The basedir xdg spec says:
"All paths set in these environment variables must be absolute.
If an implementation encounters a relative path in any of these variables it
should consider the path invalid and ignore it."
Therefore we ignore relative paths including the empty string.
Change-Id: I8f779b78981018051b16de23b2514f2e62b7ab39
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The slot object was deleted after the mutex was relocked, which caused
a deadlock in case the functor destructor locked the same mutex again.
Change-Id: I5b4fb22fdb4483f91c89915872bfd548c31b0eea
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
This fixes a serious regression from
7d72516b52.
[ChangeLog] QFileSelector: the identifier for OS X has been changed back
to 'osx' from 'mac', and 'mac' and 'darwin' have now been added as
selectors for Darwin OS (which is the base of both OS X and iOS).
Task-number: QTBUG-35073
Change-Id: I83183e34c5a697338cc1ddcac33a41bd379ded12
Reviewed-by: Alan Alpert <aalpert@blackberry.com>
When changing a filter so that a previously empty proxy model becomes
populated sorting was not applied correctly.
This was caused by using mapToSource for getting source_sort_column
from proxy_sort_column. For an empty proxy model this won't work because
no valid proxy index can be created in this case.
We now directly use the root index column mapping instead by doing
essentially the same as QSortFilterProxyModelPrivate::proxy_to_source
but without the sanity checks needed for external use.
The sorting feature of QSortFilterProxyModel has always assumed that
the number of columns is specified by columnCount(QModelIndex()) so
the behavior doesn't change.
[ChangeLog][QtCore][QSortFilterProxyModel] Fixed sorting when a
previously empty proxy model becomes populated because of a change in
the filter.
Task-number: QTBUG-30662
Change-Id: I21322122e127889dfadc02f838f0119ed322dcab
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
Even if the system is not under high load, this test can generate 4
updates on Windows. This is seems to be because the timer actually
triggers before its supposed to (sometimes the second start() triggers
an update right before the animation is done (at timestamp 399 for
instance), forcing it to generate yet another timer event in order to
reach the end)).
Under high load conditions, too many things can break this test,
usually it finishes even before the pause(), causing the
QVERIFY(animation.state() == QAbstractAnimation::Paused);
to fail.
Change-Id: Id94c0d16656675bc6165b7795f3c41a30616a3ba
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
Some comparison operators comparing to const char *
where not implemented correctly.
Task-number: QTBUG-34024
Change-Id: Idbdc64c8ed93e88d9f2b2f55213bc785b33cb543
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
The documentation implies that cd() and cdUp() functions return false if
the new directory is not readable, but that is not the case. It is an
obvious mistake in the documentation, because cd'ing into a nonreadable
directory is perfectly valid.
Provided also with a test to verify that cd() actually returns true with
nonreadable directories.
Change-Id: I4aa3e859b35c64266df510a203574e3701aea77c
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Data::allocate(0) returns a pointer to read-only memory, updating d->size
will segfault. The safety check for this exists in all other QVector ctors
already.
Change-Id: Ida0fe4182de56ee62c7f91e8652cfafbfd7b8410
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
update-mime-database does not exist, so there is no point in
trying to build this test.
Change-Id: Id60f2f8fc3af33d3f7ae2fb4f1042356b60f6596
Reviewed-by: David Faure <david.faure@kdab.com>
Reviewed-by: Wolfgang Bremer <wbremer@blackberry.com>
Fix parsing of ISO Date from being totally lenient to being semi-strict
by requiring the separator positions to be non-numeric.
[ChangeLog][Important Behavior Changes] Parsing of Qt::ISODate by
QDate::fromString() and QDateTime()::fromString() is not as lenient as
before, the date component separators are now required to be
non-numeric. This means a string like "2000901901" will no longer be
recognized as a valid ISO Date, but "2000/01/01" will still be even
though it doesn't strictly meet the ISO format of "2000-01-01".
Task-number: QTBUG-34141
Change-Id: I0e481144136c60d4cac61364b9b2c1d2cd1e78fc
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Explicitly check that the QTimeZone is valid before trying to use it.
Change-Id: Iec415a2cb07071502fe71ee5ac92a7657e818f99
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It's usually a bad idea to surround slots or signals with #ifdef since
moc may not parse it the same way the compiler does.
Change-Id: I6a3623ed7cb9fbc1b966df9d60f71b7fdf91acfe
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
There was a comment about the test failing for max() because of an
overflow. That happens if you're at UTC or ahead of it (to the East of
the Prime Meridian), which is how this test usually gets run (UTC,
Europe/Oslo, Europe/Helsinki, Pacific/Auckland). But if you're behind
UTC (to the West of the Prime Meridian), then the overflow happens for
min().
Change-Id: Iebba49d1303e9f18f5038f5cf23c77bf83e5fd4b
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Windows 7 and later have LOCALE_SSHORTTIME, which is what we need.
Task-number: QTBUG-33718
Change-Id: I4c3f113d17102a37fb752de56f06b312f27c7887
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
To be able to call SetNamedPipeHandleState on stdin in a child
process, we must create a read-end pipe handle with the
FILE_WRITE_ATTRIBUTES flag set.
This can't be done with CreateNamedPipe but only with CreateFile.
Therefore we're creating the handles for the child process always
with CreateFile now. Besides, it's conceptually cleaner to have the
server handle of the named pipe in the calling process.
[ChangeLog][QtCore][Windows] Fix regression from Qt4 in QProcess.
It wasn't possible anymore to alter pipe modes of stdin in child
processes.
Task-number: QTBUG-35357
Change-Id: I85f09753d0c924bdc8a6cef1ea5dbe6b2299c604
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
QHash::key() is O(n) and we're calling it n times. That can make repeated
calls to the meta object builder very slow, as for example QQmlPropertyMap
when inserting properties repeatedly.
Fortunately this is easy to fix, as the value in the hash map is also the
index, so we can simply iterate over the hash once. With the exception of
the class name, which we have to treat specially to ensure that it is always
the first entry in the string table.
Task-number: QTBUG-32720
Change-Id: Ic954c45c454107feee83216131f601cc69d4c63b
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
This will break network paths passed as command line arguments.
Introduced by 4ff6951550 .
Task-number: QTBUG-35432
Task-number: QTBUG-30628
Change-Id: Ice9ce15275ef69e9e9e82daf5a303e7c56294368
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Make
QCommandLineParser::add{Help,Version}Option()
QCommandLineOption::setDefaultValue()
QCommandLineOptionPrivate::setNames()
have transaction semantics: either they succeed, or they change nothing.
It's trivial to provide this guarantee, so do it.
Add a test for the surprising property that setDefaultValue("") resets
defaultValues() to an empty QStringList instead of one that contains
the empty string.
Change-Id: I61623019de3c7d2e52c24f42cc2e23ec5fddc4da
Reviewed-by: David Faure <david.faure@kdab.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Fix isValidId() which was failing valid IDs because it was splitting
name parts by \ instead of /. it was also rejecting offset from UTC
formats names. Add unit tests.
Task-number: QTBUG-35025
Change-Id: I4d23d2e54f4a9fac9afcc4eff0a02d6f4af21385
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
As the last line in the QObject destructor, we call setParentHelper(0) to
remove ourselves from the parent. In the process of that we also initiate the
QML parentChanged callback. The first thing that parentChanged callback used to
do (but now does it too late, after 26350b5ceafa0ade1328037f6234a7d288eb8f48 in
qtdeclarative) is to check if the object was deleted and then return. We could
re-introduce the check there, but I think it's cleaner to not bother calling
the callback on a dead object in the first place.
Change-Id: Ia4d43b65a9b3744a451b4c312a2d6f9c0e3b67dc
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
* Add a check for the case sensitivity of the file system, so that
unit tests can determine whether names of preference files with
different case should generate an error or not.
* Add check for OS X native file format in rainersSyncBugOnMac().
Task-number: QTBUG-32655
Done-with: Liang Qi <liang.qi@digia.com>
Change-Id: I76821653dd4ebc00e20abdbb5b79c1a80290dece
Reviewed-by: Liang Qi <liang.qi@digia.com>
Tweak a handful of tests which didn't compile on this platform.
Change-Id: I208d9eb289dfb226746c6d0163c3ea752485033b
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
The private method dateForLocalTime() was not checking that transitions
were valid, resulting in infinite looping when a time zone didn't have
any future transitions.
Change-Id: I0e5d07063861778dd86056a80c36fdd9f9d36133
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Fix parsing of TZ file abbreviations, to correctly return cases where
POSIX rule doesn't have separate DST rules, and where abbreviation
is a sub-string of another abbreviation, otherwise any toString() call
will crash.
Add test to exercise all available time zones, especially useful for TZ
file to confirm all file format variations dealt with. Fix parsing of
Version 3 of TZ file, and ICU display name, to allow all files generated
from release 2013f to pass, otherwise isValid() call will crash.
Task-number: QTBUG-34061
Change-Id: Ie0b6abc218adff1c8967eb33fdb0762041d2305f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The name Olson was misspelled as Olsen in the public api of QTimeZone
which is needed to be fixed before first public release in 5.2 would
freeze the api and prevent it being fixed. It has been decided that
renaming as IANA ID would be more future-proof.
Fixes to the private code will be done separately to keep this patch
against release branch to the minimum required.
Task-number: QTBUG-34735
Change-Id: I8ee90644862c907f6d1937b8536f0c02583ae736
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
GCC refuses to use a merely static const uint array in a constexpr function.
Fix by making the array constexpr if supported by the compiler.
Change-Id: Idd59d3f74f8f4e98aad82bc892f4a6469932df9f
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
This necessitates adding d==0 checks in QCollator. By documenting that
moved-from instances can only be assigned to or destroyed, we can
limit the functions in which to check for d==0 to the assignment
operator and the destructor.
Doing otherwise would destroy all advantages of move semantics by
introducing a heap allocation to re-populate other.d.
Add a test for this (QCollator didn't have any before).
Change-Id: Ic6ff202072822bebfd5e48259c3d0fa345a63118
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Rather than breaking at column 79 precisely, break entire words,
to improve readability.
Change-Id: Ie30db00f0e6ed95cce87480c3b91804826c6076b
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Previously matched Qt.platform.os, however that can only provide one
string. Multiple selectors can be present at once, so we can provide
both unix and linux instead of having to pick the most specialized one.
Task-number: QTBUG-34796
Change-Id: I219517d740fa7385e923a9e09cb7e241378fbaee
Reviewed-by: David Faure <david.faure@kdab.com>
Up to now, the feature classe Uniform Initialization was subsumed by the
Q_COMPILER_INITIALIZER_LISTS flag together with support for
std::initializer_list.
This caused at least two problems:
1. On QNX, the standard libray does not ship <initializer_list>, even
though the compiler (a GCC 4.6, IIRC) supports it. But since there
was only one Q_COMPILER flag for both, support for the compiler-only
part of the feature had to be disabled, too.
2. MSVC 2013 supports initializer lists, but has a bug that renders full
uniform initialization support, as required for QUuid, useless.
By splitting the feature into two, we can separate them better, and do
so in QUuid, which is the only class that currently takes advantage of
uniform initialization (to provide constexpr constructors).
Since Q_COMPILER_INITIALIZER_LISTS worked as a flag for uniform
initialization so far, with the two known exceptions above,
UNIFORM_INIT is defined whenever INITIALIZER_LIST is, except that
I don't revert UNIFORM_INIT on QNX as I do for INITIALIZER_LISTS
and that I expect the MSVC 2013 features to set INITIALIZER_LIST,
but not UNIFORM_INIT.
Task-number: QTBUG-34705
Change-Id: I81916e950a0f3aab3de7977e0326d2de3d31b14c
Reviewed-by: Yuchen Deng <loaden@gmail.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Add few cases where conversion to or from LongLong was missing
We need to make it work if we want to use variant.canConvert<qint64>
[ChangeLog][QtCore][QVariant] Fixed QVariant::canConvert with longlong
Change-Id: I0f65073802b62d99250601dd90a8cd2e4d934b60
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
The tracing API still misses some real-world exposure. Let's
re-do this in dev to have more time.
This reverts parts of following commits:
466e0dff4b7a47aebe9ea652bab6a78f0654ceb84162522edd32f27b43679ff81bdc1a
Change-Id: If97340c37b8b3363f597683336a8390d5ff386f1
Reviewed-by: hjk <hjk121@nokiamail.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
The problem is the verifyConstExpr<>() line involving the ~ operator.
The result as an integer is a value that can no longer be represented
in an int. This is known at compile time and thus template deduction,
which only has an int to match against, fails.
To fix, use an unsigned int as the first template argument of
verifyConstExpr<>().
Clang's error message for this is really sub-optimal, cf.
http://llvm.org/bugs/show_bug.cgi?id=17834
Change-Id: I3a77dc54d2bee12b016d75724ac1bd7801f4cf2d
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Make sure that the size of QIntegerForSize<N>::{Signed,Unsigned} is
actually N.
Change-Id: I221304f7c420e80758ef7b115bafb7cf1f8c8829
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
This is what ConfigLocation was meant to be. A directory shared by all
applications. Unfortunately when I wrote the fallback on Windows,
I picked DataLocation (which is app-specific) instead of
GenericDataLocation (which is shared between apps). This makes it
impossible to have config files shared between apps, e.g. for libraries.
It also makes ConfigLocation quite inconsistent (on Windows one cannot
use it to load another app's config file, while it works everywhere else).
All this is fixed by GenericConfigLocation, which is shared between apps.
Change-Id: I23a755131061d4fea01e13dd1038fbd8ef333a5d
Reviewed-by: Alex Richardson <arichardson.kde@googlemail.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
