- Remove QPlatformFontDatabase::defaultFonts() returning
a hash containing widget name ->font and the Windows
implementation.
- Add enumeration and font accessor to QPlatformTheme. The value
returned for the enumeration value overwrites the default font
of the font database.
- Implement for Windows, Mac and KDE.
- Add more Windows palettes.
Task-number: QTBUG-23686
Change-Id: I8a2abdfd216df23daa7c9630c54264cdf61295db
Reviewed-by: Morten Johan Sørvig <morten.sorvig@nokia.com>
The function can be used only with a registered type and it would fail
to compile for other types. By adding the static assert we can print
an almost user friendly compilation error message.
Change-Id: I59ab148cabf32afe0baef186b82cb03303b57780
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
We need to load OpenType tables when initializing fonts for
scripts that require them. This fixes support for many Brahmic
scripts.
Change-Id: Ib5e50f2c7e5edb4b3e3ecf9fd004f2cf62634add
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Mention in changes and document Qt version (merci à dfaure).
Follow-up to 291e2c7d54.
Change-Id: Ie5626e9cd268812c1173ca494ccd8d6bd9be2687
Reviewed-by: Honglei Zhang <honglei.zhang@nokia.com>
When an aggregate function is used for a column in a SQL resultset then
it should ensure that the right data type is reported for that column.
This also concerns expressions when the returned column does not map
directly to a table column.
Test included for this.
Task-number: QTBUG-22038
Change-Id: I07487694c0ed393d46af06e232914fe923356a99
Reviewed-by: Mark Brand <mabrand@mabrand.nl>
Reviewed-by: Honglei Zhang <honglei.zhang@nokia.com>
Mac has to work with -qtnamespace as well, and other files have gotten
this right, so follow their example.
Change-Id: I551e1843e8a0e82a82d1d5ea8c8cd5f20e6880fa
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
The studyData pointer is atomically set by the pointer assignment,
but another processor running a different thread might see the
new studyData value but not the memory it points to.
Therefore, the current studyData is returned from optimizePattern
and used by that thread.
Docs were added to optimizePattern to explain what's going on.
Change-Id: I4502c336077bb98a1751011aa93ffd4f585ed101
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Until now double clicking in Qt 5 resulted in the following sequence of mouse
events: pressed, released, double clicked, released. This is wrong, the press
belonging to the second button down is missing. In Qt 4 that pressed event is
present.
The problem is not apparent in desktop environments because the double click is
functioning properly even when the second pressed is missing. However when
using a platform plug-in like wayland, where the clients receive only press,
move and release events, double click was broken because the second click was
effectively ignored (due to receiving nothing but a button release).
Change-Id: Ief6af12c666b23e544da4a68cb835cd577265469
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
I tested, it didn't work. This is from a previous patch I had
already gotten working before 40a5ba4d3fccb449dcfd8d9a0deaf4c7f0fe12bc
was submitted.
Change-Id: I868f069fe834b3122ed9b5b3dc9af0781d6e1d0d
Reviewed-by: Laszlo Agocs <laszlo.p.agocs@nokia.com>
Removes the usage of various qdoc macros which are now deprecated.
Change-Id: I74fa70f8d2a2a1bff57cdb2bcc14a31a7198dea0
Reviewed-by: Casper van Donderen <casper.vandonderen@nokia.com>
Do not use the value of a macro before verifying that the macro
is defined.
Change-Id: I36bebe37da5f4e5e7af1e423b7f2b18091e35707
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Now that indexOfChild is 0-based, the update notifications should follow.
Change-Id: I5e0303516d503d5e23061df5894b2428c00da2ce
Reviewed-by: Jan-Arve Sæther <jan-arve.saether@nokia.com>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@nokia.com>
Certain versions of system headers will declare WCHAR_MAX like:
#define __WCHAR_MAX ( (wchar_t) - 1 )
#define WCHAR_MAX __WCHAR_MAX
In particular on ARM (see e.g.
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=598937 )
In this case, defined(WCHAR_MAX) is true, but attempting to use the
value of WCHAR_MAX in a preprocessor expression will not give the
desired results - "wchar_t" is unknown to the preprocessor, so
WCHAR_MAX silently (without -Wundef) evaluates to ( (0) - 1 ) == -1.
A simple workaround is to avoid looking at WCHAR_MAX when the
superior __SIZEOF_WCHAR_T__ is defined.
Change-Id: I439b166cffb93416737ee19025fb6e8d51c27876
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Commit 8327fa7c11 changed the type of the
Julian day member of QDate from quint32 to qint64. This changed the
QDataStream format. Keep the old behavior, with the limited date
range, if the stream version is less than Qt_5_0.
Change-Id: I800448979a1891581069f39de7f9ab9c634e4f0e
Reviewed-by: John Layt <jlayt@kde.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Using typedef causes errors due to re-definition, so #define the types
needed by the SHA-2 code to the q[u]int* equivalents instead.
Change-Id: I6fc29788dd05aeee28723820f511527d482d31f2
Reviewed-by: Oliver Wolff <oliver.wolff@nokia.com>
Calling select refreshes the query data but disrupts view
navigation.
For OnFieldChange and OnRecordChange it makes sense to only
select the row in question. This does not disturb view navigation.
Assume disruption of view navigation is not a problem
for OnManualSubmit because the user or application decides
when submitAll is called.
Task-number: QTBUG-2875
Change-Id: I1e5f68668fb9102f6296d67d543e80daa403f1c4
Reviewed-by: Yunqiao Yin <charles.yin@nokia.com>
Once an insert has been submitted, the cached record behaves like an
update. For row bookkeeping, we still have to remember that it was
originally inserted and is not in the query rows.
Between submitting a delete and selecting, we remove the values
from the deleted record. This causes a blank row to be displayed.
Read-only flag is set for cells in deleted row.
Reverting between submit and select means going back to the last
submitted values.
When removing rows, it's better to process from highest row numbers
to lowest. This avoids complications with higher rows shifting down
when lower rows are removed.
Change-Id: I8752fa11f7a1b88f2a71b9e03a020ac37e62487f
Reviewed-by: Honglei Zhang <honglei.zhang@nokia.com>
Adding some const qualifiers to members which are never written.
Change-Id: Ibb8953764c7b7790a419a5d48f2956751d5fc1f9
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
QRegularExpression::captureCount() returns the number of
capturing groups inside the regular expression pattern.
Change-Id: Ib90ce67c67d06ab2966f0c98bd91da21defc156d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
PCRE's JIT uses self-modifying code extensively, requiring full SMC
checks enabled by tools like valgrind, which slow down the execution
considerably; not enabling SMC checks lead to crashes.
Therefore, JIT is now disabled by default in debug builds of Qt.
Its usage (both in debug and release builds) can be controlled
by setting the QT_ENABLE_REGEXP_JIT environment variable.
Change-Id: Ib38952400e4219582942ce65ab9edcd89c432f3e
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Trivial change: compare dpointers first, then the data.
Added test function for operator==.
Change-Id: I33ac64a59db4ccad56c30be17622187e42415f38
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Added the proper QDebug operator to debug the
QRegularExpression::PatternOptions flags.
Change-Id: Icd00e93a0c6cc4345db528d494fc176624f7b7a2
Reviewed-by: hjk <qthjk@ovi.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Added QRegularExpression, QRegularExpressionMatch and
QRegularExpressionMatchIterator as PCRE-enabled, regexp classes.
Documentation is included, as well as a first round of autotests.
Task-number: QTBUG-23489
Change-Id: Id47031b80602c913ccd2fd740070e3024ea06abc
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
The test was testing the wrong thing, and passing even though
QNetworkRequest::AuthenticationReuseAttribute was not being
respected, until recently when I fixed username/password in URLs
Now the cache is properly bypassed when this attribute is set to
manual, and the autotest is updated to check this.
Change-Id: I87943515562d0b16b03504f0758ba265758d1c22
Reviewed-by: Martin Petersson <Martin.Petersson@nokia.com>
Loser of the race would try to delete an uninitialised pointer
Task-number: QTBUG-15765
Change-Id: Ie184ee2306e102aa8fbad752ef09b95c3ede00c2
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Use a QBasicMutex, there is no extra cost of having a mutex for this.
Change-Id: Ib5b01338649002c0c21f018b2c931a8cc68027f6
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
The qt_global_mutexpool was private API deprecated long time ago.
And there is no reason to call qt_create_tls because it is called in
QThreadData::current that is called from the QObject constructor, even
before QCoreApplication::init can be called.
Change-Id: Idf3576d8591377811b727b12edc43dc898570ba4
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
The mechanism in fontconfig which determines if a certain character
is available (FcCharSetHasChar()) may give false positives, in which
case we would load and unload those fonts per every char for which
FC gave us a false positive. This was a major performance regression.
Specifically the false positives happened when looking at e.g.
italic variants of certain multilingual fonts, since we only check
the charset of the font family as a whole and not of the specific variant,
which may only support a subset of the chars.
To optimize this, we remove the deletion of the font engines after
loading them, but also wait with loading the opentype tables until
they are actually needed. This means that for the false positives,
we will load the font, but the cached data for each unused font will
be much smaller.
Change-Id: Idfc794401a2080da5946bf65204eb947aeb635ed
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
This internal function is abused by some modules to create dynamic
types in metatype system. In Qt5 more non-optional arguments were added
to the function and to keep temporary source compatibility an overload
was created.
QMetaType code assumes that every known type has properly defined basic
operations like creation and destruction. Setting a helper function
pointer to null value is asking for a crash, because the code doesn't
check for that value, the null pointer may be called.
Change-Id: I5ca7454a70c308e01de26fab23481b3c94c22371
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Add roles to QPLatformTheme::Palette, map QWidget
subclasses to those. Port Qt4 widget palette creation
code to use the QPLatformTheme::Palette roles.
Palette entries are disabled in this commit, this
will be fixed later.
Change-Id: I07babe3d7c76d306efc4ea4813c7161fdf36227f
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
QSslCertificate can be copied around into multiple threads,
without detaching. For example, the https worker threads inside
QNetworkAccessManager.
There are const methods, which lazily initialise members of
the private class without detaching (i.e. caching results of
expensive function calls)
These functions now lock the d pointer using QMutexPool to
avoid concurrency related crashes.
autotest crashes 20% of the time in release builds without
the fix, passes 100 times in a row with the fix.
Task-number: QTBUG-20452
Change-Id: I64a01af8159216f2dd6215a08669890f6c029ca8
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Richard J. Moore <rich@kde.org>
This is due to the search in the suffix tree starting at position
fileName.length() - 1.
Change-Id: I98501c1724c7dde2626351ace8ba19faa0d2e1e1
Reviewed-by: Ivan Komissarov <ABBAPOH@nextmail.ru>
Reviewed-by: Wolf-Michael Bolle <wolf-michael.bolle@nokia.com>
Docuent the requirement that the alpha channels need to be properly
initialized by the implementation.
Change-Id: I03db81b44b43ea75feb1b983fb0725c65a3bd9f4
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
Fixes a crash when QWindow::baseSize() is invalid size.
Change-Id: I4e41f63d69ad0f218bfd35db8f30f18f92d4e9d5
Reviewed-by: Morten Johan Sørvig <morten.sorvig@nokia.com>
Similarly to change id I2f429fa7ef93bd75bb93a7f64c56db15b7283388, the capability
to arbitrarily alter the encoding of literals is very destructive, especially in
a world with libraries and plugins.
Change-Id: If0d4cd8dcf89792e39c1984cbde6b036cebfc02f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Each channel will create a socket that will allocate memory for the
read and write buffers. QNam generaly inits 6 sockets for each
connection. That means that by default 12 such buffers are created.
This will instead initialize channels when they are needed.
Change-Id: Ie3f2cf789e084fd3d17d3b2a9bb3d3a4370b3da4
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
As per discussion with Lars, intent here was to allow plugins without
a "Keys" property to still function correctly, but this particular
if statement was blocking any such plugins from being detected.
Change-Id: Icb343ca8bd95a508d62565cd816fe2a57a4f82bd
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Implicit conversion from an int would look strange in this case.
Change-Id: I2222a045c293595d7b83a2fb75ca646f5cf79bca
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Simplify logic. If the record is in the cache, even untouched
values should be there. This is also necessary for getting
the most up-to-date values between submitting and the next
select.
Change-Id: I8578d96229797ce9fb0d07fe456301358f2be071
Reviewed-by: Honglei Zhang <honglei.zhang@nokia.com>
Primary values are used to map a row in the model to a row in the
database table. It is critically important between submitting a
change and the following select (which refreshes the query) to have
updated primary values. Otherwise, if the change affected the primary
values, additional changes before select will misbehave.
Change-Id: I5d08dd70ac5d3f06cd9d3186a439f4c80a037c2d
Reviewed-by: Yunqiao Yin <charles.yin@nokia.com>
Between submitting and the next select, these values will be more
up-to-date than those that could be obtained from the query.
This will be useful for constructing primary values and reverting
changes made after submitting.
Change-Id: I8317617f3e7043ad0b79b333731c55fb88aef171
Reviewed-by: Honglei Zhang <honglei.zhang@nokia.com>
Setting the user time before mapping the window seems to prevent the
crasher from happening. We used to set the user time before mapping in
Qt 4.8 too, so it's probably the right thing to do.
Task-number: QTBUG-24462
Change-Id: Ia670b799bd1ed7a7e6399631d5242e57324918b3
Reviewed-by: Jan Arne Petersen <jpetersen@openismus.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
This event was completely unused.
In addition it leads to crashes on linux when
sending the Destroy accessibility update.
The Destroy event on linux would still query an accessible interface.
That in turn would trigger the event to be sent.
Change-Id: I8915527de067b8b70ba41b1361e3ef5d12866d7d
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@nokia.com>
Reviewed-by: Jan-Arve Sæther <jan-arve.saether@nokia.com>
Cache the first QPlatformPrinterSupport returned from the first
QPlatformPrinterSupportPlugin, and treat it as an persistent singelton.
Change-Id: Ic1c83d7c1cdf4a09723a74e0b9fd485e0b0b3acb
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
These static public functions give printsupport plugins access to the
non-exported qt_paperSizeToQSizeF() and qSizeFTopaperSize() functions in
qprinter.cpp to aid implementing QPrintEngine and
QPlatformPrinterSupport::supportedPaperSizes().
Change-Id: I3ebcdcd17e863b06ceb135e096e630b37882a293
Reviewed-by: Morten Johan Sørvig <morten.sorvig@nokia.com>
We have a QPlatformPrinterSupportPlugin on Mac OS X to return the
QPrintEngine/QPaintEngine, no need to keep this code.
Change-Id: Ie24dcfd157810ede69790fc7b27c12e24766efce
Reviewed-by: Morten Johan Sørvig <morten.sorvig@nokia.com>
Naming threads is very useful for release builds.
Enabling only on Linux/Mac for now.
The Windows port is using debugger specific API for setting thread
names, so it has to remain debug mode only.
Change-Id: I179521f65f215ff038e8230f958f6aa728ea4cbe
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Make sure that qdoc can find the same function signature in the header
than in the \fn tags in qobject.cpp
Change-Id: Iccf2ba4e8f6384e9c3bfc878a446120f03e8a813
Reviewed-by: Casper van Donderen <casper.vandonderen@nokia.com>
Imported only the minimum subset of PCRE version 8.30,
without documentation, tests, examples, etc. by using
the import_from_pcre_tarball.sh script.
Change-Id: Ib1112a6179789814cc3c90cbde59408d2ccd0f57
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Got rid of unnecessary debug output, and prefixed debug / warnings
with class name to provide more context.
Change-Id: Ia68e85ec0207de4d87d02226bdf6e4cddf464afc
Reviewed-by: Uli Schlachter <psychon@znc.in>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
In many situations it would be very nice to have setRange
as a slot. It fits good with the rangeChanged signal -
and in some situations it does make sense to synchronize
scrollbars ranges.
Change-Id: I6bcb41ed6e009e5822b56b621e4e187fe52502a6
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Changes various internal usages of QMap when the key is a pointer type.
Being ordered by the pointer value itself, it makes very little sense
to use QMap<K*, V> (esp. in cases where the key is actually allocated
on the heap).
The usages have been found with the following script:
#!/usr/bin/perl
use strict;
use warnings;
use Regexp::Common;
use File::Find;
use feature ':5.10';
my $container = qr/(?:QMap)/;
sub process {
return unless (-f and -r and /(\.c|\.cpp|\.h|\.txt)$/);
open my $fh, "<", $_ or die "Cannot open $_: $!";
while (my $line = <$fh>) {
chomp $line;
while ($line =~ /($container\s*$RE{balanced}{-parens=>"<>"})/g) {
my $tmp = $1;
$tmp =~ s/\s+//g;
say "$_:$.: $line" if $tmp =~ /^$container\s*<[^,]+\*,/;
}
}
close $fh;
}
find( { no_chdir => 1, wanted => \&process }, @ARGV );
Change-Id: Idd9819c3f4c48f98ef92831d5e8e5ac0fa42283c
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
QDoc now has support for Doxygen style commands for italics, bold
and list items. This change applies that change in QDoc to the
actual documentation.
Task-number: QTBUG-24578
Change-Id: I519bf9c29b14092e3ab6067612f42bf749eeedf5
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
updateAccessibility(const QAccessibleEvent &event) should be used,
updateAccessibility(QObject *object, int child, Event reason) is deprecated.
Change-Id: I92b32579ff202681189e9581365d2891e5d1e994
Reviewed-by: Jan-Arve Sæther <jan-arve.saether@nokia.com>
When there is no QObject subclass with Q_OBJECT macro,
we must at least set an object name
Change-Id: Ib429a9b246d9d6b4b4cfb11593e4f358850677f2
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
This change makes sure that the X11 server really supports the SHAPE
extension before it is used.
Before this, libxcb would disconnect from the server as soon as it was
told to send a SHAPE request to a server which doesn't support SHAPE. Oh
and of course it would do so without any good error message.
Change-Id: I5f9481e488925e74e62f229e9f637ef262020c04
Signed-off-by: Uli Schlachter <psychon@znc.in>
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
The application / library that uses this should know it needs to link
to ie. udev, xrendr etc.
Change-Id: I9e3ffd4d98b3bd66abb8e621378b010da7feda7a
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Laszlo Agocs <laszlo.p.agocs@nokia.com>
This means that the AT client knows it should not have to ask for
the focus child anymore.
Previously, some clients (NVDA) kept on asking infinitely because of
this.
Change-Id: Ia2bd2e1088a899f7d1a0c9189024accdacfd693b
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@nokia.com>
set QT_DEBUG_ACCESSIBILITY=(1|0) to turn logging on or off
Change-Id: Ibd5b77699decf0cf02bc6b6cc656fa237de29124
Reviewed-by: Jan-Arve Sæther <jan-arve.saether@nokia.com>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@nokia.com>
QWizard always forced Aero style even if Classic theme was used.
Added a check for Classic theme into initialization.
Task-number: QTBUG-24618
Change-Id: I0793df62e022037835e1c1a311143f06cc4382b4
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
qHash() returns uint, not int, so change all interactions with hashing to use
uint to match.
This blocks the introduction of a new (better) hashing algorithm because it
currently breaks numerous tests: rcc would (correctly) write a uint hash value
to the qrc files, but QResource would attempt to mangle it around as an int.
This wasn't a problem with the old hash, because it deliberately threw away
data (h &= 0x0fffffff), possibly because of someone not being able to
diagnose precisly this problem.
Change-Id: I46fb42acc100fdd3bedd714f6dc91aeca91d0351
Reviewed-by: hjk <qthjk@ovi.com>
Derive a data accessor class from QMimeData so application level code can
query MIME type availability through it or directly attempt to fetch
data by MIME type even if the type is not yet known to QBBClipboard.
Also make sure any MIME type used by the application in setMimeData
operations is checked for availability when being queried for formats.
Change-Id: If8174adf3f3383496a354d291ef7d36b9732be1d
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>
The application root window is at z-order == 0, all platform windows created
for Qt windows are its descendants, so their base value for z-order needs
to be 1.
Change-Id: I7c9615ffac6757c31900a7640eb0ffdd6065b722
Reviewed-by: Sean Harmer <sh@theharmers.co.uk>
Reviewed-by: Nick Ratelle <nratelle@qnx.com>
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>
Reviewed-by: Kevin Krammer <kevin.krammer@kdab.com>
Change-Id: I62dbc5a695e41179de9f6acd11aa7bc592cac6f3
Reviewed-by: Clinton Stimpson <clinton@elemtech.com>
Reviewed-by: Alexander Neundorf <neundorf@kde.org>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
This reverts commit 7c0d15a22266a425c9e9ac0120d6774e120fe01e.
The commit caused a regression whereby tooltips may be shown even if the
mouse is not over the item if it has the Qt::ItemIgnoresTransformations
flag and the QGraphicsView had been scaled.
Task-number: QTBUG-17517
Task-number: QTBUG-22663
Change-Id: Ib7fd788d9712c5e659fe07182f9505a4eb135ab2
Reviewed-by: Andy Shaw <andy.shaw@digia.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
(cherry picked from commit 15c14584199dc43e4a309fc331f3144009008128)
This allows the QCoreGraphicsPaintEngine to use QPaintEnginePrivate
members (e.g. pdev and drawBoxTextItem()).
Change-Id: I5bed3cb007ae469816afce619edd55f155b04fa9
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
Do not write Q_BYTE_ORDER to qconfig.h in the configures. Instead,
we #define Q_BYTE_ORDER in qprocessordetection.h, since many CPUs only
support a single endian format. For bi-endian processors, we set
Q_BYTE_ORDER depending on how the preprocessor sets __BYTE_ORDER__,
__BIG_ENDIAN__, or __LITTLE_ENDIAN__ (instead of using a compile test
to do so).
For operating systems that only support a single byte order, we can
check for Q_OS_* in addition to the preprocessor macros above. This is
possible because qprocessordetection.h is included by qglobal.h after
Q_OS_* and Q_CC_* detection has been done. Do this for Windows CE,
which is always little- endian according to MSDN.
Change-Id: I019a95e05252ef69895c4b38fbfa6ebfb6a943cd
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Screens need to be destroyed in reverse order to ensure the primary
screen stays valid when other screens are destroyed.
Task-number: QTBUG-24300
Change-Id: I9d9d710aa67ec045baa8bf292833ffe7d9eea935
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@nokia.com>
Do not draw on the backingstore DC when drawing
to a pixmap. Access the paintdevice for checking via
the QPaintEngine since QPainter returns the clipdevice,
which is a widget.
Fix warning about accessing handle of 0-window in the
test.
Task-number: QTBUG-24183
Change-Id: Ie91ea6ab9d09528c7ec1d35633f9a0ee667719b1
Reviewed-by: Miikka Heikkinen <miikka.heikkinen@digia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
This commit is complimentary to the commit which introduced a similar
partial specialization for single template argument types:
6b4f8a68c8
If T and U are available as metatypes, then QHash<T, U> is too.
Change-Id: I09097b954666418b424c8c23577032beb814343a
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
The Kernel's evdev module is capable of presenting up to 16 mouse
buttons, plus wheel events (UP, DOWN, LEFT, and RIGHT). This
patch updates our Plugin, so that it tranlates all of those button
numbers into corresponding MouseEvent Qt::MouseButton values.
Task-number: QTBUG-24590
Change-Id: Ib8cbb9500280c76f06a51ce95095e22ae84de1c6
Reviewed-by: Laszlo Agocs <laszlo.p.agocs@nokia.com>
Change the help output to show which options only work for plain text
logging.
Change-Id: I39eb7cd0793cdbe553c230334c6cd532b4929f61
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Also adds optional support for libmtdev (a helper library which
translates transparently from type A to B when using type A kernel
drivers).
Change-Id: Ic9c065b2fd130e9db2dd07e7dc103e9d45c08c99
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
instead of being a variable added to the makespec (via qconfig.pri),
QT_SYSROOT is now a property.
the QT_INSTALL_... properties are now automatically prefixed with the
sysroot; the raw values are available as QT_RAW_INSTALL_... - this is
expected to cause the least migration effort for existing projects.
-hostprefix and the new -hostbindir & -hostdatadir now feed the new
QT_HOST_... properties.
adapted the qmake feature files and the qtbase build system accordingly.
Change-Id: Iaa9b65bc10d9fe9c4988d620c70a8ce72177f8d4
Reviewed-by: Marius Storm-Olsen <marius.storm-olsen@nokia.com>
this is used only in qmake, not in other bootstrapped tools
Change-Id: Ie2841e69dbd82c86d2297ddf51443ee75760766c
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
Reviewed-by: Marius Storm-Olsen <marius.storm-olsen@nokia.com>
Otherwise metacity was crashing when a QWindow was destroyed immediately
after being activated, because metacity was trying to select events
(XSelectInput) for the already destroyed m_netWmUserTimeWindow.
Task-number: QTBUG-24492
Change-Id: Iedbe7bdd6b26110ca8bec6f33525209ae551ffd5
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
In previous commits we took into account bound shaping, but X also
supports input shaping, so make sure it's inside both input and bounding
shaping to consider the position as inside a window
My tests show that when unset Bound Shaping and Input Shaping return the
rectangle of the window itself, so we need to be inside both of the
rectangle sets to consider the position as a dragabble position for the
window
Change-Id: I77973e51168d2a6d274d166f29f32e1bc02c32c3
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
There's not a single in-tree user of this function, and
the concept is a broken one in MT programs: By the time
qIsDetached() returns, the result can already be
different due to another thread taking a copy, or a
copy in another thread being destroyed (note that this
doesn't require mutex use by the user, since we promise
(implicitly, if not explicitly) that you can copy from
const objects without holding a lock).
QTBUG-10813 talks about a use in QCache::trim(), but
677cf76340 removed it, so
there's no reason to keep it anymore.
Change-Id: I20380c12bdf00ac764b89d84392f0f34727b1971
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Report https://bugs.webkit.org/show_bug.cgi?id=79561 shows that
this commit will cause significant slow down in text handling
operations. Though the exact reason is unknown we should revert
it first and find out the reason later.
This reverts commit 692064bcfd.
Change-Id: I16a56c3093bdfa2119ab6a6e9049ef2925468e29
Reviewed-by: Simon Hausmann <simon.hausmann@nokia.com>
Remove the -armfpa option the config.tests/unix/doubleformat*
detection. The places where we used QT_ARMFPA and Q_DOUBLE_FORMAT
has been removed as well.
Rationale: ARM FPA with GCC does not work with EABI. Qt currently
does not support compiling without EABI, making ARM FPA an
impossibility. It is unknown whether other compilers provide ARM FPA
support with EABI. Support for ARM FPA can be re-added in the future
should the need arise, but since ARM VFP is available for ARMv5 and up,
we should encourage implementors to instead use soft-floats or VFP.
Change-Id: I3671aba575118ae3e3e6d769759301c8f2f496f5
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
This change enables us to instantiate a QFontEngineMulti that takes
the raw font's font engine as its primary engine but can use fallback
engines based on the platform. Since this can be quite expensive, we
defer the query for fallback families' names until it's needed and
we cache the resulting multi font engine.
Change-Id: I390dbc1cb2fe61d56867f29a03f313eb3eb49dc3
Reviewed-by: Simon Hausmann <simon.hausmann@nokia.com>
The purpose of this change is to allow shaping complex
text while using QRawFont.
This is needed for WebKit so that we can switch to
using QRawFont everywhere and be more in line with
what other WebKit ports do.
Since this change slightly bends the aim of QRawFont,
let's reserve this for internal use for now.
Change-Id: I5ec0881f50ce288350bd277570cb5f1fb70c355c
Reviewed-by: Simon Hausmann <simon.hausmann@nokia.com>
Specialise QTypeInfo<QPair<T1,T2>> based on the properties of
T1 and T2:
- If either T1 or T2 is Q_COMPLEX_TYPE, so is QPair<T1,T2>.
- Otherwise, if either T1 or T2 is Q_MOVABLE_TYPE, so is QPair<T1,T2>.
- Otherwise, QPair<T1,T2> is Q_PRIMITIVE_TYPE.
Change-Id: I8aecbd37e3b7924f77f38967498deabf1a19ca24
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
Causes a new test failure in QGraphicsProxyWidget::updateAndDelete,
expand the EXPECT_FAIL to cover that one as well.
Change-Id: If9da757206445d17510c4fac82a559de7b8c0563
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
The OS provides the error string in this case.
This gives more information to the developer seeing a generic
error.
Change-Id: Ia03642982f3513ee5a8a9fa98d918e948f8d97a5
Reviewed-by: Miikka Heikkinen <miikka.heikkinen@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Richard J. Moore <rich@kde.org>
When loading dnsapi.dll as a plugin, we experience crashes because
the calling convention is not specified.
The default is _cdecl, but __stdcall (via the WINAPI macro) should
be used for windows APIs.
Mismatched calling convention results in corruption of local variables,
probably because the stack pointer is incorrect and SP offsets are
used in optimised builds rather than frame pointer offsets.
Since the library has been available since Windows 2000, I don't
think that we need to load it dynamically.
(Unlike the unix version where it isn't part of the LSB)
Also checked that the current release of mingw works.
Task-number: QTBUG-24227
Change-Id: I37c0a6aa0c133799c2a6dd9391ca1435ba2539ea
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Some drivers use this as a hint to decide on weither to preallocate
mipmap memory or not.
Change-Id: I2fd438a9625b658c7f30fe39a9d63ba5396f9679
Reviewed-by: Kim M. Kalland <kim.kalland@nokia.com>
Use a _NET_ACTIVE_WINDOW client message instead of xcb_set_input_focus
for activating toplevel windows on xcb.
According to the Extended Window Manager Hints the right way to activate
a top-level window is using _NET_ACTIVE_WINDOW (when it is supported by
the WM). Other approaches like calling xcb_set_input_focus should be
avoided when possible, since the WM cannot intercept them.
Change-Id: I9be4901f56cbcfb563baf73ccd71ff17a9bdc1d2
Reviewed-by: Uli Schlachter <psychon@znc.in>
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Do this regardless of whether the event subclass
is public API or only used in examples. Examples
are examples, used by others as templates or even
copied verbatim, so they should also follow sound
engineering rules.
Anyway, there's only one in examples/...
Change-Id: I586ff16407a956c9e89288fdd4377eed73f45c0f
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
And moved the pageflipper into its own file
Done with: Paul
Change-Id: I0af34075ce8673a66025cb761c0fe4ff6c0ab0fe
Reviewed-by: Jason Barron <jason.barron@nokia.com>
Those escape sequences have a special meaning in the XML Schema 1.1
regular expressions, but not in Perl-compatible ones.
An escape sequence that has no special meaning should match the
escaped character itself; this patch fixes QRegExp's behaviour in
that regard (previously, it added a character class matching
nothing).
Change-Id: I983f923baa7c2ec19938b96353f3a205e6c06d58
Reviewed-by: Andy Shaw <andy.shaw@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The code fails to compile with the below error:
qcryptographichash.cpp:55: error: conflicting declaration 'typedef quint64 uint64_t'
/usr/include/sys/types.h:99: error: 'uint64_t' has a previous declaration as 'typedef __uint64_t uint64_t'
FreeBSDs types.h defines the used defines. Maybe it would be less
ugly to switch the code to quint*, or use a define to do so, or to
have basic os detection for stdint.h, not to include sys/types.h.
Change-Id: Ic62ae4b742c1123b4b7e17158d216374e609f59f
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Windows unhelpfully writes to only one byte of the output buffer
when getsockopt is called for a boolean option. Therefore we have
to zero initialise the int rather than initialising to -1 as was
done before.
This in general only works for little endian architecture, because
the word would look like 0x01000000 on big endian. So I have added
some compile time asserts in the assumption that windows is always
little endian. This is ok for comparisons with 0/false, but not
comparisons with true or nonzero values.
In the case of IPV6_V6ONLY, it is documented as DWORD (unsigned int)
but on some windows versions it is returned as a boolean triggering
the warning. I removed the warning, as the conversion to int works on
both LE and BE since it is only compared with zero.
Task-number: QTBUG-23488
Change-Id: I3c586d1ada76465fc045a82661f289920c657a4c
Reviewed-by: Richard J. Moore <rich@kde.org>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Andreas Holzammer <andreas.holzammer@kdab.com>
I originally tried to put Q_DECLARE_TYPEINFOs into
Q_DECLARE_OPERATORS_FOR_FLAGS, to declare not only
the flags type, but also the underlying enum as
primitive, but too many users (arguably correctly)
used Q_DECLARE_OPERATORS_FOR_FLAGS at (non-global)
namespace scope where QTypeInfo would have been
specialised in the wrong namespace.
So specialise QTypeInfo for QFlags<T> only.
Change-Id: I4af6e29aefbd9460a3d2bc6405f03cdf6b1096bc
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Make qDebug work again with the new logging framework.
Change-Id: Ib88a83182429636b274d6284933d5ea00db7279c
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
- Give QPlatformDrag a synchronous drag() function returning
the Qt::DropAction
- Move the base functionality for asynchronous event handling
to the platformsupport library as QBasicDrag (extendable base class
handling drag icon and providing new virtuals) and QSimpleDrag
(sample implementation for drag within the Qt application).
- Change the Windows implementation accordingly.
- Change XCB to be based on QBasicDrag.
- Clean up QDragManager.
Change-Id: I654f76f0e55a385ba189bd74f3ceaded6a8fe318
Reviewed-by: Morten Johan Sørvig <morten.sorvig@nokia.com>
This made sense only for Symbian where there was a special
CFbsBitmap-based backend present and it was used from the
Symbian-specific VG and GL pixmap implementations.
The generic version is merely a useless wrapper over QImage and is not
in use anywhere in the codebase.
Change-Id: I1dabe22dfb8cbbc35dce8e22703a3aff810fb5f9
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
Reviewed-by: Gunnar Sletta <gunnar.sletta@nokia.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
So this code still compiles with QStringBuilder
QTest::newRow("foo") << foo + bar;
Change-Id: Ie82a21c8fd07ec2d27ad85015aa1a0c4e94700d6
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Was missing a variable declaration and an explicit cast.
Change-Id: I4f0fb9c3d9b8472adf0d91036442adc1fe255c7e
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
The copy constructor of QCH is disabled, so there's
no point in providing an implicit conversion from
the Algorithm enum anyway, so make the ctor explicit.
Change-Id: I4ea74ffb0963b4f49415da17778c3e6050454a6b
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
switch blocks are noisy. this is nicer.
reshuffled the LibraryLocation enum to make table lookups possible and
future-safe.
using pointer-free tables to avoid adding data relocations.
Change-Id: I70ec2c2142ce02a15e67284e4b285d754d930da3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
configure always defines all of these constants. the exception is
SettingsPath which is unix-only, so make the #ifdef explicit about that.
Change-Id: I339d2d7cb9d188a8e74d79310c3a80b5d4dbb806
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Move definition of ExtraData to the implementation file.
As a side effect, we need to include qhash.h
in some other places.
Change-Id: I8bb4ec0940ae51c7d6961c9a51adb80fd444e1e3
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
As a consequence, we have to add more explicit includes.
Change-Id: Ib3137031f0554b846c7bbd08f1f7df10dfeb8e61
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
I'm not even sure whether you could implicitly
convert a T& into a QScopedValueRollback<T>,
seeing as the constructor takes a non-const
reference, but it looks wrong without explicit
and
QObject o = new QObject(parent);
also won't compile even with implicit QObject(QObject*)
under a conformant compiler because of the disabled
copy constructor, and we still make QObject(QObject*)
explicit, so add it here, too.
Change-Id: I722a6e8431644e450fe2b401ccfb707a8e982380
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
A previous patch ensures that we have exactly one section in a Span.
( see SHA : b800d8b94a )
Therefore we no longer need the sectionCount variable. We have
assess to it through the sectionSpan.count. To keep this patch
quite simple the variable sectionCount has been changed to a
function returning the count value.
Change-Id: Ibc419eafa38ab64b08f93074cb6ae4b8518995f6
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
QT_NO_KEYWORDS can be used for example to ensure that foreach can
not be used, but Q_FOREACH must be, that slots must not be used but
Q_SLOTS must be, etc.
Typically they are used to avoid symbol conflict with other
libraries that may use the same keywords (I think boost uses
signals).
For 3rd party libraries, it makes sense to use Q_SLOTS and Q_SIGNALS
instead of slots and signals, so that downstreams can still choose
to use QT_NO_KEYWORDS in their code.
The most convenient way to enforce that currently is to define
QT_NO_KEYWORDS when building the 3rd party library. However, that
has the inconvenient side effect of making foreach, forever and emit
not usable within the library implementation.
This patch makes it possible for the 3rd party library to use
QT_NO_SIGNALS_SLOTS_KEYWORDS to exclude signals and slots without
affecting whether the other keywords can be used in the library
implementation.
Change-Id: If1e16a4fa384bd3a2ddd737143499f8b587bc4f8
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Richard J. Moore <rich@kde.org>
This is a private class, but it's so close to
the classical Stack(int) example for explicit
that I just have to make this ctor explicit,
too:
QRingBuffer rb = 0; // oops: meant '*rb'
now no longer compiles.
Change-Id: I7d58c1f08c1b14d14930426159c5c8db71b4cf4d
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Why would we want copy-initialisation if we can
have the default constructor?
Change-Id: Id2de36d42ef9f63793ff4e3ec36202d3f2bf5f30
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
The compiler-generated copy assignment operator is fine,
and the user-defined one prevents the compiler from
synthesising a move assignment operator.
Change-Id: I044104a2fd4d7522a910d5c2a68d11dabeca99c4
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
a project can rightfully expect a sane environment. $$fromfile() does
not necessarily provide that. so instead use include() with a target
namespace.
Change-Id: I8d6d30ab1b760d4930c9b4453bc92f8f8ad0b0ae
Reviewed-by: Joerg Bornemann <joerg.bornemann@nokia.com>
Reviewed-by: Marius Storm-Olsen <marius.storm-olsen@nokia.com>
This allows us to create correct CMake config files when Qt is
configured with directories outside of the prefix (which Qt allows),
and also allows us to use correct values when a 'longer' relative
lib directory is used such as lib/x86_64-linux-gnu.
Change-Id: I6f88255a23752dc5b84cb20ce13fdeeee9d5ad51
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Alexander Neundorf <neundorf@kde.org>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
stdint.h is a C99-ism, which isn't available everywhere. The sha.h
header tells us we need 4 typedefs. Add these to qcryptographichash.cpp
before including sha.h and comment out the stdint.h include in sha.h.
Change-Id: I1ede9569fa7eaa84de3befeb3c58cc6a05aa522c
Reviewed-by: Oliver Wolff <oliver.wolff@nokia.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
This is dependent upon the following Change Id's:
I5ebcffb7153f4216d69921d4818051e6b3d14d8a
Iec065f528f5edd848be580807a607488dc2e401f
Change-Id: I234e3c4272d7474d8f8e20fc4fea20d95c829cb5
Reviewed-by: Kevin Krammer <kevin.krammer@kdab.com>
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>
moc is currently generating code to convert from a pointer to member
function of a slot or signal to its index.
The idea was that it could be usefull for slots to have the new syntax
do the same as the old one (connecting signal index to slot index). But
in practice, the new syntax do not use the IndexOfMethod for slots.
Also, it does not work for all the slots (no Q_PRIVATE_SLOT,
no static slots)
So since it is not used, and that it would take room in the binaries to
generate all the code to get the index of slots, we remove it.
If ever we need it, we can still add it later.
Change-Id: Ia417e3e524d7915ca86433ea86c66ac2b299c81a
Reviewed-by: Kent Hansen <kent.hansen@nokia.com>
Commit 4dabe78387 changed
these functions from virtuals to inlines that check the
return value of the remaining virtual function nodeType().
However, two of the functions call nodeType() more than
once, which we know will return the same result each time,
but requires a compiler with interprocedural optimization
capabilities to figure out by itself.
So instead of repeatedly calling nodeType(), call it once
and store its return value in a temporary, and use the
temp for further comparisions.
Change-Id: Idbeafb7fd93d275d475218c6df2ad7fdc9162cc5
Reviewed-by: Richard J. Moore <rich@kde.org>
The Windows registry and INI files use case-insensitive key.
This is a side effect of 1d01bc1e83
Task-number: QTBUG-24145
Change-Id: I862bddae68ef27569e6ffa901f98ca107d3f300b
Reviewed-by: Miikka Heikkinen <miikka.heikkinen@digia.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
The code in submitAll() had the same goal as the code in select,
so it has been unified. The new code in select() avoids sending lots
of dataChanged() signals for rows that are going to be removed
by QSqlQueryModel anyway.
Change-Id: Ic22e038223720185e47ed0cc573147745ecc8fc9
Reviewed-by: Yunqiao Yin <charles.yin@nokia.com>
Previously the silent logging mode suppressed passes, skips and internal
testlib info messages, but did not suppress debugging output, making it
hard to see the fails in a noisy test. This commit changes silent mode
so that it suppresses all output except test failures and fatal errors,
making silent mode truly useful for seeing just the important test
output.
This commit also adds a selftest to verify the behaviour of silent mode.
Change-Id: I75420aead03682306210746a87e2a3b608b58fc6
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Since we're not yet confident if they serve their purpose well enough,
we have decided to make them internal so that we are free to tune them
later
Change-Id: Id79d154e0537aca07303afea5d057cfcb0773384
Reviewed-by: Morten Johan Sørvig <morten.sorvig@nokia.com>
There is no user of this in qtbase, and I don't think
this is a desireable feature anyway.
Change-Id: Ifebba5872b0eadb0daba3c3cd5f8b19abeed5c12
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
