If the QLoggingRegistry gets called as part of the static initialization
phase, it would call into Android's QStandarPaths implementation, which
assumed that the HOME env. variable was already set. Since the variable
isn't set before main is called, QDir::homePath() returns the root path,
which would be cached and always returned.
With this fix we now call Android's getFilesDir() directly, which will
always return the right path. Since the font locations are also relying
on an environment variable being set, we no longer cache that either.
Task-number: QTBUG-65820
Change-Id: If45f3d5f0e87b808a62118ae95c31b492885646a
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
On Android, we load the application library, and its dependencies (Qt),
on Android's main thread (thread 0), and then spin up a secondary
thread (thread 1), that we call main() on.
If any QObject is constructed during loading of the application library
or any of Qt's libraries, via static initializers or constructor
functions, we will set QCoreApplicationPrivate::theMainThread to
thread 0, which will confuse Qt later on when it's being run on
thread 1, and will result in a warning during QCoreApplication
construction:
QApplication was not created in the main() thread
This situation can easily lead to a crash as well.
Unfortunately logging via qDebug/qCDebug and friends will trigger
this too, as they internally use QObject.
Fixing the root cause of this is under investigation, but for now
we will partially revert fa2a653b3b for Android. The effect
is that any qCDebug with a "qt.*" category before qApp construction
will turn into a no-op, like it was before fa2a653b3b.
This patch does not cover the case of a regular qDebug, or a qCDebug
with a non-Qt category. Those will still produce the same symptom,
as before fa2a653b3b.
Task-number: QTBUG-65863
Change-Id: I95675731d233244530d0a2a1c82a9578d5599775
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Instead of trying to detect situations where we need to send a real
expose event instead of an update request in response to a drawRect
call, we keep track of when we've asked the view to display due to
a requestUpdate call, and only deliver the corresponding drawRect
as an update request if no other code has asked the view to
display.
This should cover all cases of asking the view to display, issued
from our code or from AppKit, such as the view changing its backing
scale factor, or otherwise needing a real expose event.
Task-number: QTBUG-65663
Change-Id: I1783787823aee889dad8e48f34a1cb0f1b7b06bd
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
The backwards iteration was done under the assumption that the only
valid modification of the winEventNotifierList in a slot connected to
activated() would be the removal of the notifier itself. This is wrong.
Instead, iterate forwards, like before 85403d0a, and check the index
against the current list size in every iteration. This ensures that we
do not run out of bounds while the list is modified.
Also, retry the activation loop if the list was modified by a slot
connected to activated(). This ensures that all notifiers with signaled
handles are activated.
Task-number: QTBUG-65940
Change-Id: I25f305463b9234f391abc51fe0628d02f49b6931
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Change-Id: Ib9a1b107dd8d2cd4bf541c3edf19a602fde6356f
Reviewed-by: Cristian Maureira-Fredes <cristian.maureira-fredes@qt.io>
Reviewed-by: Martin Smith <martin.smith@qt.io>
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
This way, it's lossless.
Change-Id: I5e421e32396d44e4b39efffd150b744e40fff3a1
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
[ChangeLog][Third-Party Code] libjpeg-turbo was updated to version 1.5.3
Change-Id: I8c0caec3c92ba199b1a3baac9a523cc399c8001f
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
Without this, building Qt and Qt applications fails with GCC 8.
The errors look like this:
writing to an object of type ‘class QPointer<QQuickPointerHandler>’ with no trivial copy-assignment; use copy-assignment or copy-initialization instead [-Werror=class-memaccess]
Task-number: QTBUG-65691
Change-Id: Ie5a30814125deca7a160b9a61f5aa3f944ee1ac9
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The PRODUCT_BUNDLE_IDENTIFIER property was not defined in the Xcode
project file and therefore the build would fail.
This fixes a regression introduced by 0749ba2c5e.
Task-number: QTBUG-65673
Change-Id: I8089b36d86588223ec34859af7388c99a3574d8b
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Fix regression from commit 8b3a120a where also dialogs
got NSResizableWindowMask, making them have a Zoom
button.
Task-number: QTBUG-65668
Change-Id: I21054b3aa6fc11eab3d93f78ede44ae771522e2c
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
The idx member of EventWithNotifier is unused.
Change-Id: I0f5aacaaad4b4e82c57ff7bb020586944014f139
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
When the application is hidden then Qt will hide the popup
windows associated with it when it has lost the activation
for the application. However, when it gets to this point it
believes the popup windows to be hidden already due to the
fact that the application itself is hidden. As a result,
when the application is restored it causes a problem with
the still visible popup window as it is taking the input
events without responding to them.
Therefore we need to explicitly hide the windows right before the
application is hidden to ensure that they are actually hidden
correctly.
Task-number: QTBUG-58727
Change-Id: I4be1e1c0b1388d0c9ec872e7732185670998b7af
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
The linker complains:
LINK : warning LNK4281: undesirable base address 0x67000000 for x64 image; set base address above 4GB for best ASLR optimization
And it's not really required anymore, as the recommended /DYNAMICBASE is
the default.
Change-Id: I56b444f9d6274221a3b7fffd150caab1beecfd43
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
When original QNetworkRequest is missing "Host" header (which is the most
common case), it is provided automatically from request URL. However,
resulting header is appended to the list, i.e. after all headers specified
by user, which may include a big bunch of cookies.
To the contrary, RFC 7230 suggests: "However, it is good practice to send
header fields that contain control data first, such as Host on requests
and Date on responses, so that implementations can decide when not to
handle a message as early as possible". Many other user agents are
following this suggestion.
Task-number: QTBUG-51557
Change-Id: I1448ed3ae124f5ce86a8ca8ff35f5d05476a005d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Build systems such as Meson heavily rely on pkg-config to get build
flags.
-DQT_{module}_LIB isn't exported in .pc files which makes Meson unable
to build some codes out of the box.
Change-Id: I806998f19f815e05a47b60129977e52587b38d47
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Since braille characters are not included in Apple Unicode MS, we
extend the fallback to include Apple Symbols to cover those too.
Task-number: QTBUG-63510
Change-Id: I3977f1f9b7370a9fe9cfde643c86518e006c050a
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
Follow the example of the other modules and explicitly mention
the valid licenses on each module landing page, optionally
combining it with trademark information.
Change-Id: I9f1fea0f002e0ab607da89a0cbfe7c53060582d7
Reviewed-by: Leena Miettinen <riitta-leena.miettinen@qt.io>
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
Make it clearer what the variable actually does by mentioning
that it makes its headers available for inclusion and causes
it to be linked to the binary.
Change-Id: I72821d4bceea7a92e91175ba6c5acc4c3377d7b7
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Setting speed to 0 means the current frame will continue to be shown,
the finished signal is not emitted, and state remains QMovie::Running.
Task-number: QTBUG-65758
Change-Id: I681d902e3211c5899b21043e5177b7c73d5d3fb5
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
There's no external arm assembler to fall back on here. The
actual assembly is already disabled since 8072c36eeb,
but passing the empty assembly file to any random external
assembler could end up producing an empty object file for the
host environment instead of the target, in a cross build.
This wasn't an issue as long as the clang compiler only was identified
as g++ within mkspecs, making no_clang_integrated_as a no-op. If
the mkspec actually identifies it as clang, this config can't
be added here.
Change-Id: I0f20b9b2a8d13b5e7e1b654e391d88b639c031bf
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Passing -1 to waitForReadyRead() may cause it to wait for some time
but the data retrieved may be enough for processing. So if 0 is passed
from read, indicating that there is potentially more to come, then
it will do a waitForReadyRead() then for more data to come.
Change-Id: I75f270d1f124ecc12b18512cc20fb11f7a88f02e
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
sljit, the JIT in pcre2, doesn't quite fully work on windows on
arm yet (although the subset of functionality used by pcre2 might
mostly work). On arm64, it fails to compile (although working around
that failure is pretty easy), when using clang in MinGW mode
(where __GNUC__ is defined, making the JIT auto-enabled).
Disable it in these configurations until it has been fully tested
and fixed upstream first.
Change-Id: Ie9681cb7cdd1960586ba194c71e057eb918cb419
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
Correctly scale the requested geometry in the constructor
of QWindowSystemInterfacePrivate::GeometryChangeEvent().
Amends 3a31c70879.
Task-number: QTBUG-57608
Task-number: QTBUG-65580
Change-Id: I0ef60deda69bb61ab57972e02c342b7773e1da6b
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
This affected PE_IndicatorCheckBox in some cases.
Change-Id: I6f10dabd2ca4093f4c1bdaa2bd0ebf73c02e8d12
Task-number: QTBUG-65737
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Yulong Bai <yulong.bai@qt.io>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
mingw-w64 toolchain:
- remove 'QMAKE_CFLAGS_AESNI' and 'QMAKE_CFLAGS_SHANI' definitions,
which are already set in 'gcc-base.conf' toolchain; this is
a continuation of 77347a36 relating variables, introduced in 5.10,
- remove 'gcc-base.conf' duplicate inclusion, left after eef2d1af
was merged into 5.10.
Change-Id: I328def916ccfb93c3f6b8336eb8801defbac37ae
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Both ICC and MSVC have bugs concatenating strings and that's despite the
standard giving an example ([lex.string] table 9) that says that
u"a" "b"
is the same as
u"ab"
We can't change to preprocessor concatenation (u ## STR) because that
fails when QStringLiteral is used with a macro instead of an actual
string literal.
Task-number: QTBUG-65479
Change-Id: I39332e0a867442d58082fffd1504a0a868c291ef
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The Embedded Android build (Boot to Qt Android injection) is defined by
having both Q_OS_ANDROID and Q_OS_ANDROID_EMBEDDED flags defined,
as well as having Qt config android-embedded.
This commit enables the possibility to build embedded Android builds.
(i.e. Qt build for Android baselayer only, without JNI)
Change-Id: I8406e959fdf1c8d9efebbbe53f1a391fa25f336a
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
A large plugin dll (e.g., one with many MB of debug info) could cause
a 32-bit application to crash when the entire dll file could not fit
within the address space of the process. The code for validating a
plugin library and retrieving metadata from it first tried to map
the entire file in memory, which failed for large files, returning
NULL. In this case, the code tried then to read the entire file via
QFile::readAll(), which deep below caused a bad_alloc exception in
malloc, resulting in the termination of the application. This change
handles the case where the library could not be mapped into memory,
in spite of memory mapping being supported, by reporting the error
and returning false, making the plugin unavailable.
[ChangeLog][QtCore][QPluginLoader] Fixed a bug that would cause the
Qt plugin scanning system to allocate too much memory and possibly
crash the process.
Task-number: QTBUG-65197
Change-Id: I8c7235d86175c9fcd2b87fcb1151570da9b9ebe3
Reviewed-by: Jan Arve Sæther <jan-arve.saether@qt.io>
We do add, remove or update third party code in minor releases,
sometimes even in patch level releases. However, the documentation is
supposed to be valid for all existing Qt 5 versions, but doing this for
attributions would require infrastructure we don't have.
Therefore rather make it explicit which Qt version the attributions
apply to. Also mention since when Qt is available under LGPLv3
(starting with Qt 5.4).
Task-number: QTBUG-65665
Change-Id: I328b5bf0c143f78ea61aad51f0644c3cbb6dee49
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
QFileInfo.isRelative() deems any path starting with a slash to be
absolute; on MS-Win, such paths need a drive specifier (unless they're
UNC), so use IoUtils's more robust test for absolute paths.
Change-Id: I7d0872a87833cbf1cc1a6ef107941adc4c529624
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Gcc 8 now warns against casting functions taking references to
functions taking pointers:
warning: cast between incompatible function types from ‘QVariant (*)(const int&, const int&, double)’ to
‘QVariant (*)(const void*, const void*, double)’ [-Werror=cast-function-type]
Suppressed by casting over void(*)().
Change-Id: I48bee53253b8e87110e8a49efa761fc2add89c5a
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The __fastcall calling convention is silently ignored on other
architectures. The GNU attribute regparm is allowed but doesn't
make sense on x86_64. On other architectures, the attribute isn't
supported at all.
This fixes building with clang for MinGW/ARM and ARM64.
Change-Id: Ice1c6eadd0e90b2e5e34736542ee49a25dc67fe6
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Don't use the full size of dest in strncpy, we override the last byte
anyway afterwards.
Change-Id: I9f1618cc9019e0060f01deed58a726caa1989d3f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The native APIs don't support previous transition, only next after a
stipulated date. The prior code started its search at the epoch; if
used for a time before the first transition after the epoch, this
found no transitions so returned invalid data, when the last
transition before the epoch would have been suitable. It also wound
through all transitions since the epoch, on its way to the selected
time, which was potentially laborious.
Instead, start a year before the stipulated time; this should get a
transition if the zone uses DST. If it doesn't, start with the first
known transition and binary-chop our way to one within a year of the
last before the stipulated time; then wind forward one transition at a
time, as before. The chopping is actually faster than binary: each
time we find a transition after the interval mid-point but early
enough, we move the early end of our interval to the transition, which
is later than the old interval's middle. Using halving, starting with
a vast interval, should thus only incur modest cost, while ensuring we
give up early when no transition data is available at all or the
zone's first transition ever was after the stipulated time.
Task-number: QTBUG-65443
Change-Id: I96c14540fc2600837e6a22e480fb8dc36cb37220
(cherry picked from commit 7c0b706488)
(cherry picked from commit a090076e93)
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
Reviewed-by: Jason Dolan <jason.t.dolan@gmail.com>
use a "use" entry instead of including the transitive dep into the list
of libraries.
this also removes the redundant check of freetype features from the
fontconfig test code.
Change-Id: I86b78028255c9bf0a62be5ec0f97a62ef3fda36f
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
such a project structure violates the assumption that the referenced
resources are specified relative to the sub-project root, so we must
resolve them to absolute paths manually.
Task-number: QTBUG-65753
Change-Id: I8bcd5c6e7d7c6a713e5fcd3668be7d2f23169501
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Christian Strømme
Tor Arne Vestbø
Liang Qi
Joerg Bornemann
Alexey Safronov
Thiago Macieira
André Klitzing
Ville Voutilainen
Jake Petroules
Morten Johan Sørvig
Joni Poikelin
Andy Shaw
Konstantin Tokarev
Alexis Jeandet
Friedemann Kleint
Kai Koehne
Mitch Curtis
Val Doroshchuk
Martin Storsjö
Laszlo Agocs
Gabriel de Dietrich
Alexander Shevchenko
Jeremy LUGAGNE
Oliver Wolff
Otto Ryynänen
Alex Trotsenko
Andre de la Rocha
Tero Alamäki
Edward Welbourne
Allan Sandfeld Jensen
Oswald Buddenhagen