since the effect sets a X11 window property the only thing it has to care
about is if the widget pointer is valid and the elapsed time. moving,
resizing the widget, etc. do not matter
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
relies solely on the compositor for it and assumes window opacity is not
set by the user, Katana has animation override
(_KDE_NET_WM_SKIP_CLOSE_ANIMATION X11 atom) and the Katie fade effect
should probably have such (or just check if the opacity value is not 1.0
before running the effect) because effects may overlap with window manager
/compositor effects resulting in all sort of bad things
for that reason I am considering dropping the effects that alter the way
windows/widgets show, hide, are moved, etc. altogether and leaving that
to the window manager/compositor. effects such as colorization of pixmap
and grayscaling (used in QIcon) however are unlikely to be removed
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
since it is known what locale the X11 input method uses it is redundant to
verify that via XLocaleOfIM() (unless X11 fails to detect what locale is
set for some strange reason).
also not using the cached codec returned by QTextCodec::codecForLocale()
because the locale may change at any time (either environment variable
change or just plain setlocale() call as was done by KCatalog as gettext
workaround) and QKeyMapper has to reflect such changes
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
the desktop style is now determined by QT_PLATFORM_PLUGIN and
DESKTOP_SESSION environment variables with fallback to cleanlooks style so
even if it is the fallback style there is something to show but if it is
not the fallback the user will be able to preview the style chosen by
QGuiPlatformPlugin
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
now QSettings will behave much like opening a file via open() and the like,
see the QFile hijack test case
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
it being static means its destruction will be delayed until exit and it
holds shared data references the meta types of which may be unregistered by
the time it is destroyed
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
since fontconfig has its own cache mainting internal QFontDatabase cache
is kinda redundant, have to leak check still but font tests pass
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
the cake is a lie - even tho XGetPixel() and XPutPixel() take ulong as
pixel value it has to converted manually to value appropriate for the
XImage depth
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
for some reason on Debian XImage with depth equal to 16 can have different
bytes per line (off by 2 bits), on Arch it was not a thing when I tested
it so it could be a bug in X
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
since it is painting of image (not painting to image) the performance
impact of it not being QImage::Format_ARGB32_Premultiplied is neglectable
Signed-off-by: Ivailo Monev <xakepa10@gmail.com>
