The behavior from before libdouble-conversion is that in case of an
overflow the OK flag is set to false, but the returned number is still
infinity, rather than 0. Also, the number of processed characters is
always set to the number of characters actually processed, unless
garbage is found.
There is an important distinction between an overflow and garbage. The
client code may accept overflows and infinity may be a valid result.
Garbage is most certainly not acceptable. Having an infinity/false result
in addition to 0/false allows the client code to distinguish those.
One application where this is useful is parsing JavaScript.
Change-Id: I4b8581568144b44fca3353c4bd9685c702762af9
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This allows to drop the default ctor and isNull() method of
SectionToken again, and reduces text size on optimized
AMD64 / GCC builds by a further 456 bytes.
Change-Id: I7109c9a138dcc27c478c0827d6216a043f592992
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
Since Qt::DayOfWeek has such a limited range, it doesn't really make sense
to use it as a key in a QMap, which is optimized for much larger key spaces.
Replace by a C array of the value type, and, since the value type in question
(QTextCharFormat) doesn't have an invalid state, use a bool array to track
whether any of the seven possible keys has been inserted.
Wrap that in a small helper class which provides only the subset of QMap API
that QCalendarModel needs.
Saves 1352 bytes of text size on optimized GCC AMD64 builds.
Change-Id: Ie51390825d5933739659c4ea8e8da0af68577a9d
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Qt now requires it.
We could try and run the common/c++11 test for MSVC too, but we know it
passes, so we won't bother and thus slow down the configure step.
Change-Id: Ib056b47dde3341ef9a52ffff13ef6b53b9f8bb65
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
This function is needed across several OS', so refactor it out to
a common place.
Change-Id: I35b957029c965672739d03cd2db3e87f5bd0acdf
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@theqtcompany.com>
There really is no point in allocating a struct containing a
pointer and an int on the heap. Allocate by-value instead.
Since QList would be inefficient then, switch to QVector.
A pointer to the current token was replaced by its index.
That saves looking it up on every toPreviousToken()/
toNextToken().
This saves 816 bytes of text size on optimized AMD64 / GCC
builds, even though QVector expands to more code than QList.
Change-Id: I030ee3f6acabe76168a518495bd4462711519e54
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
On some platforms our 1200 x 1000 bytes seems to be not enough to
fill kernel buffers (socket write). But it's not a reason to fail test,
just skip it.
Task-number:QTBUG-49205
Change-Id: I13ea6f315f9318288ba054cf8bfa6cdd61e489d2
Reviewed-by: Richard J. Moore <rich@kde.org>
This also fixes the underlying cause of QTBUG-44039 and QTBUG-43885.
You can choose between system, qt, and no libdouble-conversion
support. If you choose "no", snprintf_l and sscanf_l will be
used.
By default, system double conversion is used if the system provides a
double-conversion library. Otherwise the bundled libdouble-conversion
is built. sscanf_l and snprintf_l are not used by default as the
planned "shortest" conversion mode to produce the shortest possible
string will give less precise results when implemented with snprintf_l.
Change-Id: I8ca08a0fca5c54cf7009e48e771385614f6aa031
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
This separation makes it possible to make a
canadian cross build of Qt on a linux build machine.
The canadian cross build requires an external Qt that
runs on the build system.
Change-Id: Ifd83a4c6376d3299647e74bb349a3452a6f433fc
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
The previous implementations did not check the full mantissa. The result
was that certain NaN values were seen as +/-Infinity.
A nice benefit is that the generated code for this implementation is also
faster.
Task-number: QTBUG-47692
Change-Id: I1507ec579ccd9a2ab97da8cf83dabbc5d6e28597
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
We want to be sure that the conversions are independent of the current
locale set by setlocale(3) and we want to test both directions of the
conversion.
Change-Id: I01be19e680588478d07fd65a48796e806e681863
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Some of the examples make no sense without bearer management and
QNetworkSession is not defined if QT_NO_BEARERMANAGEMENT, so
tst_qnetworkreply.cpp has to be adjusted.
Change-Id: Ic2f73746cba74f670ae5b5e99b0be1461ff6d182
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
The function is not limited to creating top level windows.
Change-Id: Iedeeba0093f251c852e9b7c9ba23c567a79a53f8
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
Change-Id: I50a3e5584a1bc2553acd4bbec1b9a290b8475e47
Reviewed-by: Markus Goetz (Woboq GmbH) <markus@woboq.com>
Reviewed-by: Richard J. Moore <rich@kde.org>
We need C++11 support. The ChangeLog is in the previous commit.
Change-Id: Ib056b47dde3341ef9a52ffff13ef6bfa55462a22
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
The new version broke the build again
-> fix it again.
Change-Id: I75761d134d97a2784f1de5076412aa814fdf9bcd
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
Reviewed-by: Gunnar Roth <gunnar.roth@gmx.net>
Reviewed-by: Mark Brand <mabrand@mabrand.nl>
These don't report ABS_X/ABS_Y, but are still a touchscreen. This is commonly
seen on Android hardware.
As a bonus, we clean up the flow control through device discovery: by using
memset, we can ignore the error value of ioctl() and just check the bits we
need.
We can also close() the FD instantly after running the ioctl calls.
Change-Id: I6405597dc34841c805552fa73ef86250eaae111c
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
Tests that window creation order doesn't affect the resulting geometry
of the parent and child windows.
Change-Id: Iff0cb5adf87107dfed4a633a67e1b4312b90e24a
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
When a child QWindow is shown by calling setVisible(true), we don't need to
create the platform window immediately if the parent window hasn't been
created yet.
We defer creation until the parent is created, or we're re-parented into
a created parent or made top level.
This optimization is more important now that we create the full parent
hierarchy once we decide that we need to create a child QWindow.
Change-Id: Ia4f0430f0d3709a12f41f6473c1cea6b0ef3c9cd
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
If the QWindow has been created and has a platform window, we need to
ensure that the windows in the parent hierarchy also are created before
we set a new parent on the platform window, otherwise the platform window
will think that it's a top level window.
Change-Id: Icb2c3cd5411cd1eb595e8673e144095feda31df9
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
To be able to create a platform window for a given QWindow we need to
sync up the parent hierarchy first, so that the newly created window
can be placed into that hierarchy.
Without creating the parent hierarchy first, the QPlatformWindow will
end up thinking it's a top level window, when in reality is represents
the platform backing of a child QWindow.
Change-Id: I2cad7759fbc118b04718e7a27ec7570ce1238757
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
Saves one iteration of the container and unneeded allocation of a list.
Change-Id: I1626dc3460e21741de3261dd68be9c1fa8bdc792
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
dragImage:at:offset:event:pasteboard:source:slideBack: on NSView is
deprecated since 10.7. Use the one on NSWindow instead.
Change-Id: Ia1c2ea367ae2ca5194b52ea57ab261461bf8b529
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@theqtcompany.com>
Add a debug proxy style that wraps around an existing style and outputs
the parameters and return values of its function calls, for debugging
QStyle.
Task-number: QTBUG-47084
Task-number: QTBUG-45055
Change-Id: I638618e83bbac8ca4fdce31441bf4c913e911154
Reviewed-by: Andy Shaw <andy.shaw@theqtcompany.com>
For Windows CE 7, a QCoreApplication was instantiated
in each test.
Change-Id: Ib579e416e47f884ef0cdf829139518fcc3910b9b
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@theqtcompany.com>
- Add new virtual QPlatformDrag::cancelDrag()
[avoiding a conflict with existing QBasicDrag::cancel()]
- Implement on Windows by returning DRAGDROP_S_CANCEL
from IOleDropSource::QueryContinueDrag() as suggested on report.
- Implement in QBasicDrag by calling QBasicDrag::cancel()
and quitting the event loop.
- Add new API static void QDrag::cancel() for it.
Task-number: QTBUG-47004
Change-Id: I4b4bb52e5fc226c8e04688ac1b0f9550daaf918e
Reviewed-by: Jørgen Lind <jorgen.lind@theqtcompany.com>
Reviewed-by: David Faure <david.faure@kdab.com>
Instead of converting the QLatin1String to a QString at the first
opportunity, keep it around until it is appended to one of the
internal QStrings in write().
Avoids a memory allocation per QLatin1String / const char* streamed.
Change-Id: Id973a9b743e5a6696defbc4ef4ed2db1ef54e9cc
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Commits 0b700fb6f9 and
0e4d94edd0 defined QTypeInfo for those
three only for Qt 6 (and forced C++11 support). Define them in Qt 5,
keeping compatibility with QList storage. That is, these types are
defined to be primitive and static at the same time, for the benefit of
QList.
Change-Id: I9cac3a70af156971022c1e6084cb5538c050f826
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@theqtcompany.com>
In case the static destructor has already been run, make sure we don't
crash. This shouldn't happen, but could happen if the QApplication
destructor is run in a weird order (after the static destructors have
begun running).
That's not usually a case we'd fix (unsupported), but since this change
improves the code and also avoids creating the pixmap cache if it hadn't
been used up until this point, the change is a net benefit.
Task-number: QTBUG-48709
Change-Id: Ia505aece07bf4e13a1faffff140f3e119cfc773e
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
On 32 bit platforms the pointers may end up being 4 byte aligned.
Happens with MSVC on 32 bit Windows. _mm_store_si128 is documented to
require 16 byte alignment.
Change-Id: I80737fedf9e7f436a51a83924117cc0bc63017cc
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@theqtcompany.com>
This bug occurs if ./configure is called with -v on systems on
which fd proc entries point to the files/devices they are open
on instead of being magic nodes which would basically dup() the
actual fds (e.g., Linux).
In this case, the command "tee $tty" appends to /dev/stderr, which
may be already opened by the parent process. This breaks the log file.
Normally, the log file starts with
This is the Qt Open Source Edition.
...
but with `-v` flag it would start with output of awkprog and maybe
some zero bytes. Zero bytes are observed on Debian Wheezy.
DEFAULT_INCDIRS=...
...
^@^@^@^@^@^@^@^@^@^@^@^@...
Done running configuration tests.
...
To fix this problem, the output of `...` is saved to a variable, and
then eval'd and echo'd (if -v).
This solution was found by Tony Theodore.
https://github.com/mxe/mxe/issues/938#issuecomment-149770348
Change-Id: Id0c28598890e813774cc92f38ee46a0697b34e77
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This is in preparation of adding a putString(QLatin1String) overload.
To keep the change simple, I kept the construction of the padChar
chain per putString() call. It probably makes sense to have a
QString::resize(int, QChar) to perform the padding operation in-place.
Let's leave that for another changeset.
Change-Id: I9ef66b8df38117e1669fd2bece6ee180a2ce3369
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This was never a documented feature, but happended to work before Qt 5.5.
It broke because the peoperty access went into the static function and are
now prefixed with '_t->'
So restore the behavior as it was by not including the parentheses in the
member name.
Task-number: QTBUG-47695
Change-Id: Ic3509ddea7ac9abc871e71f5bfbe81d04d08e9bc
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@theqtcompany.com>
Windows Runtime side callback is run from XAML thread and receiver
is at UI thread thus sendEvent asserts. Use synchronous system
interface key event handler to deliver the event.
Task-Id: QTBUG-48105
Change-Id: I91a8ef6fd29c277edfb699b688b9e7895dadda8f
Reviewed-by: Andrew Knight <andrew.knight@intopalo.com>
It was possible for Windows Runtime callback to run while integration
class was constructed. That caused an assert when handling
application state change. Fix this by connecting callbacks after
integration class fully constructed.
Change-Id: I029c2e1f932e8edc3665443cc17dbf11eaae1bf6
Task-Id: QTBUG-48109
Reviewed-by: Andrew Knight <andrew.knight@intopalo.com>
Initialize locale from current keyboard value and listen
to WM_INPUTLANGCHANGE.
Task-number: QTBUG-48772
Change-Id: I53b6ef4e2cf538bb81b41ea497ed0cb66991b104
Reviewed-by: Liang Qi <liang.qi@theqtcompany.com>
Reviewed-by: Oliver Wolff <oliver.wolff@theqtcompany.com>
Use QByteArray/QString addition instead in loops and for
test row names.
Change-Id: Ieffb429efdc14aa5932b3fcdef5a18e13a62d35f
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@theqtcompany.com>
Use QByteArray/QString addition instead in loops and for
test row names.
Change-Id: Iae76d852a1657bfb6d88e84515f30bd2f1dece6b
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@theqtcompany.com>
