olivier Tilloy on 2014-05-27

Fix FTBFS with Qt 5.3.

570. By Chris Coulson on 2014-05-28

Update Chromium to 37.0.2008.2

571. By Chris Coulson on 2014-05-28

Delete WebContents asynchronously in AddNewContents

572. By Chris Coulson on 2014-05-28

All handlers of WebView.newViewRequested must create a web view

573. By Chris Coulson on 2014-05-28

Change the popup blocker preferences after doing the page load in tst_NavigationRequest.qml. For some reason, the popup blocker preferences aren't picked up by the time we generate the click event otherwise

574. By Chris Coulson on 2014-05-28

Add a critical message to the console if an embedder doesn't create a new webview when requested. All of the tests are fixed now

575. By Chris Coulson on 2014-05-28

test_NavigationRequest3_reject relies on a timeout and debug builds are sufficiently slow to make this fail intermittently. Rearrange the test a little to make it a bit less susceptible

576. By Chris Coulson on 2014-05-28

Update the error message to point out the consequences of misbehaving

577. By Chris Coulson on 2014-05-28

Drop the hidden .channel file, as it's fairly pointless right now

578. By Olivier Tilloy on 2014-05-30

Add an 'icon' property to the QML WebView.

579. By Chris Coulson on 2014-05-31

Start decoupling some of our logic around WebContents out of WebView and in to a new class (WebViewContentsHelper). This will allow us to ensure we can observe events from a WebContents that doesn't yet have a corresponding WebView

580. By Chris Coulson on 2014-06-02

Update to Chromium 37.0.2017.2. This also removes our own implementations of SandboxIPCHandler and RenderSandboxHostLinux, as this has been reimplemented as a browser process thread in Chromium which is fine for Oxide

581. By Chris Coulson on 2014-06-02

ProcessObserver -> RenderProcessObserver

582. By Chris Coulson on 2014-06-02

Get rid of unused message declaration

583. By Chris Coulson on 2014-06-02

Stop using WebContents::SetUserAgentOverride to set the user agent string, as this is unreliable in some circumstances. Instead, have a child-process-global user agent string

584. By Chris Coulson on 2014-06-02

Use a command line switch for telling a render process that it is an incognito process. Having a dedicated IPC message is silly, because it never changes

585. By Chris Coulson on 2014-06-02

Make the form factor detection work in single process mode

586. By Chris Coulson on 2014-06-02

Refactor BrowserProcessMain to get rid of the static initializer, and also make Start() infallible. As we already CHECK() in various places that Chromium is running, allowing BrowserProcessMain::Start() to fail is fairly pointless

587. By Chris Coulson on 2014-06-02

Move GetNativeDisplay() from ContentClient. It should only be used from the browser process and it never changes, so make it an initialization parameter for BrowserProcessMain

588. By Chris Coulson on 2014-06-02

Add the shutdown hook when we start Chromium

589. By Chris Coulson on 2014-06-03

Stop tracking visibility in RenderWidgetHostView

590. By Chris Coulson on 2014-06-03

Update the RWH visibility when creating a view

591. By Chris Coulson on 2014-06-03

Loosen platform detection to recognise any platform name starting wth "ubuntu"

592. By Chris Coulson on 2014-06-03

Update warning

593. By Olivier Tilloy on 2014-06-03

Fix syntax error.

594. By Olivier Tilloy on 2014-06-03

Add a unit test for HTTP redirections to verify that the 'url' property of the WebView is updated in between the two load events.

595. By Chris Coulson on 2014-06-03

Update Chromium to 37.0.2024.2

596. By Olivier Tilloy on 2014-06-05

Fix FTBFS with Qt 5.3.

597. By Chris Coulson on 2014-06-05

The display handle is always EGL_DEFAULT_DISPLAY (ie, NULL) with Android EGL

598. By Chris Coulson on 2014-06-05

Implement ContentClient::GetDataResource (and GetDataResourceBytes for good measure) so that blink can get the images used in the media player UI

599. By Chris Coulson on 2014-06-06

Now that getting the default native display handle from Qt doesn't rely on having a ContentClient instance, use the proper display handle in IsUbuntuPhoneOrTablet() rather than EGL_DEFAULT_DISPLAY, which fails on mesa-egl in Unity 8 (where the compiled in default display is an X11 one)

600. By Chris Coulson on 2014-06-06

Support surfaceless EGL

601. By Chris Coulson on 2014-06-06

Work around a lack of pbuffer support in the Mir EGL mesa backend

602. By Chris Coulson on 2014-06-06

Select GLES when binding and creating EGL contexts. The Ubuntu QPA plugin selects OpenGL on desktop, but we depend on GLES for surfaceless contexts

603. By Chris Coulson on 2014-06-06

Don't call eglTerminate in IsUbuntuPhoneOrTablet(), as this terminates a display initialized by Qt, resulting in a crash later on

604. By Chris Coulson on 2014-06-06

Don't crash in the copy constructor for NativeWebKeyboardEvent when we don't have an OS event

605. By Chris Coulson on 2014-06-06

Back out r599 for causing a regression when resizing in Unity 7 desktop

606. By Chris Coulson on 2014-06-06

Bail out of IsUbuntuPhoneOrTablet() early when running on X

607. By Chris Coulson on 2014-06-06

Now that getting the default native display handle from Qt doesn't rely on having a ContentClient instance, use the proper display handle in IsUbuntuPhoneOrTablet() rather than EGL_DEFAULT_DISPLAY, which fails on mesa-egl in Unity 8 (where the compiled in default display is an X11 one)

608. By Chris Coulson on 2014-06-06

Bump Chromium rev to 37.0.2031.2

609. By Chris Coulson on 2014-06-09

Back out r559 for causing LP: #1328000

610. By Chris Coulson on 2014-06-09

Stop using blink::WebInputEvent::KeyDown. Aura and Android both use RawKeyDown + Char

611. By Chris Coulson on 2014-06-09

Fix a regression in unhandled key event handling

612. By Olivier Tilloy on 2014-06-10

Implement the power save blocker interface using the powerd D-Bus API where available.

613. By Chris Coulson on 2014-06-12

Bump Chromium rev to 37.0.2041.4

614. By Chris Coulson on 2014-06-17

GetViewBounds() only works when a RenderViewItem is added to a scene

615. By Chris Coulson on 2014-06-17

Actually make onWindowChanged a slot

616. By Chris Coulson on 2014-06-19

Bump Chromium rev to 37.0.2054.3

617. By Chris Coulson on 2014-06-20

Force Qt focus object changed events when the oxide focused node changes

618. By Alexandre Abreu on 2014-06-24

add devtools support

619. By Chris Coulson on 2014-06-24

Expose a Flickable-like API on the QML WebView, with content{Width,Height,X,Y} and viewport{Width,Height} attributes

620. By Chris Coulson on 2014-06-24

Stop using deprecated compositing paths by using a delegating renderer compositor alongside a synchronous browser compositor. Eventually we will delegate this to the Qml scenegraph compositor, although the extra browser compositor will still be useful for a fallback path (and for other ports of Oxide, should they happen)

621. By Chris Coulson on 2014-06-24

Evict the delegated frame and front buffer for hidden webviews

622. By Chris Coulson on 2014-06-24

Remove spurious printf

623. By Chris Coulson on 2014-06-24

Don't crash on shutdown when devtools isn't enabled

624. By Chris Coulson on 2014-06-24

Bump Chromium rev to 37.0.2062.0

625. By Chris Coulson on 2014-06-25

Get rid of a new static constructor

626. By Chris Coulson on 2014-06-26

Back out r617 for now because it causes a crash

627. By Chris Coulson on 2014-06-26

Force Qt focus object changed events when the oxide focused node changes

628. By Alexandre Abreu on 2014-06-30

add download api support

629. By Chris Coulson on 2014-06-30

Set the timeout to 5 seconds rather than 5000 seconds. It seems this was committed by accident

630. By Chris Coulson on 2014-06-30

Move the webview contents and input event handling from RenderViewItem to WebView, and delete RenderViewItem / qt::RenderWidgetHostView

631. By Chris Coulson on 2014-06-30

Bump Chromium rev to 37.0.2062.3

632. By Alexandre Abreu on 2014-06-30

Make sure to keep the default behavior for external protocol (not handled)

633. By Chris Coulson on 2014-06-30

Clear the counter between tests else the second test always fails

634. By Chris Coulson on 2014-06-30

The sandbox is not Qt-specific, so it should come out of qt/qt.gypi

635. By Chris Coulson on 2014-07-01

Make sure we notify the input method of updates to the cursor and anchor positions, and selection text

636. By Chris Coulson on 2014-07-01

It looks like ImTextBeforeCursor and ImTextAfterCursor only exist in Qt 5.3

637. By Chris Coulson on 2014-07-02

An empty preedit string doesn't mean we should do a commit. Fixes an issue where pressing the backspace key to delete the last character of a composition doesn't actually delete it

638. By Olivier Tilloy on 2014-07-03

Ensure that viewport{Width,Height} and content{Width,Height,X,Y} are integer values and that they add up correctly on devices where the device scale factor and page scale factor are floating-point numbers.

639. By Olivier Tilloy on 2014-07-03

Fix a couple of harmless compiler warnings.

640. By Chris Coulson on 2014-07-03

Don't exclude .git/index. It's required by angle, at least

641. By Chris Coulson on 2014-07-03

Don't crash when destroying a visible RWHV (happens when a render process crashes)

642. By Chris Coulson on 2014-07-03

Bump qt version to 1.2

643. By Chris Coulson on 2014-07-03

Update ContentBrowserClient::CanCreateWindow so its prototype matches that of the base class impl

644. By Chris Coulson on 2014-07-03

Don't patch Chromium's EGL code to bind GLES when creating contexts and calling eglMakeCurrent. As the bound API is thread local and the thread default is GLES, binding GLES is only necessary during initialization (which happens on the main thread where the embedder could have selected a different API)

645. By Chris Coulson on 2014-07-03

Drop add-frame-attached-to-web-contents-observer.patch. We can do this with existing content API

646. By Chris Coulson on 2014-07-03

Add WebFrame::FromRenderFrameHost

647. By Olivier Tilloy on 2014-07-08

Update the power save blocker to use the new Unity.Screen D-Bus API on touch devices.

648. By Chris Coulson on 2014-07-09

Support double-tap-to-zoom

649. By Chris Coulson on 2014-07-09

Now we are not using the GestureRecognizer, tweak ui::ScaleGestureDetector::Config::min_pinch_update_span_delta to get rid of the coarseness when pinching

650. By Chris Coulson on 2014-07-09

Get rid of a compiler warning

651. By Chris Coulson on 2014-07-09

Don't generate touch events for stationary touch points

652. By Chris Coulson on 2014-07-09

Configure the gesture detection to be closer to Android Chrome

653. By Chris Coulson on 2014-07-09

Synthesize pending touch ACK's when swapping views

654. By Olivier Tilloy on 2014-07-09

Delay the loading of data until the webview has been initialized.

655. By Chris Coulson on 2014-07-09

Don't mix up "double tap touch slop" and "double tap slop" from Android's ViewConfiguration when picking Oxide defaults

656. By Chris Coulson on 2014-07-09

Make IdAllocator non-copyable, and add some comments to it

657. By Chris Coulson on 2014-07-09

Don't crash on Qt platforms that don't implement QPlatformNativeInterface. As lack of this prevents us from getting a native display handle, also don't allow any GL implementations and guard against accesses to BrowserProcessMain::GetNativeDisplay()

658. By Chris Coulson on 2014-07-09

Use the --enable-smooth-scrolling flag rather than setting enable_scroll_animator manually

659. By Alexandre Abreu on 2014-07-10

add chrome driver support/build

[Alexandre Abreu (abreu-alexandre)]

I updated the MR, I'd happily get your feedback on this

[lp:~abreu-alexandre/oxide/content-script-injection-main-world]

660. By Chris Coulson on 2014-07-15

Bump Chromium rev to 37.0.2062.20

661. By Chris Coulson on 2014-07-18

Bump Chromium rev to 38.0.2096.0

[Maxim Ermilov (zaspire)]

"oxide" object is not injected after opening another uri in same webview.

review: Needs Fixing

[lp:~abreu-alexandre/oxide/content-script-injection-main-world]

662. By Chris Coulson on 2014-07-22

Bump Chromium rev to 38.0.2101.0

663. By Chris Coulson on 2014-07-22

Include the correct UI assets so that the scrollbars work again

664. By Chris Coulson on 2014-07-23

Add OXIDE_DISABLE_SETUID_SANDBOX and OXIDE_DISABLE_SECCOMP_FILTER_SANDBOX environment variables

665. By Chris Coulson on 2014-07-24

Assert we have a QGuiApplication on startup

[Chris Coulson (chrisccoulson)]

Thanks for working on this. So, I mentioned on Tuesday that there is potentially another approach to this which effectively eliminates the main problem with the current approach (that we're adding an API that makes oxide fundamentally unsafe).

When injecting user scripts, we have the option of emulating greasemonkey (UserScript.emulateGreasemonkey) where the script is injected in to a private scope. We achieve this in Oxide by basically running the following code within the global scope:

"(function(unsafeWindow) {" + script + "})(window)"

(see shared/renderer/oxide_user_script_slave.cc). Potentially, we could extend this to pass in the privileged oxide API for user scripts that are injected in to the main world (it wouldn't be needed for others as we make it available in the global scope). Obviously, you would only be able to do this for scripts in the main world that use UserScript.emulateGreasemonkey - as scripts that don't use this should be injected in to the global scope.

To achieve this, we need changes in a few places:
- ScriptMessageDispatcherRenderer::DidCreateScriptContext() will need to create a ScriptMessageManager for the main world
- ScriptMessageManager shouldn't add the oxide API to the global object when it's associated with the main world
- ScriptMessageManager would need an accessor for returning the oxide API
- We probably need a way to translate between a WebFrame + world ID to a ScriptMessageManager and V8 context
- For this particular case (greasemonkey emulated script in the main world), UserScriptSlave::InjectScripts() would need to be modified to construct a Function object that will run the existing script code, and use blink::WebFrame::callFunctionEvenIfScriptDisabled() to run the function with the oxide API as its argument (see how ScriptMessageManager::OxideLazyGetterInner() works).

How does this sound? I can probably help out with some of these - particularly with any refactoring of the existing code to make this possible

[Maxim Ermilov (zaspire)]

> we're adding an API that makes oxide fundamentally unsafe
exaggeration.
this api will be used only with trusted html code (embedded webview case).
+ applications will run with apparmor profile.

So It's bad reason to over complicate api.

[lp:~abreu-alexandre/oxide/content-script-injection-main-world]

666. By Chris Coulson on 2014-07-28

Stop using ContentMainRunner for starting the browser code - it messes with process global stuff that is unexpected in a public library (eg, clearing the SIGCHLD handler)

667. By Chris Coulson on 2014-07-30

Bump Chromium rev to 38.0.2107.2

668. By Chris Coulson on 2014-07-30

Add USE_GCC_VERSION option for the cross-compile toolchain file

669. By Chris Coulson on 2014-07-30

Revert r668 because of an unrelated change

670. By Chris Coulson on 2014-07-30

Add USE_GCC_VERSION option for the cross-compile toolchain file

671. By Chris Coulson on 2014-07-31

Unbreak fresh builds with the current Chromium version

672. By Chris Coulson on 2014-07-31

Fix an arm link failure caused by a recent webrtc change

673. By Chris Coulson on 2014-07-31

Tidy up the inconsistent BrowserProcessMain interface where we have a mixture of static and non-static functions

674. By Chris Coulson on 2014-07-31

CHECK that CommandLine::Init doesn't fail

675. By Chris Coulson on 2014-07-31

BrowserMainRunner should really be deleted before AtExitManager - it is in Chrome

676. By Chris Coulson on 2014-07-31

Remove comment about moving the MessageLoopForUI::Stop call to BrowserMainParts - it's ok where it is

677. By Chris Coulson on 2014-08-01

Bump Chromium rev to 38.0.2107.3

678. By Chris Coulson on 2014-08-06

CHECK that SIGCHLD isn't being ignored, and ignore SIGPIPE if the application hasn't overridden the default handler for it

679. By Olivier Tilloy on 2014-08-06

Ensure files returned by the file picker exist.

680. By Chris Coulson on 2014-08-06

Bump Chromium rev to 38.0.2114.2

681. By Chris Coulson on 2014-08-12

Don't send a URL longer than content::GetMaxURLChars() when fetching the override for navigator.userAgent, as this fails to deserialize in the browser process, causing it to abort the renderer

682. By Chris Coulson on 2014-08-12

Make sure tst_bug1349510 doesn't trigger a different bug (we'll track the issue with FrameHostMsg_OpenURL separately)

683. By Alexandre Abreu on 2014-08-12

cookie manager api

684. By Alexandre Abreu on 2014-08-12

Add bind ip to devtools API

685. By Chris Coulson on 2014-08-13

Bump Chromium rev to 38.0.2121.3

686. By Chris Coulson on 2014-08-15

Strip down BrowserContextImpl / OTRBrowserContextImpl by introducing 2 new structs for data members shared between a BrowserContext pair (BrowserContextSharedData and BrowserContextSharedIOData), which allows us to drop all of the virtual methods that proxy calls between the 2 BrowserContexts

687. By Chris Coulson on 2014-08-15

Guard BrowserContext entry points to ensure they are only called on the UI thread, and also return a strong reference from BrowserContext::Create so that ownership is explicit now

688. By Chris Coulson on 2014-08-15

Use DCHECK_CURRENTLY_ON where appropriate

689. By Alexandre Abreu on 2014-08-15

Add accept language translation support

690. By Alexandre Abreu on 2014-08-15

update pos

691. By Chris Coulson on 2014-08-19

Bump Chromium rev to 38.0.2125.0

692. By Chris Coulson on 2014-08-20

Build with host_clang == 0

693. By Chris Coulson on 2014-08-20

Fixes for WebContext + WebContextDelegateWorker threading:
- IODelegate is no longer owned by WebContextAdapterPrivate with a raw pointer in OxideQQuickWebContextPrivate. It is now owned by OxideQQuickWebContextPrivate with a weak pointer in WebContextAdapterPrivate.
- UIThreadController is merged in to OxideQQuickWebContextDelegateWorkerPrivate
- IOThreadController is now accessed via QSharedPointer rather than owning this in OxideQQuickWebContextDelegateWorkerPrivate and accessing it from WebContextIODelegate through a raw pointer. This avoids us having to hold a lock for the entire duration of calls in to the worker on Chromium's IO thread, ensuring that setting a new worker or deleting an existing one will no longer block the UI thread if the IO thread is calling in to the existing worker.
- HelperThreadController is shared between IOThreadController and OxideQQuickWebContextDelegateWorkerPrivate, ensuring it remains alive as long as either the UI thread or IO thread entry points remain accessible.
- Add assertions in various places that we are on the correct thread.

694. By Chris Coulson on 2014-08-20

Rename run_qmlscene.sh to run_qmlapp.sh and allow it to run any QML app

695. By Chris Coulson on 2014-08-21

Drop the random operator overloads for allowing bitwise operations on FrameMetadataChangeFlags. Instead, use QFlags in the qml plugin and introduce a new macro for creating bitwise operator overloads for enums (OXIDE_MAKE_ENUM_BITWISE_OPERATORS), partly modelled on Mozilla's MOZ_MAKE_ENUM_CLASS_BITWISE_OPERATORS

696. By Chris Coulson on 2014-08-22

Don't leak the default WebPreferences when setting a new one

697. By Chris Coulson on 2014-08-22

Make revocation checking work by ensuring that NSS has a MessageLoop and URLRequestContext. This is required for Extended Validation to work correctly, and also fixes a bunch of console warnings: "No URLRequestContext for NSS HTTP handler"

698. By Chris Coulson on 2014-08-23

Bump Chromium rev to 38.0.2125.8

699. By Chris Coulson on 2014-08-26

Don't leak GeolocationPermissionRequest if there are no handlers on WebView.geolocationPermissionRequested

700. By Chris Coulson on 2014-08-26

Update for the Chromium GIT migration

701. By Chris Coulson on 2014-08-27

Revert r700 - it breaks the current checkout and doesn't make it possible to check out 38.0.2125.24 either

702. By Chris Coulson on 2014-08-27

Update for the Chromium GIT migration

703. By Chris Coulson on 2014-08-27

Make sure we fetch branch heads

704. By Chris Coulson on 2014-08-27

Bump depot_tools to 8f93f79bcee79072a70cf9c8c392ad140f0cecf1

705. By Chris Coulson on 2014-08-27

Bump Chromium rev to 38.0.2125.24

706. By Chris Coulson on 2014-09-01

Add WebView.securityStatus, WebView.blockedContent, WebView.setCanTemporarilyDisplayInsecureContent, WebView.setCanTemporarilyRunInsecureContent, SecurityStatus, CertificateError and SslCertificate API's. I'm still working on tests for these API's, but this shouldn't block webbrowser-app work

707. By Chris Coulson on 2014-09-01

Fix various issues with the CookieManager code:
- getNextRequestId() always returns zero
- getNextRequestId() depends on undefined beahviour (signed integer overflow)
- CookieMonster calls the callbacks synchronously if nothing needs to be loaded from permanent storage. This means that the behaviour of setting one cookie and waiting for the callback to run before setting the next causes the stack to grow depending on the amount of cookies being set (and also means lots of reentry). Rework it to prevent this behaviour.
- Ensure that the responses at the API level always do happen asynchronously.
- Setting a session cookie always fails because the undefined expiry date gets interpreted as a date far in the past, so Chromium considers the cookie expired.
- Ensure that session cookies returned from the API don't have an expiry date way in the past.

There are a few minor changes to the API too:
- CookieManager.gotCookies does not return a status code now. The only possible errors when reading cookies occur synchronously (ie, the CookieStore is unavailable), so the API just returns an invalid request ID now.
- The addition of CookieManager.getCookies(), which allows you to retrieve all cookies for a URL.
- The expirationdate attribute now works with the built-in Date type rather than having to give it the expiry in milliseconds since the epoch.

The tests have been improved as well to actually test the various cases rather than just calling in to the API and making sure we get a response back

708. By Chris Coulson on 2014-09-01

Set the gettext domain to oxide-qt rather than oxide

709. By Chris Coulson on 2014-09-01

Bump qt version to 1.3

710. By Chris Coulson on 2014-09-02

Bump Chromium rev to 39.0.2138.3

711. By Chris Coulson on 2014-09-02

Some minor tweaks to the CookieManager API:
- Bring back setNetworkCookies as this is being used from C++ code in webbrowser-app. Note that the url parameter is no longer a QString though (it's a QUrl)
- Remove the status parameter entirely from the response to setCookies/setNetworkCookies - the problem with this is that it provides no way to know which cookies failed and which ones were set successfully. Instead, replace it by a new parameter (failedCookies) which returns a list of cookies that failed to set.
- Having setCookies + cookiesSet and getCookies + gotCookies is a bit confusing. These are now changed to setCookies + setCookiesResponse and getCookies + getCookiesResponse.

712. By Chris Coulson on 2014-09-02

Add back support for ms since epoch for the expirationdate, as well as ISO8601 strings

713. By Chris Coulson on 2014-09-02

Clear the temporary mixed content blocker settings for navigations that aren't within the same page (ie, where more than the URL fragment changes) and aren't due to a content-initiated reload. The last one is needed to avoid ending up in an infinite loop

714. By Chris Coulson on 2014-09-03

Don't initialize NSS with a user DB

715. By Chris Coulson on 2014-09-03

Add CookieManager.deleteAllCookies and use this in place of clear-test-cookies-hack.py

716. By Chris Coulson on 2014-09-03

Get rid of run-qmltests.cmake - just use run_qmlapp.sh for this instead

717. By Chris Coulson on 2014-09-03

Add @run-at document-start to TestUtilsSlave.js to hopefully fix some random test failures that look like a race

718. By Chris Coulson on 2014-09-05

Bump Chromium rev to 39.0.2145.4

719. By Chris Coulson on 2014-09-05

Split the event loop implementation in runtests.py in to its own module

720. By Chris Coulson on 2014-09-05

Tidy up the cmake files a bit. The include files are gone - they aren't needed anyway and it wasn't really obvious where new variables should go. The changes also fix a problem where running cmake in a cross-compile set up with a cache file already seems to fail

721. By Chris Coulson on 2014-09-08

Give each test its own datadir. Should hopefully fix LP: #1278315

722. By Chris Coulson on 2014-09-08

Split the HTTP server in to its own module

723. By Chris Coulson on 2014-09-08

Move mock plugin to out/plugins

724. By Chris Coulson on 2014-09-08

OxideQQuickGlobals -> OxideQQuickGlobal

725. By Chris Coulson on 2014-09-09

Add tests for WebView.securityStatus, SecurityStatus and SslCertificate

726. By Chris Coulson on 2014-09-09

Add some tests for navigating away from a HTTPS URL

727. By Chris Coulson on 2014-09-09

ssl-api => ssl

728. By Chris Coulson on 2014-09-09

Revert accidental commit

729. By Chris Coulson on 2014-09-09

Split the certificate tests in to their own test

730. By Chris Coulson on 2014-09-09

Remove bogus chunk from patch

731. By Chris Coulson on 2014-09-10

Verify the certificate fingerprint in tst_SecurityStatus.qml. Whilst we test the SslCertificate API in a separate test, SslCertificate objects have 2 sources in Oxide and we should verify that both of these create them correctly

732. By Chris Coulson on 2014-09-10

Add tests for CertificateError

733. By Chris Coulson on 2014-09-10

Revert accidental change again

734. By Chris Coulson on 2014-09-10

Add tests for WebPreferences.can{Display,Run}InsecureContent

735. By Chris Coulson on 2014-09-11

Make sure we add some checks for WebView.blockedContent

736. By Chris Coulson on 2014-09-11

Ensure WebView.blockedContent is set appropriately after a reload

737. By Chris Coulson on 2014-09-11

Add tests for WebView.setCanTemporarily{Display,Run}InsecureContent() API's

738. By Chris Coulson on 2014-09-11

B'ah, I keep committing this change

739. By Chris Coulson on 2014-09-11

Fix a nullptr deref with non-overridable certificate errors

740. By Chris Coulson on 2014-09-11

Make all subframe and subresource SSL errors non-overridable, as overriding them currently doesn't always result in the security status API indicating a degraded security level

741. By Chris Coulson on 2014-09-11

Add WebContext.hostMappingRules API and add a set of mappings in TestWebContext. This allows us to load content from different domains (that get mapped to localhost), and made it possible to discover bug 1368385

742. By Chris Coulson on 2014-09-11

Test SecurityStatus for insecure content in subframes

743. By Chris Coulson on 2014-09-12

Merge tst_CertificateError_allow/tst_CertificateError_denial

744. By Chris Coulson on 2014-09-12

Test CertificateError.isCancelled

745. By Chris Coulson on 2014-09-12

Sigh...

746. By Chris Coulson on 2014-09-12

Test the mixed content preferences with subframes

747. By Chris Coulson on 2014-09-12

Add subframe tests for the mixed content override

748. By Chris Coulson on 2014-09-12

Fix a regression on clean builds introduced by r720

749. By Chris Coulson on 2014-09-12

Append to QML2_IMPORT_PATH, as Qt adds these in order to the import path list, and the most recently added is used first

750. By Chris Coulson on 2014-09-12

Fix various issues with make package

751. By Chris Coulson on 2014-09-12

Fix a rounding error when calculating the view size in DIP which results in one axis overflowing by a pixel and the other axis underflowing by a pixel

752. By Chris Coulson on 2014-09-16

Back out r751 because it breaks window.screen{X,Y}

753. By Chris Coulson on 2014-09-16

Fix a rounding error when calculating the view size in DIP which results in one axis overflowing by a pixel and the other axis underflowing by a pixel

754. By Alexandre Abreu on 2014-09-16

Make the gesttext domain oxide-qt a proper define

755. By Chris Coulson on 2014-09-16

Bump Chromium rev to 39.0.2159.4

756. By Chris Coulson on 2014-09-16

Fix undefined variable error when running cmake

[Maxim Ermilov (zaspire)]

any news ?

[lp:~abreu-alexandre/oxide/content-script-injection-main-world]

757. By Chris Coulson on 2014-09-17

Auto scroll the focused editable node in to view when the input method appears

[Alexandre Abreu (abreu-alexandre)]

> any news ?
about to update the branch soon

[lp:~abreu-alexandre/oxide/content-script-injection-main-world]

758. By Chris Coulson on 2014-09-18

Make sure we can set domain cookies

759. By Chris Coulson on 2014-09-18

Deprecate WebView.loadingChanged, introduce WebView.loadEvent for delivering LoadEvents and WebView.loadingStateChanged for notifying of changes to WebView.loading

760. By Chris Coulson on 2014-09-18

Add LoadEvent.TypeCommitted, as this is when the security origin of the page changes. It could be useful for clearing any context an application might, related to the security origin of the page

761. By Chris Coulson on 2014-09-19

Don't save "cachedir = None" in .client.cfg when --cache-dir isn't specified, as it breaks checkouts

762. By Alexandre Abreu on 2014-09-19

Merge trunk

763. By Alexandre Abreu on 2014-09-19

fix

764. By Alexandre Abreu on 2014-09-22

small fix

765. By Alexandre Abreu on 2014-09-23

fix crasher w/ gm script compilation; fix issue when reloading

766. By Alexandre Abreu on 2014-09-24

fix tests

767. By Alexandre Abreu on 2014-09-24

remove clutter

768. By Alexandre Abreu on 2014-09-24

allow the handle to escape the scope

769. By Alexandre Abreu on 2014-09-24

fix tests and main frame injection

[Alexandre Abreu (abreu-alexandre)]

should be ready to review

[lp:~abreu-alexandre/oxide/content-script-injection-main-world]

770. By Alexandre Abreu on 2014-09-30

small fix

[Chris Coulson (chrisccoulson)]

Thanks for working on this. Some general comments:
- The special URL "oxide-private://main-world-private" would be fine as just "oxide://main-world"

- I'm not sure using greasemonkey metadata for this is a good idea. Scripts that define @inject_in_main_world in their header effectively change the behaviour of the UserScript.context API (it looks like it would be ignored if UserScript.emulateGreasemonkey is true). Also, @inject_in_main_world behaves differently depending on whether UserScript.emulateGreasemonkey is set or not.

Not only that, but the design is inherently unsafe - imagine the webbrowser grows the ability to install and run custom user scripts in the way that Chrome can with extensions. At the moment, the browser can safely isolate scripts based on their source by injecting them in to their own world (like Chrome, which gives each extension ID its own world). Having this additional metadata in the script header means that scripts can bypass this mechanism entirely.

We should drop this additional metadata and probably just rely on the special URL for UserScript.context. That would also mean you wouldn't need the additional accessors on oxide::UserScript. If UserScript.context is set to the special URL, the script should be injected in to the main world regardless of the state of UserScript.emulateGreasemonkey (but obviously, they wouldn't get the privileged API if UserScript.emulateGreasemonkey isn't set).

Note, this would also require a test to verify that scripts injected in to the main world don't have access to oxide.sendMessage if UserScript.emulateGreasemonkey is set to false.

- For tests in qt/tests/qmltests/api, I try to name them as close as possible to the API or class that they are testing. tst_WebView_scriptMainWorld doesn't really make much sense here - it would probably be better off in qt/tests/qmltests/core (and there are already some related tests there).

There are some additional comments inline too.

review: Needs Fixing

[lp:~abreu-alexandre/oxide/content-script-injection-main-world]

771. By Alexandre Abreu on 2014-10-07

updates

772. By Alexandre Abreu on 2014-10-07

fix small but important piece that had to be removed

773. By Alexandre Abreu on 2014-10-07

fixes

774. By Alexandre Abreu on 2014-10-08

move test file to core/

[Alexandre Abreu (abreu-alexandre)]

Updated,

There is one thing that in the inline comments I did not do (since I did not necessarily agree), it is about the specialized ScriptMessageManagerForWorldId for the main world along w/ a dedicated variable, I have the impression that it will complicate the code a bit and I find it relevant (eventhough it is only used from one context at this point) to have a very dedicated member function there,

=== added file 'qt/tests/qmltests/core/tst_WebView_scriptMainWorld.html'
--- qt/tests/qmltests/core/tst_WebView_scriptMainWorld.html	1970-01-01 00:00:00 +0000
+++ qt/tests/qmltests/core/tst_WebView_scriptMainWorld.html	2014-10-08 13:23:51 +0000
@@ -0,0 +1,12 @@
+<html>
+
+<script>
+</script>
+
+<body>
+  <h1>SCRIPT MAIN WORLD test</h1>
+  <iframe src="empty.html"></iframe>
+  <div id="result"></div>
+</body>
+</html>
+

=== added file 'qt/tests/qmltests/core/tst_WebView_scriptMainWorld.qml'
--- qt/tests/qmltests/core/tst_WebView_scriptMainWorld.qml	1970-01-01 00:00:00 +0000
+++ qt/tests/qmltests/core/tst_WebView_scriptMainWorld.qml	2014-10-08 13:23:51 +0000
@@ -0,0 +1,112 @@
+import QtQuick 2.0
+import QtTest 1.0
+import com.canonical.Oxide 1.0
+import com.canonical.Oxide.Testing 1.0
+
+TestWebView {
+  id: webView
+  focus: true
+  width: 200
+  height: 200
+
+  property var addedScript: null
+
+  Component {
+    id: userScript
+    UserScript {}
+  }
+
+  Component {
+    id: scriptMessageHandler
+    ScriptMessageHandler {}
+  }
+
+  TestCase {
+    name: "WebView_scriptMainWorld"
+    when: windowShown
+
+    function init() {
+      if (webView.addedScript) {
+        webView.context.removeUserScript(addedScript)
+        webView.addedScript = null
+      }
+    }
+
+    function resultUpdated(testapi) {
+      var res;
+      try {
+        res = testapi.evaluateCode("\
+  var el = document.getElementById(\"result\");\
+  if (!el) throw Exception();\
+  return el.innerHTML;", true);
+      } catch(e) {
+        fail("#result not found: " + JSON.stringify(e));
+      }
+      return res;
+    }
+
+    function test_WebView_canInjectInMainWorldWithOxideApi() {
+      webView.addedScript = userScript.createObject(null, {
+        context: "oxide://main-world",
+        emulateGreasemonkey: true,
+        url: "tst_WebView_scriptMainWorld_user_script.js"});
+      webView.context.addUserScript(webView.addedScript);
+
+      webView.url = "http://testsuite/tst_WebView_scriptMainWorld.html";
+      verify(webView.waitForLoadSucceeded(),
+             "Timed out waiting for successful load");
+
+      var testApi = webView.getTestApi();
+      var res = webView.waitFor(function () { return resultUpdated(testApi); })
+
+      compare(res, "Main world content script found oxide.sendMessage",
+              "Unexpected result message");
+    }
+
+    function test_WebView_verifyRegularMainWorldScriptDoesNotHaveAccessToOxideApi() {
+      webView.addedScript = userScript.createObject(null, {
+        context: "oxide://main-world",
+        url: "tst_WebView_scriptMainWorld_user_script.js"});
+      webView.context.addUserScript(webView.addedScript);
+
+      webView.url = "http://testsuite/tst_WebView_scriptMainWorld.html";
+      verify(webView.waitForLoadSucceeded(),
+             "Timed out waiting for successful load");
+
+      var testApi = webView.getTestApi();
+      var res = webView.waitFor(function () { return resultUpdated(testApi); })
+
+      compare(res, "Main world content script DID NOT found oxide.sendMessage",
+              "Unexpected result message");
+    }
+
+    function test_WebView_receiveMessageFromMainWorldUserscript() {
+      webView.addedScript = userScript.createObject(null, {
+        context: "oxide://main-world",
+        emulateGreasemonkey: true,
+        url: "tst_WebView_scriptMainWorld_user_script.js"});
+
+      var received = null;
+      var frame = webView.rootFrame;
+      var handler = scriptMessageHandler.createObject(null, {
+        msgId: "from-user-script",
+        contexts: ["oxide://main-world"],
+        callback: function(msg, frame) {
+          received = msg.args;
+        }
+      });
+
+      frame.addMessageHandler(handler);
+
+      webView.context.addUserScript(webView.addedScript);
+      webView.url = "http://testsuite/tst_WebView_scriptMainWorld.html";
+
+      verify(webView.waitForLoadSucceeded(),
+             "Timed out waiting for successful load");
+
+      verify(received != null, "Did not receive message from the main frame's userscript");
+      compare(received.data, "mydata" , "Did not receive message from the main frame's userscript");
+    }
+  }
+}
+

=== added file 'qt/tests/qmltests/core/tst_WebView_scriptMainWorld_user_script.js'
--- qt/tests/qmltests/core/tst_WebView_scriptMainWorld_user_script.js	1970-01-01 00:00:00 +0000
+++ qt/tests/qmltests/core/tst_WebView_scriptMainWorld_user_script.js	2014-10-08 13:23:51 +0000
@@ -0,0 +1,10 @@
+try {
+  oxide.sendMessage("from-user-script", {"data": "mydata", "values": [1, 2, 3]});
+  oxide.addMessageHandler("from-user-script", function(msg) {});
+  document.getElementById("result").innerHTML =
+	"Main world content script found oxide.sendMessage";
+} catch(e) {
+  document.getElementById("result").innerHTML =
+	"Main world content script DID NOT found oxide.sendMessage";
+}
+

=== modified file 'shared/common/oxide_constants.cc'
--- shared/common/oxide_constants.cc	2014-06-02 15:35:10 +0000
+++ shared/common/oxide_constants.cc	2014-10-08 13:23:51 +0000
@@ -27,3 +27,8 @@
 const char kIncognito[] = "incognito";
 
 } // namespace switches
+
+namespace oxide {
+const int kMainWorldId = 0;
+const char kMainWorldContextUrl[] = "oxide://main-world";
+} // namespace oxide

=== modified file 'shared/common/oxide_constants.h'
--- shared/common/oxide_constants.h	2014-06-02 15:35:10 +0000
+++ shared/common/oxide_constants.h	2014-10-08 13:23:51 +0000
@@ -29,4 +29,9 @@
 
 } // namespace switches
 
+namespace oxide {
+extern const int kMainWorldId;
+extern const char kMainWorldContextUrl[];
+}
+
 #endif // _OXIDE_SHARED_COMMON_CONSTANTS_H_

=== modified file 'shared/renderer/oxide_isolated_world_map.cc'
--- shared/renderer/oxide_isolated_world_map.cc	2014-02-21 17:37:53 +0000
+++ shared/renderer/oxide_isolated_world_map.cc	2014-10-08 13:23:51 +0000
@@ -22,6 +22,8 @@
 #include "base/lazy_instance.h"
 #include "base/logging.h"
 
+#include "shared/common/oxide_constants.h"
+
 namespace oxide {
 
 namespace {
@@ -39,6 +41,9 @@
 int IsolatedWorldMap::IDFromURL(const GURL& url) {
   CHECK(url.is_valid());
 
+  if (url == GURL(kMainWorldContextUrl))
+    return kMainWorldId;
+
   WorldIDMapIterator it = g_isolated_world_map.Get().find(url);
   if (it != g_isolated_world_map.Get().end()) {
     return it->second;
@@ -52,6 +57,9 @@
 
 // static
 GURL IsolatedWorldMap::URLFromID(int id) {
+  if (id == kMainWorldId)
+    return GURL(kMainWorldContextUrl);
+
   for (WorldIDMapIterator it = g_isolated_world_map.Get().begin();
        it != g_isolated_world_map.Get().end(); ++it) {
     if (it->second == id) {

=== modified file 'shared/renderer/oxide_script_message_dispatcher_renderer.cc'
--- shared/renderer/oxide_script_message_dispatcher_renderer.cc	2014-09-09 22:43:03 +0000
+++ shared/renderer/oxide_script_message_dispatcher_renderer.cc	2014-10-08 13:23:51 +0000
@@ -24,11 +24,15 @@
 #include "base/memory/ref_counted.h"
 #include "content/public/renderer/render_frame.h"
 #include "content/public/renderer/render_thread.h"
+#include "content/public/renderer/render_view.h"
 #include "ipc/ipc_message.h"
 #include "ipc/ipc_message_macros.h"
+#include "third_party/WebKit/public/web/WebLocalFrame.h"
 #include "third_party/WebKit/public/web/WebFrame.h"
+#include "third_party/WebKit/public/web/WebView.h"
 #include "url/gurl.h"
 
+#include "shared/common/oxide_constants.h"
 #include "shared/common/oxide_messages.h"
 #include "shared/common/oxide_script_message_request.h"
 
@@ -50,10 +54,6 @@
 void ScriptMessageDispatcherRenderer::WillReleaseScriptContext(
     v8::Handle<v8::Context> context,
     int world_id) {
-  if (world_id < 1) {
-    return;
-  }
-
   v8::HandleScope handle_scope(context->GetIsolate());
 
   for (ScriptMessageManagerVector::iterator it = script_message_managers_.begin();
@@ -159,6 +159,19 @@
   CHECK(rv.second);
 }
 
+linked_ptr<ScriptMessageManager>
+ScriptMessageDispatcherRenderer::ScriptMessageManagerForWorldId(int world_id) {
+  linked_ptr<ScriptMessageManager> message_manager;
+  for (ScriptMessageManagerVector::iterator it = script_message_managers_.begin();
+       it != script_message_managers_.end();
+       ++it) {
+    if ((*it)->frame() == render_frame() && (*it)->world_id() == world_id) {
+      message_manager = *it;
+    }
+  }
+  return message_manager;
+}
+
 ScriptMessageDispatcherRenderer::~ScriptMessageDispatcherRenderer() {
   // RenderFrameObserver has already cleared it's pointer to our RenderFrame
   for (ScriptMessageDispatcherMap::iterator it = g_dispatcher_map.Get().begin();
@@ -182,10 +195,6 @@
 void ScriptMessageDispatcherRenderer::DidCreateScriptContext(
     v8::Handle<v8::Context> context,
     int world_id) {
-  if (world_id < 1) {
-    return;
-  }
-
   script_message_managers_.push_back(
       linked_ptr<ScriptMessageManager>(new ScriptMessageManager(render_frame(),
                                                                 context,

=== modified file 'shared/renderer/oxide_script_message_dispatcher_renderer.h'
--- shared/renderer/oxide_script_message_dispatcher_renderer.h	2014-05-01 15:25:31 +0000
+++ shared/renderer/oxide_script_message_dispatcher_renderer.h	2014-10-08 13:23:51 +0000
@@ -45,6 +45,8 @@
 
   static ScriptMessageDispatcherRenderer* FromWebFrame(blink::WebFrame* frame);
 
+  linked_ptr<ScriptMessageManager> ScriptMessageManagerForWorldId(int world_id);
+
   void DidCreateScriptContext(v8::Handle<v8::Context> context,
                               int world_id);
 

=== modified file 'shared/renderer/oxide_script_message_manager.cc'
--- shared/renderer/oxide_script_message_manager.cc	2014-09-15 10:45:20 +0000
+++ shared/renderer/oxide_script_message_manager.cc	2014-10-08 13:23:51 +0000
@@ -29,6 +29,8 @@
 #include "third_party/WebKit/public/web/WebScopedMicrotaskSuppression.h"
 #include "ui/base/resource/resource_bundle.h"
 
+#include "shared/common/oxide_constants.h"
+
 #include "oxide_isolated_world_map.h"
 #include "oxide_script_message_handler_renderer.h"
 #include "oxide_script_message_request_impl_renderer.h"
@@ -94,11 +96,9 @@
       mm->Value())->OxideLazyGetterInner(property, info);
 }
 
-void ScriptMessageManager::OxideLazyGetterInner(
-    v8::Local<v8::String> property,
-    const v8::PropertyCallbackInfo<v8::Value>& info) {
-  v8::Isolate* isolate = info.GetIsolate();
-  v8::HandleScope handle_scope(isolate);
+v8::Handle<v8::Object> ScriptMessageManager::GetOxideApiObject(
+      v8::Isolate* isolate) {
+  v8::EscapableHandleScope handle_scope(isolate);
 
   base::StringPiece raw_src(
       ui::ResourceBundle::GetSharedInstance().GetRawDataResource(
@@ -121,7 +121,7 @@
   v8::Local<v8::Function> function(script->Run().As<v8::Function>());
   if (try_catch.HasCaught()) {
     LOG(ERROR) << "Caught exception when running script";
-    return;
+    return v8::Handle<v8::Object>();
   }
 
   v8::Local<v8::External> local_data(closure_data_.NewHandle(isolate));
@@ -152,9 +152,20 @@
   }
   if (try_catch.HasCaught()) {
     LOG(ERROR) << "Caught exception when running script function";
-    return;
+    return v8::Handle<v8::Object>();
   }
 
+  return handle_scope.Escape(exports);
+}
+
+void ScriptMessageManager::OxideLazyGetterInner(
+    v8::Local<v8::String> property,
+    const v8::PropertyCallbackInfo<v8::Value>& info) {
+  v8::Isolate* isolate = info.GetIsolate();
+  v8::HandleScope handle_scope(isolate);
+
+  v8::Handle<v8::Object> exports(GetOxideApiObject(isolate));
+
   v8::Handle<v8::Value> val = info.This();
   if (val->IsObject()) {
     v8::Handle<v8::Object> object = v8::Handle<v8::Object>::Cast(val);
@@ -347,10 +358,12 @@
   closure_data_.reset(isolate(), v8::External::New(isolate(), this));
   v8::Local<v8::External> local_data(closure_data_.NewHandle(isolate()));
 
-  v8::Local<v8::Object> global(context->Global());
-  global->SetAccessor(v8::String::NewFromUtf8(isolate(), "oxide"),
-                      OxideLazyGetter, NULL,
-                      local_data);
+  if (world_id_ != kMainWorldId) {
+    v8::Local<v8::Object> global(context->Global());
+    global->SetAccessor(v8::String::NewFromUtf8(isolate(), "oxide"),
+                        OxideLazyGetter, NULL,
+                        local_data);
+  }
 }
 
 ScriptMessageManager::~ScriptMessageManager() {}

=== modified file 'shared/renderer/oxide_script_message_manager.h'
--- shared/renderer/oxide_script_message_manager.h	2014-02-28 12:53:41 +0000
+++ shared/renderer/oxide_script_message_manager.h	2014-10-08 13:23:51 +0000
@@ -56,6 +56,8 @@
 
   GURL GetContextURL() const;
 
+  int world_id() const { return world_id_; }
+
   content::RenderFrame* frame() const { return frame_; }
 
   v8::Isolate* isolate() const { return isolate_; }
@@ -70,6 +72,8 @@
 
   ScriptMessageHandlerRenderer* GetHandlerForMsgID(const std::string& msg_id);
 
+  v8::Handle<v8::Object> GetOxideApiObject(v8::Isolate* isolate);
+
  private:
   friend class ScriptMessageRequestImplRenderer;
   typedef std::map<std::string, linked_ptr<ScriptMessageHandlerRenderer> > ScriptMessageHandlerMap;

=== modified file 'shared/renderer/oxide_user_script_slave.cc'
--- shared/renderer/oxide_user_script_slave.cc	2014-06-02 15:35:10 +0000
+++ shared/renderer/oxide_user_script_slave.cc	2014-10-08 13:23:51 +0000
@@ -29,15 +29,19 @@
 #include "third_party/WebKit/public/platform/WebURLRequest.h"
 #include "third_party/WebKit/public/web/WebDataSource.h"
 #include "third_party/WebKit/public/web/WebLocalFrame.h"
+#include "third_party/WebKit/public/web/WebScopedMicrotaskSuppression.h"
 #include "third_party/WebKit/public/web/WebScriptSource.h"
 #include "third_party/WebKit/public/web/WebSecurityOrigin.h"
 #include "url/gurl.h"
+#include "v8/include/v8.h"
 
 #include "shared/common/oxide_constants.h"
 #include "shared/common/oxide_messages.h"
 #include "shared/common/oxide_user_script.h"
 
 #include "oxide_isolated_world_map.h"
+#include "oxide_script_message_dispatcher_renderer.h"
+#include "oxide_script_message_manager.h"
 
 namespace oxide {
 
@@ -105,6 +109,76 @@
   content::RenderThread::Get()->RemoveObserver(this);
 }
 
+void UserScriptSlave::InjectGreaseMonkeyScriptInMainWorld(
+      blink::WebLocalFrame* frame,
+      const blink::WebScriptSource& script_source) {
+
+  ScriptMessageDispatcherRenderer * dispatcher_renderer =
+      ScriptMessageDispatcherRenderer::FromWebFrame(frame);
+  DCHECK(dispatcher_renderer != NULL);
+
+  linked_ptr<ScriptMessageManager> message_manager =
+      dispatcher_renderer->ScriptMessageManagerForWorldId(kMainWorldId);
+  DCHECK(message_manager != NULL);
+  if (!message_manager.get()) {
+    LOG(ERROR) << "Could not get a proper message manager for frame: "
+               << frame
+	       << " while trying to inject script in main world";
+    return;
+  }
+
+  v8::Isolate* isolate = message_manager->isolate();
+  v8::HandleScope handle_scope(isolate);
+  v8::Context::Scope context_scope(message_manager->GetV8Context());
+
+  v8::Local<v8::String> wrapped_script_head(v8::String::NewFromUtf8(
+      isolate,
+      "(function(oxide) {\n"
+  ));
+
+  v8::Local<v8::String> src(
+      v8::String::NewFromUtf8(
+          isolate,
+	  script_source.code.utf8().c_str()));
+  DCHECK(!src.IsEmpty() && src->Length() > 0);
+
+  v8::Local<v8::String> wrapped_script_tail(
+      v8::String::NewFromUtf8(isolate, "\n})"));
+
+  v8::Local<v8::String> wrapped_src(
+      v8::String::Concat(wrapped_script_head,
+          v8::String::Concat(src, wrapped_script_tail)));
+
+  v8::Local<v8::Script> script(
+      v8::Script::Compile(wrapped_src));
+
+  v8::TryCatch try_catch;
+  v8::Local<v8::Function> function(script->Run().As<v8::Function>());
+  if (try_catch.HasCaught()) {
+    LOG(ERROR) << "Caught exception when running script: "
+               << *v8::String::Utf8Value(try_catch.Message()->Get());
+    return;
+  }
+
+  v8::Handle<v8::Value> argv[] = {
+      message_manager->GetOxideApiObject(message_manager->isolate())
+  };
+
+  {
+    blink::WebScopedMicrotaskSuppression mts;
+    frame->callFunctionEvenIfScriptDisabled(
+        function,
+        message_manager->GetV8Context()->Global(),
+        arraysize(argv),
+        argv);
+  }
+  if (try_catch.HasCaught()) {
+    LOG(ERROR) << "Caught exception when calling script: "
+               << *v8::String::Utf8Value(try_catch.Message()->Get());
+    return;
+  }
+}
+
 void UserScriptSlave::InjectScripts(blink::WebLocalFrame* frame,
                                     UserScript::RunLocation location) {
   blink::WebDataSource* data_source = frame->provisionalDataSource() ?
@@ -151,6 +225,16 @@
     }
 
     blink::WebScriptSource source(blink::WebString::fromUTF8(content));
+
+    if (script->context() == GURL(kMainWorldContextUrl)) {
+      if (script->emulate_greasemonkey()) {
+        InjectGreaseMonkeyScriptInMainWorld(frame, source);
+      } else {
+        frame->executeScript(source);
+      }
+      continue;
+    }
+
     int id = GetIsolatedWorldID(script->context(), frame);
     frame->executeScriptInIsolatedWorld(id, &source, 1, 0);
   }

=== modified file 'shared/renderer/oxide_user_script_slave.h'
--- shared/renderer/oxide_user_script_slave.h	2014-04-17 12:57:07 +0000
+++ shared/renderer/oxide_user_script_slave.h	2014-10-08 13:23:51 +0000
@@ -32,6 +32,7 @@
 
 namespace blink {
 class WebLocalFrame;
+class WebScriptSource;
 }
 
 namespace oxide {
@@ -57,6 +58,10 @@
                                 blink::WebLocalFrame* frame);
   void OnUpdateUserScripts(base::SharedMemoryHandle handle);
 
+  void InjectGreaseMonkeyScriptInMainWorld(
+      blink::WebLocalFrame* frame,
+      const blink::WebScriptSource& script_source);
+
   Vector user_scripts_;
 
   DISALLOW_COPY_AND_ASSIGN(UserScriptSlave);