| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
| |
We create an EGL_KHR_fence_sync on the chromium thread
but try to verify if the wait function implementation
is available by testing for the EGL_KHR_reusable_sync
extension in the scene graph thread.
Since the latter extension is not supported by most of
our devices, we never actually waited for the rendering
to finish.
Change-Id: Ied829d4035d42899cfb3f86de018bc48e61c0cc0
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
|
|
|
|
|
|
|
|
|
|
| |
This prevents the crash when the application requests
persistent storage access using the Quota API (e.g. on
http://codepen.io/matt-west/full/CrfKh). The request has been
disallowed for now, because it should rely on user permission.
Change-Id: I4ae057c9485d3f06f45a637c7eeda9dd69fe6b54
Reviewed-by: Andras Becsi <andras.becsi@theqtcompany.com>
|
|
|
|
|
|
|
|
|
| |
Cookies were not flushed on application exit because objects deleted
lazy on the UI-thread never had their destructors called because we
did not flush the queue on exit.
Change-Id: If53cdb5547e15a9f39de8b99b1da313e307b1c90
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
|
|
|
|
|
|
|
|
|
|
|
| |
This fix deploys the following files into qtbase without make install:
- qtwebengine_resources.pak
- translations/qtwebengine_locales/*.pak
- icudtl.dat
- plugins/qtwebengine/libffmpegsumo.so
Change-Id: Id2c02f03fb01571942d3135b215ebb8ad7a3221b
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
| |
Since we use the system ICU we do not need the extra data file.
This fixes a crash caused by missing icudtl.dat file.
Change-Id: Ic7a6ad83d6595ea3a89c736c281499636b05945e
Reviewed-by: Zeno Albisser <zeno.albisser@digia.com>
|
|
|
|
|
| |
Change-Id: I67d0040c2e9f00e8a7b842ac1228212c0e93873c
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We are mapping a single QTouchEvent to
multiple MotionEventQt for Chromium.
For gesture recognition it is important
that these motion events are being processed
in a sorted manner, as a move event might
trigger a different gesture depending on
how many ACTION_DOWN or ACTION_POINTER_DOWN
were received before.
It is particularly illegal to process an ACTION_MOVE
with multiple touch points without having received
a ACTION_POINTER_DOWN before.
Change-Id: I75d22dd845774a14b5f590e0e0ce46263c4a49f4
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
|
| |
Callers inside Chromium expect an implementation of AccessTokenStore to exist.
This fixes crashes with the geolocation-based APIs, exposed on e.g. google.com
when faking an Android UA (as well as other places, probably).
Change-Id: I45e6b483e096d5165fefd86927cdf34e799cc4d9
Reviewed-by: Pierre Rossi <pierre.rossi@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
ANGLE isn't thread-safe and we're already taking a risk by letting
Chromium do texture upload from the GPU thread with a shared GL
context.
Any kind of drawing happening concurrently between the GPU thread and
the Qt scene graph thread might currently end up drawing on the same
device context and this is creating visual artifacts and Direct3D
errors.
We're already disabling WebGL but we should also disable accelerated
2D canvas rendering to lower that risk. We should ultimately disable
ANGLE from within Chromium or fix multithreading issues to properly
support GL rendering scenarios across threads.
Task-number: QTBUG-41627
Change-Id: I37a7a778f0e1066a454f51bb38e98df4dea8767e
Reviewed-by: Pierre Rossi <pierre.rossi@gmail.com>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@theqtcompany.com>
|
|\
| |
| |
| | |
Change-Id: I01615a492c958bf123deeef5d2ddf9fb01b1a9d0
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We should save our cache in the configured CacheLocation, and ensure
we are prepared for multiple BrowserContext by caching to a Default
subdir. This matches how chromium caches on linux.
Change-Id: Ibe914fca35435ea1658565be87747cfabc71e97b
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
There is currently no convenient way to deploy QtWebEngine into an application
bundle on OSX. macdeployqt copies frameworks into a .app bundle's Frameworks
directory but this makes no sense unless all the needed files are also
distributed with the bundle.
This patch moves:
- The ffmpegsumo.so library into Libraries/
- Locale .pak files, qtwebengine_resources.pak and icudtl.dat into Resources/
- QtWebEngineProcess into its own .app bundle, itself into Helpers/
QMAKE_BUNDLE_DATA is used to copy files into the bundle while INSTALLS is
used when installing normally. A LOCALE_LIST is explicitly listed since
QMAKE_BUNDLE_DATA can't handle the * glob to match all source files.
Change-Id: I5c0df57b4b9e93f9cce34a74a6e024bf90d37b5c
Task-number: QTBUG-41611
Reviewed-by: Zeno Albisser <zeno.albisser@digia.com>
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|/
|
|
|
|
|
|
|
|
| |
core_module.pro is the one responsible for deployment of the binary so
it makes sense to also let it handle the other files to install.
This leaves core.pro to only handle sub-projects dependencies.
Change-Id: Ibe433eceec7fb0a9f806dfa89dccb0bedeaccc59
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
|
| |
This avoids using a General matrix for tile transformations which are
usually only translations in our case.
Change-Id: Ic947d7ef9b78de6d5d16ea718cc03255e38a8097
Reviewed-by: Michael Bruning <michael.bruning@digia.com>
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
| |
Use explicit url in quick API too and use GetLastCommittedURL instead of
GetVisibleURL for getting the expected urls in tests.
Change-Id: If3251323645979643f3dc0e16491fe35b584f51c
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
| |
Chromium is only supported on ARMv6 and ARMv7 for ARM architecture.
Change-Id: I0493c9ba2baf4a537f759286622056093c0d4a25
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
| |
This prevents the crash when a webpage requests geolocation permission.
Change-Id: I7331b869ec1439d678cee7941691289220718f3d
Reviewed-by: Zeno Albisser <zeno.albisser@digia.com>
|
|
|
|
|
| |
Change-Id: I972d3560d6aa2640a2702f219cf802d2476b1c6e
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
| |
That code was borrowed straight from Chromium and used the Chromium
coding style.
Change-Id: I212aea23f19056b901cbc770640faed4998f02fd
Reviewed-by: Zeno Albisser <zeno.albisser@digia.com>
|
|
|
|
|
|
|
|
|
| |
This is required so that embedders can implement custom error pages,
and is also needed for tests, where error-page loading has side-effects
on subsequent page loads.
Change-Id: I5ae276a58864a2fa9d7b608bea3340b482a92f4e
Reviewed-by: Peter Varga <pvarga@inf.u-szeged.hu>
|
|
|
|
|
|
|
|
|
| |
This should be removed once desktop notifications
have been implemented.
Task-number: QTBUG-41893
Change-Id: Ia004ee8946ae4441027161f5dddf05d304df8b1b
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
|
| |
We did not load any localization data on OS X until now,
so the error page did not have any strings, and some
tests asserted because of missing localization.
Change-Id: Icb6de1f7e3aaa6815483d379efaf00e8ed6a1dbf
Reviewed-by: Michael Bruning <michael.bruning@digia.com>
|
|
|
|
|
|
|
| |
Doing so causes a check to fail in URLRequest::BeforeRequestComplete.
Change-Id: I487042b922b3f357bd8d9579167d9b372963ff29
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
| |
As it turns out failing to do so can mess up the page state.
Change-Id: I8cc785b01c73bde5443bdb78554d7b3809d96afe
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
|
| |
The current delegate just downloads files to the default download
location. By not setting the delegate for downloads, we effectively
disable them.
Change-Id: I1b416f76ac9b4395ee0bcf9986134f4bf056d8ed
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
|
|
|
| |
This is used to represent tables, lists and trees to screen readers.
There are some properties still not implemented, notably selection which
is also not available in blink.
Change-Id: I57053881445e39a4967a1aabd85a594271402f63
Reviewed-by: Zeno Albisser <zeno.albisser@digia.com>
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
|
|
|
|
|
|
|
|
|
| |
It's faster to use QLatin1String in places where we don't actually need
a QString object (like for comparison, or when creating a string with
QStringBuilder).
Change-Id: I9b8e28040dc2b79a767e15015b85822d06141bbe
Reviewed-by: Zeno Albisser <zeno.albisser@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The requestedUrl function didn't return empty URL even if
the empty URL was really requested. It was assumed if
GetOriginalRequestURL returned empty string that means the requested url was
not set in the navigation entry.
This fix handles that case when empty url is set in the navigation
entry as requested url. If the navigation entry is in pending state that means the
request url has not been set yet thus the actual URL should be returned.
Change-Id: Ic2eff5c487686f7c0e349a7a34a86b80551a002f
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
|
|
| |
If the error page loading is interrupted by loading of another page the
signals of the second page should not be suppressed.
Change-Id: I537daabf80ffed005acaaf2a6019e52a3250c523
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
Reviewed-by: Zeno Albisser <zeno.albisser@digia.com>
Reviewed-by: Adam Kallai <kadam@inf.u-szeged.hu>
|
|
|
|
|
|
|
|
|
|
| |
Add the missing default_encoding to the web preferences.
Use a 0ms timer instead of the 100ms in order to apply the settings
as soon as possible and avoid waiting mechanisms in the test.
Change-Id: I12f1a6dd68b29c2b287b160eea1ad5a5fb169c14
Reviewed-by: Zeno Albisser <zeno.albisser@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The build configuration requires the following changes:
- toolkit_views=1: for native_theme_aura, as otherwise
nine_image_painter_factory.cc would be excluded.
- use_system_harfbuzz=0: we do not want to rely on this
being provided on the system, therefore being explicit.
- want_separate_host_toolset=1: assures that icuuc gyp
target will be available.
The file gl_surface_ozone.cc must be added to the exclusion
list to avoid duplicate symbols with gl_surface_at.cpp.
The defines EGL_API_FB and LINUX must be set to 1,
to explicitly define that we are using EGL and are building
for a linux target.
Change-Id: If6722ecbfd0655fe7852d52a606f27c18304f717
Reviewed-by: Zeno Albisser <zeno.albisser@digia.com>
|
|
|
|
|
|
|
|
| |
Adding a local LoadLibrary function to wrap the new
call to base::LoadNativeLibrary and do the error handling.
Change-Id: Iaab23d3731508fbbf2dadb27d34c53fd65095830
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
|
|
| |
The Ozone platform base class has received several
new pure virtual functions that needed to be implemented.
Most of the functions are implemented according to
how it is done in OzonePlatformGbm.
Change-Id: Ibab75ae0891b67d03f705da611949ed6cf75ad9f
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
| |
This makes deployment easier and was also what was done in QtWebKit
for the WebKit2 WebProcess executable
Change-Id: I2245c1f3337f19509c08c08c0baf40dc0a964890
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
PbufferGLSurfaceWGL is a reference counted class, hence deleting
it manually triggers an assert in debug builds if Release is not
called or the reference count is non-zero.
Using a scoped_refptr prevents this and deletes the object
correctly.
Change-Id: I7e5ff783afc367e06d872b8d118c2ff83627b836
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Until now we relied on the PATH to pick the right clang version on
OS X which turned out to be a source of issues on build nodes that
have a lot of old cruft lying around which was picked up by gyp.
Set make_clang_dir so that gyp uses the configured clang version
from Qt and remove the env manipulation from our gyp script.
Change-Id: I4ac679ea56fa874eecaa578aad77b462445c0caa
Reviewed-by: Michael Bruning <michael.bruning@digia.com>
|
|
|
|
|
|
|
|
| |
Enabling rendering into a WGL backbuffer, in addition to the EGL/angle
one.
Change-Id: I8f2e3f5ecf52b6db22712b1129059f462725a256
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
|
|
|
| |
This patch fixes widgets rendering on HiDPI displays and notifies
about the a potential device pixel ratio change when the window is
moved. This addresses the blurriness when a web view is moved
between an attached LoDPI screen and HiDPI screen resulting from
the device pixel ratio not being updated.
Change-Id: I97af0199c07deac8bbe4688c665aaa6df1608ae9
Reviewed-by: Michael Bruning <michael.bruning@digia.com>
|
|
|
|
|
|
|
|
| |
If we get a replacement content, we can see a DidFinishLoad event for a frame.
This error page should be ignored based on the frame.
Change-Id: I3e1cd1773e8c5fc608605197c957011ddf258123
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
|
|
|
|
|
| |
Change-Id: Ia75177e8759c5ca10f74cc635d600999fb8205fb
Reviewed-by: Michael Bruning <michael.bruning@digia.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Since the default arm_tune option applied by chromium is not
recognized by our gcc toolchain set arm_tune to empty string
to prevent chromium from setting a default. Our toolchains
set the arm version which already implies the corresponding
mtune option anyway.
Change-Id: I0291bd4f0e5706597d37f914e4eeefbd75397bbf
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
|
|
|
|
|
|
|
|
|
|
|
| |
This logic was removed in 3a30ed4ecc9c828641daef85f88f93baf78826b6
but was still needed by the velocity tracker used for fling gestures.
Also disable swipe gestures in the gesture config since they have
unwanted side-effects on the gesture detection and we don't use them.
Change-Id: Icbb44a4c27e2cd243d631484d03a956e9dce64dc
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
| |
Change-Id: I94ccd40dbba9a3c6a962ebaca6d3d5ab316de8d5
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Configuring Qt with -force-debug-info -no-strip should carry the
needed cflags to ninja through gyp.
Also prevent stripping the library in release with force_debug_info.
qmake will already call strip --strip-unneeded by default when
running make install for release builds so remove that condition
completely from core_module.pro.
Change-Id: Ic31fca5b4365d61161f2500c2de011faa3e19c14
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
|
|
| |
Use the WebFrame's more reliable contentAsMarkup() and
contentAsText() methods instead of going down to the Document
element.
Change-Id: I332e24375e074124099b59e8a784a0a038c6f70b
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
| |
Change-Id: I1edd16e6875198d7480897e7b2e2fbf68b245883
Reviewed-by: Jocelyn Turcotte <jocelyn.turcotte@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Both QWebEnginePage::createWindow and QQuickWebEngineNewViewRequest
behave as if JavascriptCanOpenWindows was disabled by default, when
the application isn't handling new window signals.
LocalStorageEnabled should have no negative impact to be enabled
beside the used disk space, but is still less inconvenient than
cookies or disk cache, which are also enabled by default.
Change-Id: I3f5343c2410fac50bec08b386defe22baf3df440
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
| |
They are disabled by default at the moment, enable them conditionally
to a touch screen being available.
Change-Id: I5827fe2326b7bc019fa5613121d1cfa666608e52
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
| |
Most importantly, get rid of QOpenGLContextPrivate::globalShareContext
which will be removed before the final.
Change-Id: I5796fa822d2962c21fd8b6d8d3c196686d536e3d
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The --log-level switch sets the log level of the chromium messages. It affects
the browser and web processes but has no impact on the Qt messages.
The following levels can be set:
0 -> LOG_INFO
1 -> LOG_WARNING
2 -> LOG_ERROR
3 -> LOG_FATAL
Log level 3 set per default (fatal messages are shown only) since
lower level messages might be irrelevant for those who use QtWebEngine API.
Change-Id: Ie2ba10718d3c86bbfe3d847471b3facb346be446
Reviewed-by: Andras Becsi <andras.becsi@digia.com>
|