Both ARM and x86 can convert fp16 much faster in bulk than one at a
time. This also enables hardware accelerated conversion on x86, when
F16C isn't unconditionally available at compile time.
This code is implemented in C to ensure that there's no leakage of
inline symbols from the .obj file that was compiled by Visual Studio
with AVX support. Unfortunately, simd.prf uses $(CXX) instead of $(CC)
for all its sources, which means the file gets interpreted as C++ by
g++, clang++ and icpc. Those compilers at least don't leak any symbols.
Done-with: Thiago Macieira <thiago.macieira@intel.com>
Change-Id: I9d26d99e83392861fb09564e0e8e8d76cd8483b3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Tested with Clang, GCC 4.5 & up, ICC 17 and MSVC 2017. No current
version of MSVC supports C11 and GCC implemented the features slightly
later in C than in C++.
Change-Id: I57a1bd6e0c194530b732fffd14f45c5074c9a052
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Replaces the Qt Accessibility Windows back end, formerly based on legacy
MSAA, with a new implementation based on UI Automation. Fixes issues with
accessibility tools like screen readers and magnifiers, and with the
automatic showing and hiding of the virtual keyboard in touchscreen-based
Windows computers.
[ChangeLog][Windows] The Windows Accessibility back end, formerly based on
Microsoft Active Accessibility, was replaced with a new implementation
based on Microsoft UI Automation.
Task-number: QTPM-487
Task-number: QTBUG-53024
Task-number: QTBUG-43190
Task-number: QTBUG-61926
Task-number: QTBUG-38499
Task-number: QTBUG-38337
Task-number: QTBUG-38501
Task-number: QTBUG-38502
Task-number: QTBUG-38504
Task-number: QTBUG-38505
Task-number: QTBUG-38507
Change-Id: I20b4f8f5e938fef791c6e9c577fcd919140999bd
Reviewed-by: Jan Arve Sæther <jan-arve.saether@qt.io>
A source of constant confusion is the QFontMetrics::width() function,
which does not return the bounding width of the text, but the advance
width. We deprecate this and add horizontalAdvance() instead, to avoid the
confusion in the future.
Note that there was an internal width() overload which was only there
for the purpose of supporting the Qt::TextBypassShaping flag. This
flag has already been replaced by public API, so no such
overload is added. Instead, we deprecate the TextBypassShaping flag
as well, which makes sense, since a replacement has been made.
Also note that there was a consistency problem with QFontMetrics and
QFontMetricsF, which are supposed to be interchangeable. The
QFontMetrics::width() functions for strings took an optional int length
argument, while the floating point version did not. This error is
corrected in the advance() functions.
[ChangeLog][QtGui][Text] Added QFontMetrics::horizontalAdvance() and
QFontMetricsF::horizontalAdvance() to replace the confusingly named
width() function. The latter has now been deprecated.
Change-Id: I0dfda43aa65c8235be32c62fade82cae05b29c79
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
This allows us to make sure that the PID we read is from the same boot
as we are right now running. The collision could happen on embedded
systems where the boot sequence is fixed, so all the same processes
would have the exact same PIDs after reboot as they did before.
[ChangeLog][QtCore][QLockFile] QLockFile can now properly conclude that
a lock file from a previous boot of the same device is stale and can be
removed. This is implemented only for Linux and Apple operating systems.
Task-number: QTBUG-63425
Change-Id: I0b48fc8e90304e0dacc3fffd14e8e3a197211788
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
This allow retrieving the value of the known PMTU for the current
socket. This works on Linux (IPv6 and IPv4) and FreeBSD (IPv6 only) --
the other OSes don't have the necessary API.
Note: do we need add IP_MTU_DISCOVER?
Change-Id: I6e9274c1e7444ad48c81fffd14dcaf97a18ce335
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
We document that -1 is not a valid value. The Windows structure's type
is unsigned, so the value is actually ULONG_MAX.
Change-Id: Ic632b4163d784b83951cfffd14f668645c4da3a9
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
[ChangeLog][QtNetwork][QNetworkInterface] Added dnsEligibility() to
QNetworkAddressEntry to indicate whether the address is eligible or not
for publication in DNS or similar mechanisms.
Change-Id: Id3ae5f853d964358ac1ab19b525334a426e0e052
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
[ChangeLog][QtNetwork][QNetworkInterface] Added preferredLifetime() and
validityLifetime() to QNetworkAddressEntry that report the remaining
lifetime of the address in the network interface.
Change-Id: I292b84e2193979446e43344b0727642812cba630
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
The milli-seconds since epoch value for an invalid transition is,
of course, invalidMSecs(), not invalidSeconds().
Added a comment while I was at it, explaining why we expect a
transition before the epoch, if such transitions are supported.
Change-Id: I0f376f9d69c0e6e79a309dc011943baa41175135
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
... that become apparent after switching qtestlib to use enhanced mouse
event (a37785ec76). With the old code path,
where QGuiApplication was deducing event type it would deduce mouse release
event even when there wasn't one. The new code path doesn't do that, which
revealed an obscure problem when mixing QTest::mouse* APIs (where QWindow
overload goes through QWindowSystemInterface API and QWidget overload goes
through QApplication::notify() and sets mouse_buttons from there). What
happened in this specific test case "./tst_qtreeview selection statusTip" was:
// tst_QTreeView::selection sets mouse_buttons = Qt::LeftButton from QApplication::notify
QTest::mousePress(widget, Qt::LeftButton, ..)
// tst_QTreeView::statusTip
QTest::mouseMove(window, )
The old code path sees that position and state has changed, creates a fake
mouse event, which gets deduced as mouse release even if there wasn't one.
And by luck this happened to set mouse_buttons=Qt::NoButton. So when we use
mouse_buttons later to create QMouseEvent everything works as expected. With
the enhanced mouse we don't clear the pressed button from mouse_buttons (set
in tst_QTreeView::selection) as this is done only from press/release events,
then pass it to QMouseEvent and later because of that QApplicationPrivate::
pickMouseReceiver() returns nullptr.
The fix here is to use e->buttons when constructing QMouseEvent, instead of
relying on mouse_buttons which gets changed from various places and has other
issues that can not be solved without invalidating the current documentation
of QGuiApplication::mouseButtons() (e.g QTBUG-33161). Tests and any Qt code
in general should avoid using the fragile QGuiApplication::mouseButtons() API.
This patch does not affect the old code path (it continues working as before)
and fixes the issue described above for the enhanced mouse API. The enhanced
mouse API actually is better in a way that it does not get affected by button
state from test functions that run earlier, as opposed to the old code path
where every subsequent test function uses mouse_buttons in whatever state it
was left by the test functions that run earlier.
Not relying on mouse_buttons when creating QMouseEvent helped also to discover
other logic error. This caused an in incorrect button state for a mouse move
event that is generated for a release event that simultaneously changes a mouse
position.
Task-number: QTBUG-64043
Change-Id: I6ad8e49d8437ab0858180c2d0d45694f3b3c2d60
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
We can use the D-Bus / systemd machine-id file (which is a UUID without
the dashes) on systems with D-Bus. On Windows, there's a value in the
registry that is filled when Windows is installed, like on Linux. For
BSD systems, the kernel has a UUID we can use too, so extract that.
Task-number: QTBUG-63425
Change-Id: I27eaacb532114dd188c4ffff13d32f2e3c1d74bb
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
ultrix and reliant have not seen a release since 1995. dgux not since
2001. bsdi not since 2003. irix not since 2006. osf not since 2010.
dynix... unclear, but no later than 2002. symbian needs no mention.
All considered obsolete, all gone.
sco and unixware are effectively obsolete. Remove them until someone
expresses a real need.
Change-Id: Ia3d9d370016adce9213ae5ad0ef965ef8de2a3ff
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Discussed during Qt Contributor Summit 2017. We concluded that we don't
want to make these functions public, as they do not follow Qt coding
style API. Specifically,
qStartsWith(a, b)
is not easily understood which argument is the needle and which argument
is the haystack (same problem memcpy() has). Compare that to
a.startsWith(b)
which can clearly be read in English as a subject-verb-object sentence.
This commit removes the unit tests that called compare().
Discussed-on: http://lists.qt-project.org/pipermail/development/2017-October/031060.html
Change-Id: Icaa86fc7b54d4b368c0efffd14ee6205eb9043fb
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Right now,this does really nothing. This commit is just to allow us to
transition the other modules (besides qtbase) to use the syntax that
will become the API.
I've marked three places to use the system CSPRNG:
1) the QHash seed
2) QUuid
3) QAuthenticator
I didn't think the HTTP multipart boundary needed to be
cryptographically safe, so I changed that one to the global generator.
Change-Id: Ib17dde1a1dbb49a7bba8fffd14ecf1938bd8ff61
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
It fails for me on Windows 10. Therefore, this is wrong.
FAIL! : tst_QHostInfo::initTestCase() 'networkSession->waitForOpened(30000)' returned FALSE. ()
S:\qt\qt5-msvc2017-x64\qtbase\tests\auto\network\kernel\qhostinfo\tst_qhostinfo.cpp(194) : failure location
Change-Id: Ib17dde1a1dbb49a7bba8fffd14ed5691472a4760
Reviewed-by: Jesus Fernandez <Jesus.Fernandez@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
tst_noqteventloop.cpp 'event' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtreeview.cpp 'canFetchMore' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtreeview.cpp 'fetchMore' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtreeview.cpp 'hasChildren' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtreeview.cpp 'rowCount' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtreeview.cpp 'columnCount' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtreeview.cpp 'index' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtreeview.cpp 'parent' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtreeview.cpp 'data' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtextedit.cpp 'begin' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtextedit.cpp 'end' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtextedit.cpp 'updateState' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtextedit.cpp 'drawPixmap' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
tst_qtextedit.cpp 'type' overrides a member function but is not marked 'override' [-Winconsistent-missing-override]
Change-Id: I2a0c5da15994619383c1f90fee311927e58d7af0
Reviewed-by: Jesus Fernandez <Jesus.Fernandez@qt.io>
1. Recently we have updated our receive window size to a larger value.
Unfortunately, this also results in auto-test pumping through
more data, which probably takes more time on CI.
At the moment we do not have any public API on QNAM's level to
customize HTTP/2 parameters (aka 5.10/FF and so on). So we use the fact
that QNAM is QObject and we can set a property on it. This property
is our Http2::ProtocolParameters object that allows us to configure:
- HPACK parameters (in 5.10 - noop)
- session receive window size
- different SETTINGS as described by RFC 7540, 6.5.2.
2. Undocumented environment variable to set ENABLE_PUSH is not needed
anymore.
3. In 5.11 Http2::ProtocolParameter will become a public API
and we'll introduce a new setter in QNAM.
Change-Id: If08fd5e09e7c0b61cf9700b426b60b5837b6b2e6
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
The previous version was good, just not optimal. Because the input was
an unsigned 64-bit number, compilers needed to generate extra code to
deal with HW instructions that only convert 64-bit signed input. And
that was useless because a double uniformly distributed from 0 to 1 can
only have 53 bits of randomness.
The previous implementation did exactly what the Microsoft libstdc++ and
libc++ implementations do. In my opinion, those implementations have an
imperfect distribution, which is corrected in this commit. In those, all
random input bigger than 0x20000000000000 has a different frequency
compared to input below that mark. For example, both 0x20000000000000
and 0x20000000000001 produce the same result (4.8828125e-4).
What's more, for the libc++ and MSVC implementations, input between
0xfffffffffffff001 and 0xffffffffffffffff results in 1.0 (probability 1
in 2⁵³), even though the Standard is very clear that the result should
be strictly less than 1. GCC 7's libstdc++ doesn't have this issue,
whereas the versions before would enter an infinite loop.
Change-Id: Ib17dde1a1dbb49a7bba8fffd14eced3c375dd2ec
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
The path wouldn't match if the cookie's path was root ('/') and the
URLs path was empty.
Change-Id: I6dcd10f1fdf4f48f14e50f1b169cbdfda7005849
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
We need to ensure that simultaneous touch and tablet input is possible.
Change-Id: I0c0d06e382b6b89e11ed06c1e4ad0db63b3b7c82
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@qt.io>
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
Before this patch a very accurate drop position below
or above an index was needed. Therefore it was not that
easy to do.
This patch increases the above/below area to be about
18% of the item (still leaving the most space for the item).
An average user will likely be 2-3x faster with dropping
below or above (while not losing much when dropping on items).
[ChangeLog][QtWidgets][ItemViews] Made it easier to
drop above and below items.
Change-Id: I47f0f80c76878c17ebf3f93d0a0cc82755971c2a
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
All systems must implement it by now. If there's any system still
without it, that means it has no IPv6 support, so they can disable
QtNetwork entirely.
[ChangeLog][Deprecation Notice] Starting with Qt 5.10, IPv6 support is
mandatory for all platforms. Systems without proper IPv6 support, such
as the getaddrinfo() function or the proper socket address structures,
will not be able to build QtNetwork anymore.
Change-Id: I3868166e5efc45538544fffd14d8c28046f9191b
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Proximit events are sent to the QGuiApplication, catch them via
event filter.
Change-Id: I7f896e7d9f5c90347b9e3c708feb69abd1c5fc95
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
[ChangeLog][QtNetwork][QHostAddress] Added isGlobal(), isLinkLocal(),
isSiteLocal(), isUniqueLocalUnicast(), and isBroadcast() classification
functions to complement isLoopback() and isMulticast().
Change-Id: I3868166e5efc45538544fffd14d8fca6e9042c04
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
All POST requests that were redirected would previously turn into GET
requests. This does not follow the standard for HTTP codes 307 and 308.
Task-number: QTBUG-63142
Change-Id: Ibd25a9566066e589670a9bc34e5dc5111f8139d5
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
In some cases when a session isn't needed (i.e. for localhost), the
session is not opened at all. If a program (e.g. our tests) redirects
from localhost to a different system (e.g. the qt network test
servers, or the internet) it will wait for a session forever. So, we
need to check if a session is needed for the redirect-target and then
open one. It is usually opened in
QNetworkReplyHttpImplPrivate::_q_startOperation
Change-Id: Id3b78182a3fb3f63f0235ecb1fb665df8bd0c4ca
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
From: https://tronche.com/gui/x/xlib/events/exposure/expose.html
"The circumstances in which the X server generates Expose events
are not as definite as those for other events."
On windows with XCB_GRAVITY_NORTH_WEST flag set we should not get
expose events according to e2665600c0,
but as stated earlier this might not always be true.
Nevertheless, sometimes we get expose event from X server when shrinking
window, but most of the time we don't. Make the test not flakey by
checking that we get at least 1 expose event, instead of exactly 1.
Now running test 500 times in a loop does not fail.
Task-number: QTBUG-63424
Change-Id: I8004e622020cc09e11b7d592faf6d9ee1b9cfee2
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
(cherry picked from commit 542e11ab2b)
Reviewed-by: Tony Sarajärvi <tony.sarajarvi@qt.io>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Liang Qi <liang.qi@qt.io>
Tests should not use QCursor to emulate mouse move, see QCursor::setPos() docs.
The flakiness of the test on XCB is not surprising when the test queries geometry
even before the window has been shown. With the re-factored version I could not
reproduce flakiness anymore.
Removed Q_OS_MAC and closed QTBUG-26274 as test passes on macOS from which I
assume that the underlying issue has been fixed.
Removed Q_OS_QNX ifdef as test does not rely on QCursor anymore.
This patch also fixes the issues on minimal / offscreen platform plugins.
QCursor::setPos() is evil for auto test purposes.
Note:
We intentionally use QTest::mouseMove(QWindow *window, ..), not the QWidget overload.
The QWindow version gets routed through QWSI, which ensures that all necessary events
are generated as expect. In QWidget code path this is currently disabled by
QTEST_QPA_MOUSE_HANDLING.
Change-Id: I285c26cff09e3f2750f8c2abbb1f46c8f7be984a
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
(cherry picked from commit 1af927976a)
Reviewed-by: Tony Sarajärvi <tony.sarajarvi@qt.io>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Liang Qi <liang.qi@qt.io>
We already cleared 'cookedHeaders', which is a QHash for 'known headers'
(enumerators as keys instead of strings), now do the same for 'rawHeaders'-
not to end up with some weird mix of headers from all possible redirect
responses and the final response.
Task-number: QTBUG-61300
Change-Id: Ifd6655c4167840bb00d29446d36ce65ba2d5491a
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
A few tests would QSKIP depending on the inclusion of SSL, producing
multiple lines of noise in the output.
And one test used https in one of its configurations without checking to
see if it could, causing an UnknownProtocolError.
Change-Id: I5f54bf1005f962cc027c099b816fbe245dc43d3f
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
QStandardItemModel::setItemData replaces the content of an item data
with the new values rather than updating/inserting which is
the behavior for QAbstractItemModel. This patch aims to unify the
behavior.
[ChangeLog][QtWidgets][QStandardItemModel] Fixed setItemData()
incorrectly deleting unmodified data. That behavior is not
following QAbstractItemModel's documented behavior which is
no modification of data not provided in parameter.
Task-number: QTBUG-45114
Task-number: QTBUG-10872
Change-Id: I2be40cee372b68d9f71c976548ecda6dc3011241
Reviewed-by: David Faure <david.faure@kdab.com>
On macOS, absence of a QPA menu means that we should be using our
own internal logic since there's no entity on the QCocoaMenuDelegate
to take care of the shortcuts.
Change-Id: I35ed8f0b55445f61d0528709d4debb636a502002
Task-number: QTBUG-61039
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
We were using the default ones, provided by RFC7540. It appears they are way
too restrictive and conservative: when downloading something relatively big,
a stream keeps spending the whole session/its own 'recv' windows and thus
we have to constantly send WINDOW_UPDATE frames. This significantly slows
down our HTTP/2 implementation, making it orders of magnitude slower than
HTTP/1.1. To fix this:
- We send SETTINGS_INITIAL_WINDOW_SIZE in the first SETTINGS frame
to inform our peer that per-stream WINDOW is bigger than 64Kb
- We increase the session's receive window size.
Task-number: QTBUG-63722
Change-Id: I31312fcfd5f0fc0aee6aaa5d3562cc7d1b931adc
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Error out if it's missing or broken (Mersenne Twister not present).
This ensures that we never have a low-quality random generator in Qt.
Change-Id: I0a103569c81b4711a649fffd14ec80649df7087e
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Auto-reset events are automatically reset to non-signaled when we get
notified about the signaled state. This implies that we cannot check the
event state again in activateEventNotifiers. Instead, store the signaled
state in a QAtomicInt and test and decrement that.
This amends commit 85403d0a.
Task-number: QTBUG-63555
Change-Id: I0adee3d2929783f98d91060f9106c8b5266d72fa
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Allan Sandfeld Jensen
Thiago Macieira
Andre de la Rocha
Eskil Abrahamsen Blomfeldt
Edward Welbourne
Liang Qi
Gatis Paeglis
Jake Petroules
Friedemann Kleint
Timur Pocheptsov
Sami Nurmenniemi
Mårten Nordheim
Orgad Shaneh
Shawn Rutledge
Thorbjørn Lund Martsum
Samuel Gaist
Gabriel de Dietrich
Joerg Bornemann