| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
| |
This reverts commit d4c621f6a6b87f2a86069fa393b9f7c4f9e7b9ad,
which seems to break QtWebEngine in B2Qt devices
Task-number: QTBUG-63341
Change-Id: I2224b8c68c85baa602ed3d2aab88cd2cf622eeb8
Reviewed-by: Michal Klocek <michal.klocek@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
After the recent fix for rendering intersecting quads, DelegatedFrameNode no
longer builds scene graph nodes for DrawQuads that are outside the visible area.
This means that the structure of the scene graph now depends on the size of the
visible area, however the logic for deciding whether to update or rebuild the
scene graph was not updated to reflect this fact. As a result we may try to
update e.g. a QSGImageNode as if it were a QSGRectangleNode leading to a crash.
Task-number: QTBUG-62112
Change-Id: I6e2e9dee4238d208fc2be98669281c2d4d4962d7
Reviewed-by: Peter Varga <pvarga@inf.u-szeged.hu>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
| |
Task-number: QTBUG-60006
Change-Id: I75149082f36cd4d56da508283b766df680ed88b8
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
| |
The mentioned LICENSE file is not available; add it under
the name LICENSE.Chromium.
Task-number: QTBUG-60006
Change-Id: Ib2b3cb64467046c7ba1f53b0bbe14c3ec5a5291d
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
| |
Task-number: QTBUG-60006
Change-Id: Ibc0507f300f52154e6f131056d826a4dcef009c2
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
|
|
| |
Split intersecting quads using a BSP tree in the same manner as Chromium's
rendering algorithm. Note that these split quads still will not be correctly
rendered in software mode as Qt Quick's software renderer is not at all aware of
non-rectangular QSGClipNodes.
Task-number: QTBUG-62112
Change-Id: Ibfb72b9220817baebf828bc6183af7bd9c25d050
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
| |
Valgrind does not approve of the member cookie_ being read by the destructor but
never initialized.
Change-Id: I419fca616034fb04c4f3ee29edb068881583dac0
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
| |
Refactoring only, no algorithmic changes.
Task-number: QTBUG-62112
Change-Id: I871105ea48cd513e03b957d7daa9fe095c952a52
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Quote from Linux man-pages:
This function should never be used. Use free(3) instead. Starting with
version 2.26, it has been removed from glibc.
Arch Linux has switched to glibc 2.26 now, so cfree(ptr) is no longer available.
Since cfree(ptr) is equivalent to free(ptr) (even on SunOS?), this patch
replaces it with free(ptr).
Change-Id: I50859ecfe45c965a9477541b3b7644fcbcc330e2
Reviewed-by: Michal Klocek <michal.klocek@qt.io>
|
|
|
|
|
|
|
|
|
|
|
| |
Fix is based on afc9e2d9674f7ab5800df4803cc68c71d1ae691a
Moreover, new quick auto test has been added to check that the commit
happens in case mouse and touch input events.
Task-number: QTBUG-62942
Change-Id: Ie9d55e0bb5b3bbc34c099502e735b94f37c5d5f8
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A user can explicitly request software OpenGL rendering via
opengl32sw.dll instead of Skia rendering, by passing the
--enable-webgl-software-rendering command line argument.
This will make sure not to disable GPU processing if and only if
software OpenGL rendering is used on Windows (via
QT_OPENGL=software environment variable, or the
Qt::AA_UseSoftwareOpenGL attribute), so that WebGL is not disabled.
Software OpenGL rendering will still use Skia on platforms other than
Windows.
Change-Id: I8b3601b144e09e3a732abfb74dbbf6f924889b5a
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
This ensures the scripts and webchannel are added before
RunScriptsAtDocumentStart is called. It also means we shouldn't add
them later from the changed signal.
Task-number: QTBUG-62898
Change-Id: Ib4cd3a69134d0eafc7f2ac4b7927a1c79b63bfa6
Reviewed-by: Viktor Engelmann <viktor.engelmann@qt.io>
|
|
|
|
|
|
|
|
|
| |
Note that the Referer header still won't be sent if the download is triggered
via an anchor element with the 'download' attribute: crbug.com/455987 .
Task-number: QTBUG-61354
Change-Id: I5af971af916b2190756e3e58f19736072a213922
Reviewed-by: Michal Klocek <michal.klocek@qt.io>
|
|
|
|
|
|
|
|
|
|
|
| |
VK_MENU is the virtual key code for the Alt key, but
Qt::Key_Menu is the context menu key which maps to VK_APPS.
Now the menu key can be used to invoke the context menu on web pages.
Task-number: QTBUG-58306
Change-Id: I8674ab84848b5c8732e4d0cf189909191b85d562
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
| |
QT_ARCH is the current architecture, and if the host is Arm or MIPS, we
also need the flags there.
Task-number: QTBUG-61846
Change-Id: I38f826e061eff900f6e27f4de0fc913323cf5e3e
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
| |
Change-Id: Ibad2b8ebb90e79ff8337d994b2234eed5e230daa
Reviewed-by: Romain Pokrzywka <romain.pokrzywka@gmail.com>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
| |
Change-Id: Ifa305045f3594f06b0ee106cd43f9d35853f8958
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Reviewed-by: Michal Klocek <michal.klocek@qt.io>
|
|
|
|
|
|
|
|
|
| |
Set host_cpu to ensure GN gets on the right path matching the host of
the toolchain we give it in host_toolchain.
Task-number: QTBUG-62673
Change-Id: I99d6dfdbe4cee30bc240049b656d1ec73becce8a
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
Chromium itself does not really support processing touch events on
macOS, but rather uses native gestures sent by the OS. Warn the user
in case they send simulated QTouchEvents.
Task-number: QTBUG-62130
Change-Id: If47f25bc06f3d6f99ba70164a585182354ed9b5a
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
| |
The software backend version of QSGLayer defaults to not mirroring
vertically, where the OpenGL does, and we rely on that.
Task-number: QTBUG-62112
Change-Id: Id5368728fd7046d2e4dc35bd4b2b61f2db35dac6
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
|
|
|
|
|
|
|
|
|
| |
Chromium handles the flag for the render process, but with multi-process
we need to enable it for the browser process as well.
Task-number: QTBUG-62808
Change-Id: I1111dd7a2fd58876978ec5299bfb7abea8cb8487
Reviewed-by: Michal Klocek <michal.klocek@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
DropData::file_contents and DropData::file_description_filename are
filled by Chromium if the user drags an image from a web page. Write the
file's content into a file with the suggested file name in a temporary
directory, and set the drop action to "move".
This enables users to drag images from the browser to the desktop or a
file browser.
The file is moved by the client to the target location.
The temporary directory is removed when the page is closed.
Note that Chrome doesn't use this temporary directory trick, but employs
techniques like XDS [1] on Linux to transfer the data via the native
clipboard/DnD MIME object. Unfortunately QDrag doesn't support this.
Support could be added, but I consider this solution as good enough.
[1] https://www.freedesktop.org/wiki/Specifications/XDS/
Task-number: QTBUG-60790
Change-Id: I44f2b3170a124e861ed4131c421903d895b70715
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Instead of explicitly loading libraries from hard-coded locations,
we now just call dlopen(NULL, RTLD_LAZY). This returns a handle to
the host process'es context, which already contains the symbols of
both these libraries, because we link against them.
It was necessary to bypass LoadLibrary, because that expects a non-NULL
file path, so we couldn't pass NULL through that interface.
Task-number: QTBUG-57761
Change-Id: I29f037dfe542222b5188a33c7727c81a464a87bb
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Reviewed-by: Michal Klocek <michal.klocek@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
With QtWidgets the QHoverEvent handling is slightly incorrect
(HoverEnter and HoverLeave triggering Q_ASSERT), quite unnecessary
(Chromium works fine with just MouseMove events when mouse tracking is
enabled), and mostly unused (QHoverEvents are only delivered if the
WA_Hover widget attribute is set, which it usually is not).
QtQuick however does not have the equivalent of QtWidgets mouse
tracking, so to get mouse movement information into Chromium we have to
use HoverMove QEvents. But the HoverEnter and HoverLeave QEvents are not
used or useful for QtQuick either.
Task-number: QTBUG-62200
Change-Id: I333de2b6adcc24544935d36645036aedb07e51ac
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
|
|
|
|
|
|
|
|
|
| |
We currently don't support unbundling SRTP because Chromium uses a too
new unreleased version, but we were still testing for it and claiming to
use the system one if found.
Change-Id: I250f0d7fc1e09398b3196895bdd529d366cd0a86
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
Chromium has since version 50 overridden the global font settings with
the settings of every web contents created, so we need to first set the
global settings in the web contents, to still follow system defaults.
Task-number: QTBUG-62146
Change-Id: If10847c81beda08ecb7bfdf7556e39b1d2989754
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
| |
Task-number: QTBUG-61910
Change-Id: I894d1f180ba9952213b6f67a2445aa927dfd7e9b
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
|
|
|
|
|
| |
Change-Id: I4533e28a2c1a338abdb1d373a0a3bfe598a5edce
Reviewed-by: Peter Varga <pvarga@inf.u-szeged.hu>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The previous code tried to find out whether a user has set an
application proxy by checking the type of the applicationProxy.
This is wrong, because a system proxy will actually also change the
applicationProxy type.
Instead, we now rely on QNetworkProxyFactory::usesSystemConfiguration
to decide whether to use QtNetwork's application proxy, or Chromium's
logic for the system proxy. We also save the state of
QNetworkProxy::useSystemConfiguration to be able to track changes.
[ChangeLog][Networking] Fixed an issue where system proxy settings were
not picked up correctly.
Task-number: QTBUG-61910
Change-Id: I1d9af3f6006ba187266fe50c645f425a46632e41
Reviewed-by: Peter Varga <pvarga@inf.u-szeged.hu>
|
|
|
|
|
|
|
|
| |
Re-export missing C memory handling functions.
Task-number: QTBUG-61521
Change-Id: I8f94d93a70c81117735a9610e391a1e31c54f15a
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
This command-line argument makes it possibly to block exposure of
internal IP addresses when WebRTC access hasn't been granted, using
--force-webrtc-ip-handling-policy=default_public_interface_only
Task-number: QTBUG-57505
Change-Id: I0cddd1b20e0814811894204cb31e0c463a75f7be
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Unhandled key presses for printable keys were not received by the parent
widget anymore. This was a regression from 5.8.0.
The propagation is suppressed if NativeWebKeyboardEvent::skip_in_browser
is true.
Commit 4501b9d8 accidentally removed the wrong skip_in_browser
assignment.
Task-number: QTBUG-61621
Change-Id: I7eafa1fac5fb9a7edc4af2bc1aac21c106264713
Reviewed-by: Peter Varga <pvarga@inf.u-szeged.hu>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In Chromium 56 the --touch-events switch controls both the availability
of the Touch Events API in JavaScript (presence of 'ontouchstart' in
'window' object), and whether touch events are dispatched at all.
In Chromium 57, the switch controls only the Touch Events API
availability, and touch events are always dispatched.
In Qt 5.9.0 which is based on Chromium 56, we always dispatched touch
events, ignored the value of --touch-events, and determined the
availability of the Touch Events API by checking if QTouchDevice lists
any touch screen devices (essentially Chromium's --touch-events=auto
option).
This commit changes the behavior of WebEngine to match that of Chromium
57, so that users can choose to enable or disable the Touch Events API
via the --touch-events switch, whereas the default will be 'auto' mode.
Touch events will always be dispatched as usual. Users that wish to
stop dispatching touch events can install an event filter on
WebEngine's view focus proxy.
Task-number: QTBUG-61045
Change-Id: I07404af3336619869aa87a90a1b426036548dd55
Reviewed-by: Leena Miettinen <riitta-leena.miettinen@qt.io>
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
|
|
|
|
|
|
|
|
|
|
|
| |
Make sure that a non-Core Profile OpenGL Context is created on the
Chromium side, so that sharing between the Qt global context and the
Chromium ones works.
Change-Id: Ifc5f39268395615fe50c8aa9467f833c0de233a8
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
Reviewed-by: Viktor Engelmann <viktor.engelmann@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously when a default QSurfaceFormat was set with an OpenGL Core
profile, all the contexts created on the Qt side would obey the profile,
but Chromium would still use the Compatibility profile for its contexts
leading to warnings when trying to create shared contexts.
The fix is to check which OpenGL profile is used in the Qt global
shared context, and pass that information along to Chromium.
Note that this works only on macOS for now, and the default non-Core
profile is used on other platforms, even though Core was requested.
Passing CoreProfile to Chromium on Windows and Linux currently leads
to crashes.
Task-number: QTBUG-60605
Change-Id: I27a77091923624d19ccc2019953a5b07f9282916
Reviewed-by: Peter Varga <pvarga@inf.u-szeged.hu>
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
|
|
|
|
|
|
|
|
|
|
|
| |
Debug symbol level 1 is not enough to generate all debug symbols on
Windows. The symbols for the files found in src/core are generated with
symbol_level=1, but no symbols are generated for the files in
src/3rdparty.
Change-Id: I30729981d5956c1c6a8040008bc00af1f19c15d2
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
|
|
| |
We need to ensure that the path is using forward slashes to avoid any
chances of a trailing backslash escaping the final quote. Additionally
we need to quote the whole string in case there is a space, we can't use
shell_quote for this as this preserves the backslashes which can cause
a problem.
Change-Id: Iadde9a10e2d80ad48efd695d414af58d891c7412
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Zygote process is responsible for initializing ResourceBundle on Linux.
Thus it is expected to be initialized in Renderer process. When zygote
is disabled the Renderer process can't reload locale resources since the
ResourceBundle hasn't been initialized yet. However, there is no need
for reload because without zygote, Renderer Process initializes
ResourceBundle for itself.
Pulls in Chromium changes:
5352138 [Backport] Initialize RenderSandboxHostLinux in --no-zygote
mode to not crash.
Change-Id: Iace3585e23b4f18850783a813b1466a3ff5fe12f
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
| |
SelectionRegion::caret_rect is only documented to be supported on
macOS, so use anchor, focus instead. Also make sure that the rect
always has a width > 0; Both the Windows QPA and the native API
require this.
Task-number: QTBUG-60334
Change-Id: I8436fbfaa204d243578b33cbcaa3f28871b462ee
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Reviewed-by: Peter Varga <pvarga@inf.u-szeged.hu>
|
|\
| |
| |
| |
| |
| |
| | |
Conflicts:
mkspecs/features/functions.prf
Change-Id: I53d65ea49f546c7d4aadfdaff178fd2f3f4cdf11
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
With the no-accessibility feature moved to qtgui-config, we don't
always include it where web_content_apapter_client.h is included, which
gives it an inconsistent binary layout.
Solve it by making the optional method always defined as it doesn't
rely on anything from accessibility.
Task-number: QTBUG-61200
Change-Id: I65f34ab2b6763f3166b945e700994bd8d019a835
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Read the correct QT_ARCH and QT_TARGET_ARCH variables from QMAKE. This
fixes cases where toolchain arch does not match host arch (e.g. x86 tools
on x64).
Also removes unnecessary WEBENGINE_ARCH variable, and a mismatch in GN
mips names.
Task-number: QTBUG-61528
Change-Id: I11b050977ce6203c10d1776a15eb32efa4693290
Reviewed-by: Rolf Eike Beer <eb@emlix.com>
Reviewed-by: Michal Klocek <michal.klocek@qt.io>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
|
|/
|
|
|
|
| |
Task-number: QTBUG-61429
Change-Id: I936f1c924ae8158a5243fdf20d12e6c5508b684a
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Setting httpCacheType followed by persistentCookiesPolicy
may lead to io thread calling generateHttpCache just before
generateCookieStore, which triggers assert in line 344.
Recall generateHttpCache in case generateHttpCache was
called before generateCookieStore.
Task-number: QTBUG-61047
Change-Id: I1cfe5f18a884bd0d76fce4b30da0e22359cab2af
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
We were using the wrong spellcheck preferences, they have moved to the
spellcheck component. It mostly worked because their internal names had
stayed the same.
Change-Id: I485b11d88ee9c6e13d9c2b769c271d27bee4ba30
Reviewed-by: Michal Klocek <michal.klocek@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
If we don't give Chromium a host toolchain it will pick one of its own
overriding which compiler should be used, and breaking native builds
on non-x86 architectuers.
Task-number: QTBUG-61128
Change-Id: Ia25a4f43a28214003abcb11dc75a0e57064f76a7
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
The /WHOLEARCHIVE option was introduced in Visual Studio 2015 Update 2.
Since we support no older MSVC version we can unconditionally use this
option and remove our hack that simulates whole-archive by writing
object file names into a linker reponse file.
Change-Id: I32d2cb556c15f9a8aa2f77f608268e0af7ab6a1d
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
Clean up the logic for cross-building to x86 on x86-64 windows, avoiding
enabling debug symbols when force_debug_info is set, remove them 32-bit
builds even it _is_ set, and use the standard logic for getting the GN
target arch.
Change-Id: Ia6a5a5e7429ffb1cf07629deaeebfb2466280a9b
Reviewed-by: Peter Varga <pvarga@inf.u-szeged.hu>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously when a screen sharing request was accepted on macOS, instead
of showing the captured screen, a black rectangle was displayed. This
was due to passing an incorrect screen id to the screen capturer, which
resulted in a failed attempt to find the main active screen.
The fix is to query for all available screen ids from the operating
system, and choose the first id as the id to pass along to the screen
capturer.
Task-number: QTBUG-55165
Change-Id: Id8e648e59755aa2820b05b990adeaa9b58fd26f0
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
| |
Add dependency to spellchecker build_feature
Task-number: QTBUG-61165
Change-Id: Ib730280f00b07c9f8a0ffb59399e0a25bc9829a3
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|