It's just a test, but it's in the way of automatic tracking
of inefficient QLists.
Change-Id: I2dcfd81c9e208dab57bb256d7c276ad5303f196c
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
ints are only half the size of void* on 64-bit, so
QVector<int> uses only 50% of per-element memory,
compared to a QList.
Saves ~1800B of text size on GCC 4.9 optimized C++11 AMD64
Linux builds, even though it wasn't the last instance of
QList<int> in QtCore, yet.
Change-Id: Ibf04b26cff78c428e4253aaab7c280866906db58
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
The QMeta*Private classes are larger than a void*, and weren't
marked as movable, so QList<QMeta*Private> is horribly inefficient.
Fix by holding them in QVector instead. Saves ~900B in text size
on GCC 4.9 optimized C++11 AMD64 Linux builds, and tons of memory
allocations.
Change-Id: I313c965d7a0fea16f79e9fde04a972fc248e33aa
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
Instead of QHash::keys(value).isEmpty(), use QHash::key(value, 0) == 0
(0 is used a failure indicator three lines down, so it should not be
a key in the hash table).
Change-Id: I75cc4f7f7540dc5c51a7b8e3add09a0ec6a75e05
Reviewed-by: Alex Blasche <alexander.blasche@theqtcompany.com>
In case of QT_NO_LIBRARY there are no plugins after all.
Change-Id: Iad891d0cf0edf9e4418a4fe4ac49cf6497ceeb79
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@theqtcompany.com>
We have established the maximum size qAllocMore can deal with in
commit 880986be23 and we should use
it.
The maximum size for byte arrays is reduced by one byte as with
the previous code we could make qAllocMore produce ((1 << 31) - extra)
by passing (1 << 30). That is not a problem for qAllocMore itself
(as long as extra > 0) but it's hard to verify that no related
code casts the total sum back to signed int, which would overflow
to -1.
To make the compiler inline access to the maximum size, a private
enum MaxByteArraySize is provided, which can be used in internal
code.
This fixes the merge of commits
880986be23 and
c70658d301
Change-Id: Idb04856f7c2e53ef383063e7555d3083020ff2b7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Add a cleanup function for the check and disable animations
in tst_qwidgetaction to prevent effect widgets from interfering
(vista style animations).
Change-Id: I043ecb131c8dcd07b6ef10bc75c9e010ab569e85
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@theqtcompany.com>
When deleting selected text, Google Keyboard will call
deleteSurroundingText() with a negative beforeLength. This does
not make any sense, and not all our controls are able to handle
the resulting input method events.
This patch interprets a negative beforeLength as a positive
afterLength. This works with the cases I have seen so far, but
since the keyboard is not following the specification, there
may be more weirdness later.
Task-number: QTBUG-46637
Change-Id: I410832417f6fb21139c338355e8fcfa57b9e544c
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@theqtcompany.com>
Do not attempt to resolve dynamically in static builds. This is not
wrong, but if the GL library is statically linked just like the rest
of Qt, it won't work. Therefore just call the function directly in static
builds, it should be no problem on modern systems.
Task-number: QTBUG-46499
Change-Id: I35fd7c3b5b180d753becbb57377b27dd1a8747ad
Reviewed-by: Andy Shaw <andy.shaw@theqtcompany.com>
Reviewed-by: Jørgen Lind <jorgen.lind@theqtcompany.com>
Setting the version breaks on Android as it has no .so versions, leading
to aborting apps on devices that support GLES 3.x.
Change-Id: Id4381e08a2615a0898def8f075ce5a5c5b9c7a7d
Reviewed-by: Jaeyoon Jung <jaeyoon.jung@lge.com>
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@theqtcompany.com>
The "Fixed CE build of sqlite3" patch is preserved in this change,
Change-Id: I3073b51ea51403fa0011b78a510cd90b34b01068
Reviewed-by: Andy Shaw <andy.shaw@theqtcompany.com>
With Xcode7, Apple added support for compiling apps to bit code
instead of binary (*). But this is only supported when the deployment
target is at least 6.0. And in Qt-5.5, the deployment target is
still set to 5.1.1. The result is that every Qt application will fail
building once people move to Xcode7.
Instead of bumping the deployment target (which we plan to do
for Qt-5.6), we choose to switch off bit code for now.
*: https://developer.apple.com/library/prerelease/ios/documentation/IDEs/Conceptual/AppDistributionGuide/AppThinning/AppThinning.html
Change-Id: I23001563439a7726506b7cd2dc77a82533b8a27b
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@theqtcompany.com>
Not only should using the "Ex"-versions be the rule and not the
exception on Windows, but it's only the only way to share as much
code as possible between Desktop Windows and WinRT (which is pushed
by Microsoft a lot). The current rule of Desktop and WinCE vs WinRT
does not make a lot of sense any longer, as WinCE is getting less
and less important. By moving these #ifdefs in favor of WinRT,
WinCe code might be removed easier in the future.
Change-Id: I0ef94fb14fbf8add9c2dfa2a3fb8036d25fb697d
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This can lead to situations in which the gap item ends up in the
saved state
'subinfo' is owned by the item, but the const-ness is not transferred to
the pointer. So calling a non const function (unplug) that changes the
subinfo changes it in the original list as well.
This is a problem because that list might be shared with the saved state.
Change-Id: I90a5faafea5958043db610a84c069494f2e48137
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
Reviewed-by: Jocelyn Turcotte (Woboq GmbH) <jturcotte@woboq.com>
We cannot do this if no library support is present. Incidentally that's
the case when bootstrapping, but you can also set the flag manually.
Change-Id: I51e167176d0839af5858122630ef623a1c69a106
Reviewed-by: Alex Blasche <alexander.blasche@theqtcompany.com>
Verify the bounding rect of the left hand side of copy/assignment,
not the right hand side (which is passed by const ref).
Change-Id: I5044d269fe0acb5f4484c82da7e030ca33958792
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
When a font style-name is not directly available from the font we
need to generate it with the data we have, so we should include all
the new supported weights.
Change-Id: I45cc0114f8c76f45bc7a639f9d0d8096a064bc51
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Reviewed-by: René J.V. Bertin <rjvbertin@gmail.com>
Reviewed-by: Pierre Rossi <pierre.rossi@theqtcompany.com>
Otherwise, the handler can't be called for a stack overflow.
Change-Id: I5d1e6f7607404caa96e4ffff13e7fabd66011785
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
Reviewed-by: Jason McDonald <macadder1@gmail.com>
Use a QStringRef instead of a QString.
Change-Id: I63bf9775606b071cd9614306375e1316dada9f61
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
Addition of new options SSL_KEY, SSL_CERT, SSL_CA, SSL_CAPATH and SSL_CIPHER
to allow SSL-encrypted connections to MySQL databases.
When needed, these options must be specified in the function call
QSqlDatabase::setConnectOptions() before the call to QSqlDatabase::open().
SSL_KEY = the path name to the key file
SSL_CERT = the path name to the certificate file
SSL_CA = the path name to the certificate authority file
SSL_CAPATH = the path name to a directory that contains trusted SSL CA
certificates in PEM format.
SSL_CIPHER = a list of permissible ciphers to use for SSL encryption.
These options replace CLIENT_SSL (which should not be used any more).
Example:
db.setConnectOptions("SSL_KEY=client-key.pem;" \
"SSL_CERT=client-cert.pem;" \
"SSL_CA=server-ca.pem");
[ChangeLog][QtSql] SSL support for MySQL database connections has been added.
Option CLIENT_SSL replaced by SSL_KEY, SSL_CERT, SSL_CA, SSL_CAPATH and
SSL_CIPHER, so that the keys, certificates and cipher can be specified.
Task-number: QtBUG-3500
Change-Id: I8197234b169a818658678d6fcc953c90e83db23e
Reviewed-by: Mark Brand <mabrand@mabrand.nl>
... and enable auto-deletion on it.
This allows users of the function to get rid of their own
cleanup code. They just need to keep the shared pointer alive
for as long as they need it.
Drive-by changes:
- replaced QStringLiterals that were only used as the rhs of op+
- replaced an instance of mid() used as the rhs of op+ with midRef()
- enabled NRVO
Change-Id: I161d39461e020c9e8d473c0810dea2109fe0d62d
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Print the full backtrace including the values of all variables, for
improved diagnostic analysis.
Change-Id: Iffcaa08123599e3610cdc2e316627167bfff0f36
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
They didn't show up in the "old" CI runs because they usually pass the second
time they are executed - which the testrunner does. The new CI doesn't do that
anymore, instead we now mark those tests explicitly and will track their record
of passing and failing in the new metrics database.
Change-Id: Id34dd6f792f38995b07b6fec88f833df64de2f8b
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Merge QXcbWindow::setNetWmWindowFlags(), which was called from
QXcbWindow::setWindowFlags(), into QXcbWindow::setWmWindowType().
Now setWindowFlags() can't override window type set by
QXcbWindowFunctions::setWmWindowType().
Also reorder _NET_WM_WINDOW_TYPE atoms in QXcbWindow::setWmWindowType()
as it was in Qt 4.
Change-Id: Id1752d78f91caf04e9d16bb1ac40ed180597df7b
Reviewed-by: Jørgen Lind <jorgen.lind@theqtcompany.com>
Otherwise you cannot link it statically against QtCore.
Change-Id: I4ac35602cea2192974f3e96ecad35edac976ce27
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
The d_ptr variable is introduced by QGLCustomShaderStage and
needs to be fully handled by it. Delete the d_ptr in the dtor
Fixes Coverity CID #10995.
Change-Id: I76bf05206acf23363419afc95094f5593e808a2e
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
Change-Id: If9ef172bda6488c4516a5e0577d6b233bd2ed1ad
Reviewed-by: Mikko Harju <mikko.harju@jolla.com>
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
Any plugin loading is impossible then and the code won't even compile.
Change-Id: I81e49ff87ae1a609521d526f5098a0d44cc28f7c
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
This allows for example properties with QMap<Foo, Bar>
[ChangeLog][QtCore] Types in the Q_PROPERTY macro can now contain commas
(for example, QMap<Foo, Bar>)
Change-Id: Ibf5c8c9cf20a7c8b3dfec9e891fb8a9ca1bdba7c
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@theqtcompany.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Objective-C classes must be forward declared in the
global namespace.
Change-Id: Ic4818cef4b61f1578c478da9c30b77237262b283
Reviewed-by: Erik Verbruggen <erik.verbruggen@theqtcompany.com>
This prevents compiler warnings on less common branches of certain
ifdefs.
Change-Id: I89bae68491151b21cf6887a854cc4c9f68838cf9
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
QFile::decodeName should be used instead of a simple QString::fromUtf8.
Change-Id: I76955ab01af55dd373f860f6e3dbffa37e60892c
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
readlink does not append a NUL character to buf. If readlink places
PATH_MAX characters into buf, then an unterminated character buffer
would have been passed to QString::fromUtf8.
Change-Id: Ib1865b8df760fa7da91c3be746dc701a165d93ee
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Size policies can be compared for equality,
so qHash should be overloaded, too.
[ChangeLog][QtWidgets][QSizePolicy] Added qHash(QSizePolicy).
Change-Id: Id219f47ac6cb236efdd73cad7b892c8efd034d5c
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
Avoids the copying of the contents when returning devices().
Remove an unneeded #include (already includes by public
header; unneeded there, too, but leaving for SC purposes).
Change-Id: Ifa9a3dc2c7fada01fb624527dd4d16d0cb32a658
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
QPainter::clipRegion() is expensive especially for certain clip paths.
QPainter::clipBoundingRect() does not promise to return an exact (tight)
bound but delivers the desired result quickly. In addition, this avoids
discretization.
Change-Id: Ib35406edc12fb8206ca978bc140a7c5e21279ca2
Task-number: QTBUG-46578
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
It is exposed in the QtQml.StateMachine import, but note that the metaobject
version is just incremented (no need to match QtQml.StateMachine, which
is in a different module).
Change-Id: I50773d9dec5252aa93846b7e5f743958ee838023
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
Zero is a legitimate size to be returned by bytesFree/bytesAvailable
functions, so change those functions to return some 'invalid' size
in case of an invalid drive.
This is also consistent with the original version from the Qt Systems
framework.
[ChangeLog][QtCore][QStorageInfo] Fixed sizes returned for invalid drives.
Task-number: QTBUG-45724
Change-Id: I312fba521fdf8d52d7a0ac0e46cacca625775e80
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
