This is good for performance in terms of avoiding unnecessary
database activity and keeping the cache smaller.
Detail:
This change was not included in the big refactoring of QSqlTM. The
idea was that the model shouldn't second guess the intention of the
application and maybe the application wants to cause a submit.
It was a marginal consideration.
Now I think it's clear that our interest in not unnecessarily
expanding the cache outweighs that. In addition, applications can now
call selectRow() if they worry that the database values for the row
have changed and want to set a value back again.
Test added.
Change-Id: I63814dcb63a96c6ba1c8cc227807725a954a0b68
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: David Faure <david.faure@kdab.com>
QMacStylePrivate needed to be revised in order for it to have access to
QCommonStylePrivate that is planned to manage style animations in a
generic fashion in the near future.
Change-Id: I03185ae210d2cfc253a236d6b2da35d0201f7e13
Reviewed-by: Jens Bache-Wiig <jens.bache-wiig@digia.com>
The STL-style iteration over the cache in submitAll() assumed the
iterator would remain valid until reaching cache.end(). This failed
to consider that virtual selectRow() might be overridden so that
it removes rows from the cache. For example, it might call select()
which would empty the cache.
The new approach checks at each iteration whether the row is
still in the cache. Using foreach here is justified by its fitness
for purpose and readability.
New test included.
Change-Id: Idee8807ede239c3ba56ff1604574c49f47385ad2
Reviewed-by: David Faure (fixes for KDE) <faure@kde.org>
Change-Id: Id0137400f18c8dfe7be7ca44670c16615401d424
Reviewed-by: Sean Harmer <sean.harmer@kdab.com>
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
Reviewed-by: Peter Hartmann <phartmann@rim.com>
QWidget::underMouse() did not report correct widget in cases where
mouse was grabbed by popup, which was especially disruptive in case
of QCompleter popup, as that wouldn't close anymore with off-popup
clicks.
Root problem was that mouse capture in Windows caused enter/leave
events for QWindows to be generated incorrectly.
QPlatformWindow documentation specifies that enter/leave events
should be sent independent of explicit mouse grabs and only automatic
mouse grabbing done when button is pressed should suppress
enter/leave events. Updated Windows mouse handling to conform to
this.
Task-number: QTBUG-27283
Change-Id: Iecf786a702f7d29e6026c42ff8ec4c9cbf1b6ac3
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
mOpenPanel is set to 0 if save file dialog is used so call
mSavePanel instead.
Fixes regression introduced in
2e902799df
Change-Id: Idb93e44437731cad3c985d51b57b793a3ff2f292
Reviewed-by: Pasi Matilainen <pasi.matilainen@digia.com>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
This test actually passes on Mac OS X, so removing the QSKIP
Task-number: QTBUG-24374
Change-Id: I0b761ca9c30afc9d511e9962ee1c0958b863b374
Reviewed-by: J-P Nurmi <jpnurmi@digia.com>
The intention probably was to use __FUNCTION__, but since
Qt logging can now be configured to print the function name
by setting environment variables, just remove it.
Change-Id: Ic87aebfea633dd91eda7376b33ca3b14d1cab93f
Reviewed-by: Miikka Heikkinen <miikka.heikkinen@digia.com>
The solution is similar to that
in b84e180263 which affected
QSortFilterProxyModel.
Task-number: QTBUG-25370
Change-Id: I6bbb9d9786bcb2c9fa8027ab8a7cc13664784b8d
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
When rendering transient scrollbars, we instanciate a specialised
NSGraphicsContext since those need non-opaque rendering. This context
is being used as current context, but it was never properly released
and the previous context never properly restored.
Task-number: QTBUG-27327
Change-Id: I4e867e64a09368dbe64b84b17f07fd8e57bbae24
Reviewed-by: Teemu Katajisto <teemu.katajisto@digia.com>
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
We need to handle CMAKE_INCLUDE_CURRENT_DIR for include directories.
Otherwise generated files located in the current binary directory are
not found as expected.
e.g. *.json file as meta data for Qt5's plugins generated at build time.
Change-Id: I14ae1e7013f9d8b485aa990d50db4a03ca4f4b81
Reviewed-by: Yuchen Deng <loaden@gmail.com>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
On Cocoa if the modality of a dialog changes then the style of the
window needs to change to reflect this. So we add a variable to
cache the windows modality when it is created to compare against when
being made visible.
Task-number: QTBUG-22316
Change-Id: I7bfd016321510a9ec70ccb90672f5203a0f3a468
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
qUpperBound was replaced by std::upper_bound.
Unfortunately the STL of MSVC 2008 enforces
the definition of the operator in both
directions.
Change-Id: I3e0f775c23e43332d106e0847d3611e488da6c06
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Enable eglfs build against any android > 4 tree,
linking to native libs without emulation layers,
running on top of surfaceflinger.
No GNUs where harmed in the process.
Yes, any android.
Tested on maguro, tf300, eeepc-x86
x86-64 compiles but broken elsewhere.
You don't need an Android.mk, but you must compile from within
a shell setup with androids "lunch" or an equivalent that set TOP and OUT.
Since we do callbacks to androids build system,
the same env restrictions apply (must use gnu bash, and gnumake 3.81)
Done-with: Samuel Roedal <samuel.rodal@digia.com>
Done-with: Robin Burchell <robin+qt@viroteck.net>
Done-with: Brian Avery <brian.avery@nokia.com>
Change-Id: Iec0178cdeadbeefc79e4fe6ef449d399ac8ca666
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
qwidget_qpa.cpp was setting cursor directly to platform window, so
QWindow didn't actually know what cursor was supposed to be shown.
Since QWindow now supports setting cursor and has override cursor
handling there, set cursor via QWindow instead of platform window.
Also changed QGuiApplication override cursor handling so that
it will query the cursor from QWindow if there is no override cursor.
With these changes, widget's custom cursor will be properly shown
when final override cursor is restored.
Task-number: QTBUG-24674
Change-Id: I2d2bb1027779256f9d6de560b6533f45e205ffe9
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
The library name changed back from quick1 to declarative, so the
debug option should be renamed accordingly.
Change-Id: I192520a0a4de4764cf93da8cdbb14e24ba7de093
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
On linux with the new plugin debug output was
a bit too generous. Instead use environment var:
If QT_DEBUG_ACCESSIBILITY is set, give the full a11y output.
Change-Id: I1c9e2078e38799f91468cfc2d9ddcaa36ba4fa03
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
The mouse events were being stolen by the modal dialog, preventing the
combo box to update itself. Key events were being properly forwarded,
but this was probably due to the different way key events are being
delivered.
Change-Id: I5bfc4800fad9fd4122a212e6fdec13811a7e647b
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
The function wglCreateContextAttribsARB will fail if we request a
context version higher than is supported. We therefore upper-bound the
requested version by the version of the static context. This results
in context creation succeeding and having the closest possible match
to the requested format.
The xcb qpa plugin is modified to operate similarly to the windows
plugin in that it now creates a "static" context which is used to
limit the versions of contexts requested by the user.
Change-Id: I277ad7cc82edfdf7b9d8502ad921c8175feb1a4a
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
Mac OS X 10.7 comes with the family of Stix fonts, some
of which exposed an ugly layout bug in the QFontComboBox
because the ascent/descent ratio is very large due to
a very high ascent, so centering the text vertically might
cause most of the text to be clipped away.
The solution is to detect when the ascent is larger than the
height of the destination rectangle (hence a large part of the
characters will be clipped) and use the actual bounding rect
for centralizing instead. Since this only happens for a
very few of the fonts, the overhead of getting the bounding
rect should be tolerable.
This is a port of 4679c6901fc7c388fdf6c022d3499708222ef1f1 from
Qt 4.8.
Task-number: QTBUG-26691
Change-Id: Ia2014775e5baf0568df3290f2dc4ad64fb5a74bd
Reviewed-by: Jens Bache-Wiig <jens.bache-wiig@digia.com>
for the case when the boundary finder is assigned to an invalid one.
Change-Id: I5b60984ff3fd99972fcae21895684bd83b012780
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
This event can be used by any platform plugin to implement
special application panels/overlayed menus.
Currently used by QNX only.
This replaces sending fake Qt::Key_Menu presses in the QNX
plugin. Qt::Key_Menu is already used when invoking context
menus with the keyboard.
Change-Id: I9c8f1743fd147a07c11883323800017376915ae1
Reviewed-by: Kevin Krammer <kevin.krammer@kdab.com>
Reviewed-by: Sean Harmer <sean.harmer@kdab.com>
Reviewed-by: Nicolas Arnaud-Cormos <nicolas@kdab.com>
Reviewed-by: Paul Olav Tvete <paul.tvete@digia.com>
On Mac usually release frameworks are used during development so asserts
don't tend to get fired. Since the menu code did not crash in Qt 4.8,
it was decided that we should ensure that the same code does not crash
now so instead we warn instead of asserting for those cases.
Change-Id: I923e5591202cb2ad801074ce3ad2a11d7190cee8
Reviewed-by: James Turner <james.turner@kdab.com>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Keep the original QString that triggered the parsing error, instead of
just one QChar. This provides more powerful error messages, like:
Invalid IPv6 address; source was "http://[:::]"; scheme = "http", host = ""
(QUrl cannot keep invalid hostnames)
Invalid port or port number out of range; source was "http://example.com:abc"; scheme = "http", host = "example.com"
(QUrl cannot keep a non-numeric port number)
Invalid path (character '%' not permitted); source was "foo:/path%?"; scheme = "foo", path = "/path%25%1F"
(the tolerant parser runs first, so the faulty component is fixed)
This stores the error state in a special structure which is not
allocated under normal conditions, keeping the memory consumption
down. On 32-bit systems, QUrlPrivate does not increase in size; on
64-bit systems, it grows by 8 bytes.
Change-Id: I93d798d43401dfeb9fca7b6eed7ea758da10136b
Reviewed-by: David Faure <faure@kde.org>
Make both invalid hostname messages start with "Invalid hostname". And
split the empty port error from the invalid port one.
Change-Id: I870d1ed6fb07ec494f553871a37ed167141ffc06
Reviewed-by: David Faure <faure@kde.org>
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
Just so the code generation is a little better.
Change-Id: I2a43a4df0ae67900c465a6c2b4f2b8ba284dbbaa
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
Reviewed-by: David Faure <faure@kde.org>
They're never accessed outside of qurl.cpp anyway, so let the compiler
know that it doesn't need to generate a full out-of-line copy for them
in case it does inlining.
Change-Id: I2be069b3fd2658eff9ad3023c21c8ae653c389ab
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
Reviewed-by: David Faure <faure@kde.org>
It used to return bool because setHost returns bool and, therefore,
setAuthority could fail. However, the return value is never checked,
in either parse() or QUrl::setAuthority(), because there's no error
recovery.
This is a small optimisation.
Change-Id: I25660d66cfad64ca5b9706cc38afa0e97ba3ee0b
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
Reviewed-by: David Faure <faure@kde.org>
That's what we have QUrl::errorString() for. This will become evident
especially now that QUrl::toString() / toEncoded() return empty if
there are errors.
Change-Id: I64a84e9c6ee57c0fc38cc0c58f5286ddc1248d1f
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
Reviewed-by: David Faure <faure@kde.org>
As the comment says, an empty scheme is not permitted. However, if
that error were to happen, QUrl falls back to parsing the URI as an
"URI reference", starting with the path.
E.g., ":/foo" is a path of ":/foo", which will in turn trigger the
compound "colon before slash" error.
Also, we don't percent-decode in the scheme.
Change-Id: I438a61e17323c7722ddcc64792577a9ecb869c4b
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
Reviewed-by: David Faure <faure@kde.org>
If an URL is invalid, let's indicate that in the test output. To be
helpful, let's make QUrl::errorString() include the component form of
the URL.
Change-Id: Iaafe16973ded79c7ea688fbb23808d91253e8c14
Reviewed-by: David Faure <faure@kde.org>
These two errors can only happen if one calls setPath() explicitly. They
cannot happen for parsed URLs, which is why they are only caught with
isValid(). It's not possible to set the error condition in setPath()
either because they depend on the presence / absence of the authority
and scheme.
Also update all the unit tests that set a path not starting with a slash
and were just "freeloaders" on the previous behaviour.
Change-Id: Ice58cd4589a850452d7573a5b19667bbab2fb43e
Reviewed-by: David Faure <faure@kde.org>
Position the QIODevice correctly for DIB (==BMP without file
header).
Task-number: QTBUG-26690
Change-Id: I7922b4e830a92dcc19f1b52e305ca9ff3fd5d9f5
Reviewed-by: Thomas Hartmann <Thomas.Hartmann@digia.com>
Since we no longer do the sanity checking against
gtk-qt-theme at startup, there
is no longer any need to inherit from cleanlooks.
Cleanlooks should no longer be needed as it has been
completely replaced by Gtk style and this dependency
is not required at this point.
Change-Id: I05b59278cd3ff915df49a7ad2d04494258db0d42
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
Like Qt 4. Use the presence of the "-psn_" command-
line argument to detect Finder launches.
Link QtGui against the Cocoa framework due to CFBundle
usage.
Change-Id: I947ae53462072fbdb23afcc0a13ce6b26d02c191
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
