The protocol may be namespaced, in which case the string lookup would
fail, so we iterate the protocols of QNSWindow instead (of which there
is only one, QNSWindowProtocol).
Change-Id: Ic45752c9e3a40f5d42ec82c4287402a3d7a47b09
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@qt.io>
QMetaCallEvent references up, so we need to drop by one. In all other
cases, we need to drop the reference to zero and delete.
[ChangeLog][QtCore][QMetaObject] Fixed a memory leak that happened when
the new-style call to invokeMethod() was used.
Task-number: QTBUG-65462
Change-Id: I39332e0a867442d58082fffd15034b99e31c92a1
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Remove usage of outdated header.LGPL21 and replace those with proper
one (header.LGPL in src, header.GPL-EXCEPT in tests)
Change-Id: Ia4d1c0d84b77f09787fe7c30670747a1fe2aff29
Reviewed-by: Liang Qi <liang.qi@qt.io>
Original documentation commit included unresolved link target names.
This fix replaces them with the actual OpenGL function names.
Change-Id: I36a24eb237ef35d7207f3bae0771dc96476d7b19
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
The referenced static meta object for the superclass might be in a
different DLL. In this case, the whole QMetaObject can't be initialized
all via preinitialized data in the data section of the binary, but must
run code at runtime to fill in the value of the dllimported pointer.
In these cases, both GCC and MSVC initialize as much as possible statically,
while only filling in the dllimported values (QMetaObject::d::superdata)
at runtime. Clang, on the other side, initializes the whole struct
at runtime if some part of it needs runtime initialization, leaving
the struct completely uninitialized before constructors are run.
In C++, there are no guarantees for in what order constructors in
different translation units are executed. This in particular means
that there are no guarantees as to whether qRegisterWidgetsVariant()
in qwidgetsvariants.cpp runs before or after the runtime initialization
of QWidget::staticMetaObject.
With GCC and MSVC, this doesn't seem to have mattered since only the
superdata pointer of the staticMetaObject was uninitialized - everything
else was initialized, and the superdata pointer doesn't seem to be
accessed during qRegisterWidgetsVariant.
With clang, the whole staticMetaObject is uninitialized, unless the
staticMetaObject has been initialized before (and the initialization
order is undefined).
By setting a manual priority (which is a GCC extension that also
clang supports) for the staticMetaObjects, we can be sure that
these are initialized before the actual explicit constructor
invocations (without any explicit initialization priority) that
can access the staticMetaObjects.
Change-Id: I64a82f12d690528567509791bae088b6304e189b
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
As the HANA ODBC driver does not provide scrollable
results per default, the connection has to be established
with a specific option.
Change-Id: I14751a94f61d6fffd6d56c39076c24ec4d1f88ef
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Joachim Puhr <jpuhr@gmx.de>
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
The code used to fall back to anonymous login independently for username
and password; however, it should only use a fall-back password if the
username is missing or (case-insensitive) "anonymous". When a
non-anonymous username is given without password, we should simply skip
he PASS message to FTP.
If the FTP server requests a password, in the latter case, QFtp will
signal authenticationRequired; in all cases, if the server rejects the
given credentials, QFtp signals authenticationFailed. Either way, the
client code can then query the user for credentials as usual.
Task-number: QTBUG-25033
Change-Id: I2a4a3b2725819ab19c8a7e4baa431af539edcd8d
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
setData()'s default implementation for QListWidget, QTableWidget, and
QTreeWidget treats Qt::EditRole and Qt::DisplayRole as referring to the
same data. Used the same sentence from QStandardItem::setData for
consistency.
Task-number: QTBUG-11549
Change-Id: I41d06bdaaa8e7d4a86e24147b3d8222ad7823a6c
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
qt5_add_binary_resources() macro did not recompile for CMake generated
input before this patch.
Adding the input files to the DEPENDS option corrects this issue:
Task-number: QTBUG-60714
Change-Id: I0f46918c6f1079fed7ee1b21305b18ff38f863f8
Reviewed-by: David Faure <david.faure@kdab.com>
When applying fallback fonts to characters that are joined by
ZWJ or ZWNJ, we also have to set the same font for the control
characters, otherwise we will split the text and the necessary
shaping will not take place. This was reported for emojis, but
will probably also happen for Indic scripts where joiners are
used predominately.
[ChangeLog][QtGui][Text] Fixed ZWJ and ZWNJ control characters
when fallback fonts are in use.
Task-number: QTBUG-65519
Change-Id: Ia37233f3319b95af68ae6053c29997eac65448e0
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
After we stopped sanitizing the fallback font list (with change
6ca48a847a), we now need to make
sure it is ordered so that the fonts that support the writing
system in question are always tested first, otherwise we can end up
loading a lot of fonts that will never be used.
Task-number: QTBUG-65605
Change-Id: Id2a65bbff3e64e6d6e6b4f72500778ee3e811e84
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Introduce a special hash modeled on the one used for QFileSystemWatcher
on Windows.
Task-number: QTBUG-31103
Task-number: QTBUG-64147
Change-Id: I69ebabe841716e4957ae3fb04fa5c43d233a3552
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
With libinput we now get a hardcoded resolution that is unrelated to
the hardware. So avoid using that as a real pixel delta and document
pixel deltas as being driver specific and unreliable on X11.
Task-number: QTBUG-59261
Change-Id: I9fe86d80e7ccd290ed2e4091d7eafa52cb537d34
Reviewed-by: David Edmundson <davidedmundson@kde.org>
Reviewed-by: Marco Martin <mart@kde.org>
Reviewed-by: Gatis Paeglis <gatis.paeglis@qt.io>
When we request fallback fonts, we cannot discriminate the fonts
based on the writing system support. This is especially important
since common script is now merged with other scripts, meaning that
a common script character will always go through the fallback
mechanism when not supported by the main font. When drawing
for instance a string of Devanagari characters on macOS, we would
get a list of 33 fallback fonts, but almost all of them would be
the default Devanagari font, since none of the other fallbacks
would support that script. Meaning that we would just check the
same font over and over, which makes no sense. The fallback list
has been retrieved specifically for the given script, so we do
not need to consider that when fetching the fonts.
For most of the common set, we will not have noticed the bug,
because at least one of the writing system-specific fallbacks will
have had support for latin characters as well. But when trying to
mix emojis and some non-common script, we would get a box in
place of the emoji, which had been adopted to the main script and
would only be looked for in the fonts supporting this.
Note that this exposed an issue with the QRawFont test on some
systems. When the sample text contained a space, it would
be possible to get a fallback font for this character, since
we now effectively support fallbacks. This is not the correct
behavior, but it is unrelated to this fix, and it was not what
the QRawFont::unsupportedWritingSystem() test was written to
check. I have therefore removed the space from the sample text
to make the test pass, and will make a separate task of fixing
the issue of merging fonts for whitespace characters.
[ChangeLog][QtGui][Text] Fixed a bug where mixing different
writing systems with emojis could lead to missing glyphs.
Task-number: QTBUG-61882
Change-Id: I00f6043bb01af1f2277723ccf643034aebf3e18f
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
sqlite will reuse the index for bound parameters when the named
placeholder is duplicated so we only need to call bind one time for each
placeholder. Therefore we need to have just one instance of each value
when doing the bind.
Task-number: QTBUG-65150
Change-Id: I75c4bcc4563e43c180a59a7a4cbb770dbe994642
Reviewed-by: Jesus Fernandez <Jesus.Fernandez@qt.io>
While adding the possibility to access values for QSqlRecord with
decorated field names (table.field), some string-allocations were added
which created a remarkable slowdown. Replace the QString allocations
with QStringRef avoids those allocations and restores the speed for
normal operations (apart from on QString::indexOf() call and some
integer comparisons)
Task-number: QTBUG-65226
Change-Id: I9e458523891421abce9e4a7ed931fec000dcbe76
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
This reverts commit d7bcdc3a44.
Now that QPalette::setBrush() is a bit smarter (see
c564779c07), we can be
confident the palette's cache key won't change so often.
Therefore, we can use it again as part of every pixmap's
cache key.
Change-Id: I5cab2296a937f9149c7745291e0442eea24f4e9d
Task-number: QTBUG-65475
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
These options are used by winrtrunner/Qt Creator to control logging
and debugging behavior and should not influence applications.
Without this change applications that use
QCommandLineParser::process have to explicitly remove these values
from the arguments list or the application will exit with an
"unknown option" error when run from Qt Creator.
As winrt applications do not natively support command line arguments,
breaking the use case of having these values as valid command line
options is acceptable.
[ChangeLog][qtmain][winrt] -qdevel and -qdebug are removed from the
command line arguments and not passed to the application.
Change-Id: I60e5565148fd3dd0155b946607fdf5042139d2e1
Reviewed-by: David Schulz <david.schulz@qt.io>
Reviewed-by: Maurice Kalinowski <maurice.kalinowski@qt.io>
Clang errors out on casting from a pointer to a smaller integer
type.
This is an adapted backport of a newer upstream harfbuzz
commit, 03b7a221f701a9b003890878e730ad175c3fdd86.
Change-Id: I0de62d90008021aed6acf4e946ce1d346f9aac8c
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Setting the same brush on the same group and role should
not detach nor alter the result of QPalette::isCopyOf().
Task-number: QTBUG-56743
Change-Id: Ic2d0dd757d703b01e8c5d835a8c124b3317653f4
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
MinGW's headers and libraries are lowercase, and MSVC is generally
run on a case-insensitive file system. Referring to libraries with
a lowercase name is more compatible.
This fixes cross-building using MinGW, and matches a similar
change for headers in dfaa32a945.
Change-Id: I728376f239a23c700416643cae98e271bbcdff32
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
This fixes a regression from ba44cdae38.
Task-number: QTBUG-65516
Change-Id: I989b075d3f96170660042349437e4a6a4a115cf9
Reviewed-by: Liang Qi <liang.qi@qt.io>
So invert the list and let "msvc-2017" be the latest, at least until we
know about a different version.
Change-Id: I39332e0a867442d58082fffd1504c316e27f51d0
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
QTreeViewPrivate::itemAtCoordinate() did not calculate the correct item when
non-uniformRowHeights is enabled and vertical scroll mode is ScrollPerPixel.
This results e.g. in an activation of the item above when the click happens
on the very first pixel line of an item. Another, more problematic effect was
that once a drop happened on the very first line of an item,
QAbstractItemViewPrivate::dropOn() calculated the root as the drop index
because the visualRect of the calculated item is compared with the drop
position which did not match.
Task-number: QTBUG-57538
Change-Id: If89571bf0fb76fc01d60ba7139f1d985b49d70db
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
Reviewed-by: Thorbjørn Lund Martsum <tmartsum@gmail.com>
Add an env variable QT_MAC_TABLET_IGNORE_BUTTON_MAPPING which, if set,
reverts to the old behavior.
The issues that it fixes are:
1) With a Genius tablet, it's not possible to generate a middle-click
at all, regardless of preferences, unless you set this variable
2) With a Wacom mouse (passive mouse used with an Intuos 3), the
side-buttons (normally mapped to forward and back buttons) do not
work unless you set this variable
[ChangeLog][macOS][Tablet support] If you have changed
the button mapping of your Wacom tablet in System Preferences,
it will be respected in Qt applications. To revert to the
old behavior, set the environment variable
QT_MAC_TABLET_IGNORE_BUTTON_MAPPING
Task-number: QTBUG-57487
Task-number: QTBUG-54160
Change-Id: I0a1c9c6550e4dfe6ee2daa2d993f22691430826c
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
Unconditionally apply window cursors restoreOverrideCursor() also for
platforms with override cursor capability in case
a cursor was changed while the override cursor was active.
Amends b05d1c2ebf.
Task-number: QTBUG-65001
Change-Id: I7fd38a27dcf045f2fb48a16abb484342feaab5df
Reviewed-by: Orgad Shaneh <orgads@gmail.com>
If the values vector is empty then we know already that the paramCount
will still be invalid, so we should just accept that and not check the
reused named placeholders.
Task-number: QTBUG-64923
Change-Id: Ifaa755540c4574f1f76d3f9f129bf0f66b837b70
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Binding without an interface and expecting the OS to select something is
not supported in all OSes. On FreeBSD, I keep getting EADDRNOTAVAIL. So
modify our test to only join, leave and send to multicast groups with an
interface selection.
With this, all tests either pass or are skipped for me on Linux,
FreeBSD, and macOS. On Windows, this revealed an inconsistency in
behavior, which this commit adds a workaround for.
Change-Id: Ifb5969bf206e4cd7b14efffd14fb6815456494d2
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Putting parentheses around the call to (std::numeric_limits<T>::min)()
works, but the trick cannot apply to the min() function declaration on
the same line. So we really need to #undef.
I hope no one after the 1990s still needs these macros.
Task-number: QTBUG-65414
Change-Id: I39332e0a867442d58082fffd15024f8edb293311
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
The C standard library functions cannot handle UTF-8 filenames. Instead,
we need to use the wide-character versions which accept UTF-16 input.
Task-number: QTBUG-65492
Change-Id: If4b3b4eeeec4f3bbb428b8f6b0311a65d01463b0
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Like in commit 75cdf654bc, we use the
pointer value of the QMap header (which is of type QMapNodeBase) as a
sentinel. It's never dereferenced. In that commit, the issue was calling
end() directly; in here, it happens when iterating forward from any
element, which means nextNode() can reach back to the header.
However, using static_cast is wrong, so we use reinterpret_cast.
Change-Id: I39332e0a867442d58082fffd15040317704a87ce
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
This removes at least one special-case we had to have, in Android's lack
of /etc/mnttab. Bionic's _PATH_MOUNTED is already /proc/mounts.
Change-Id: I9407dcf22de6407c83b5fffd14fedc638586d0f9
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
This fixes only those that look fixable. There are a number of errors in
the online documentation that look like qdoc tool bugs, like missing
functions (operator==) and documentation text that does not exist in the
source code. This fixes:
- QRandomGenerator(System) constructor showing up
- Links for C++ reference documentation not showing up as links
Change-Id: I9e2892cb6c374e93bcb7fffd14fe21db5a6969d9
Reviewed-by: Martin Smith <martin.smith@qt.io>
Commit 77942a1bdf introduced the QScopedValueRollback, but without
assigning it to a local temporary, so the value was rolled back
immediately, resulting in always sending touch events async.
Change-Id: Ic7f65c3d38c46813ff06694e883dae3df138b9d4
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
Instead of allocating the full backing store size for
the transparency layer, we make sure we only allocate
exactly as much as needed by the scrollbar.
Change-Id: I55c3172fe3dd2a1f3fd46828463497f9f35cb1ae
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
This one fell off the truck while we were removing
HITheme calls. We add it back by simply rendering a
BLACK RIGHT-POINTING TRIANGLE character.
We also fix smaller issues, such as not displaying
any shortcut related to a submenu action — this is
simply not a thing. The spacing between the menu
item's text and the submenu indicator has also been
slightly improved.
Change-Id: I6c768a5506a5eb9528b0dd76acd52b561266d67b
Task-number: QTBUG-64405
Reviewed-by: Jake Petroules <jake.petroules@qt.io>
On Windows, QFileInfo.isWritable() was returning true in situations
where the file would only be writable with elevated privileges. Using
AccessCheck instead of GetEffectiveRightsFromAcl to get the correct
results.
Done-with: Friedemann Kleint <Friedemann.Kleint@qt.io>
Done-with: Edward Welbourne <edward.welbourne@qt.io>
Task-number: QTBUG-30148
Change-Id: I7a3468ac069bf782ca312078e3a84107b6cd468c
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Tor Arne Vestbø
Orgad Shaneh
Thiago Macieira
Jani Heikkinen
Liang Qi
Paul Wicking
Rainer Keller
Martin Storsjö
Joachim Puhr
Ryan Chu
Kavindra Palaraja
Kevin Funk
Eskil Abrahamsen Blomfeldt
Friedemann Kleint
Allan Sandfeld Jensen
Andy Shaw
Christian Ehrlicher
Gabriel de Dietrich
Oswald Buddenhagen
Mårten Nordheim
Oliver Wolff
Jake Petroules
Shawn Rutledge
Simon Hausmann
Lars Knoll
Dyami Caliri