Use the same logic for creating the pages of this book control as for
all the other ones and keep only wxTreebook-specific "depth" and
"expanded" nodes handling in this class.
This struct is finally going to have to be declared in the header as it
will be needed by wxTreebookXmlHandler when it's modified to derive from
wxBookCtrlXmlHandlerBase too.
Also add GetImageId() function to the struct while moving it.
Add new wxBookCtrlXmlHandlerBase and inherit the XRC handlers of all the
wxBookCtrlBase-derived classes except for wxTreebookXmlHandler, which
will require special handling, from it to avoid duplicating the same
code in all of them.
This commit is best viewed with --color-moved git option.
This class is only implemented when wxUSE_BOOKCTRL==1 as its code uses
it, so it shouldn't be declared when wxUSE_BOOKCTRL==0 neither.
This fixes wxiOS build problem after the latest changes.
Convert wxCoTaskMemPtr<wchar_t> to wxString explicitly as doing it
implicitly required 2 user-defined conversions which is not allowed in
standard C++ (but works with MSVC).
Also get rid of a not really necessary temporary wxString variable and
just perform this case when constructing wxStringTokenizer.
Closes#22170.
Skip the event in wxEVT_SIZE handler in wxGrid and wxDataViewCtrl as it
was already done for wxListCtrl in 98c2eef619 (Fix scrollbar position in
wxListCtrl in wxUniv, 2021-09-05).
Closes#22145.
Add a menu item to the dialogs sample for simulating having unsaved
documents in the application. Then demonstrate how to do something
when the user attempts to close the application, e.g. show a dialog,
and possibly cancel closing.
This code was originally written for debugging some issue related to
closing an application, but besides that it serves as an example of a
very common pattern a lot of real world applications need in some form.
Mention the example in samples/dialogs in wxCloseEvent documentation
On recent FreeBSDs, the "Linux-only" <sys/epoll.h> exists in the default
include path, and is thus detected by configure. However, the header belongs
to the epoll-shim emulation library, and it is doubtful whether using it
on FreeBSD is a good idea. Therefore check that the operating system
is actually Linux before enabling wxEpollDispatcher.
Closes#22146.
It seems like this function sometimes doesn't return non-zero exit code
even though it fails to fetch some packages (e.g. due to a network
error), so log its status code to check if this is really the case.
This property expands to either (just) "d" (even in Unicode builds) or
nothing depending on whether we're building in debug or release and can
be used to construct the configuration-independent names for the third
party libraries, most of which use just this suffix and not wxSuffix,
which expands to "ud" in (Unicode) debug build.
This allows to use exactly the same link dependencies for both debug and
release configurations, as demonstrated by the updated minimal MSVS
project file, which is convenient when creating new projects not using
wxwidgets.props, as it allows to enter the dependencies just once and
use them for all build configurations instead of having to enter them
separately for the debug and the release ones.
Make the results look somewhat better by using wxIMAGE_QUALITY_NEAREST
which preserves sharp horizontal and vertical edges in the images which
are common in the icons, with which this function is often used, instead
of blurring them as wxIMAGE_QUALITY_HIGH does.
This is also much (factor of ~40) faster, which shouldn't hurt neither.
Check the speed of wxImage::Scale() using this algorithm too.
As expected, it's between normal and high quality when upscaling and
exactly the same as high quality when shrinking (because it is actually
the algorithm used for shrinking for wxIMAGE_QUALITY_HIGH).
Use the string parameter as the name of the file containing the image to
resize and the numeric parameter as the scale factor in percents (so
that fractional factors could also be specified using this integer
parameter).
This is more practical than running it a given number of times, which
may result in very long running times for slow functions, such as
wxImage resizing tests using wxIMAGE_QUALITY_HIGH.
Also show the standard deviation in addition to the average and min/max
values.
With overlay scrollbars disabled, it was not possible to move the scrollbar
slider all the way to the end, and there would be some flickering when trying
to do so. Fixed by always setting the preferred size to the virtual size, and
returning the proper value for GtkScrollable policy. GtkScrolledWindow uses
the child's preferred size as the virtual size, so the preferred size must
be set for the scrollbars to be managed correctly. Additionally, there was
some confusion during the initial GTK3 port over the difference between
GtkScrolledWindow policy and GtkScrollable policy, resulting in using the
minimum size rather than the natural size.
See #22093
This reverts commit de10f054c4 and a bunch
of commits improving it done since then:
6f888df474 (Restore button size calculation when not using manifest in wxMSW, 2021-04-24)
de10f054c4 (Improve calculating wxButton best size under wxMSW, 2021-04-08)
6f888df474 (Restore button size calculation when not using manifest in wxMSW, 2021-04-24)
75d508b6e6 (Fix regression in sizes of buttons with bitmaps in wxMSW, 2021-07-08)
051418ac00 (Fix recent wxMSW buttons with bitmaps appearance regression, 2021-08-26)
fbc6462375 (Fix a gcc warning introduced by the previous commit, 2021-08-26)
The reason for reverting is that using BCM_GETIDEALSIZE doesn't seem to
have any advantages compared to using our old code, as it doesn't take
the margins into account neither and we still have to add them ourselves
and so it's simpler to just always do it without using this message.
They are deprecated by libtiff in favour of C99 standard types, but
replace them with wxUintNN for now, as those are definitely available on
all platforms, while we support some platforms without C99 stdint.h
(notably ancient MSVC versions).
Do it for consistency with the existing GetBitmap() overload and also
because this can be useful when we need to test for the bitmap bundle
presence before using it.
Also document both the new overload and the original one, which was
mistakenly left undocumented when it was added.
This replaces 4 calls to GetBitmap() with a single one.
Note that we now use SetBitmapPressed() and SetBitmapCurrent(), which
take wxBitmapBundle, instead of deprecated SetBitmapSelected() and
SetBitmapFocus(), which do not.
We also search for the parameter node only once instead of doing it
twice in a row if it was found, so the new code is slightly more
efficient.
Update the remaining ports to take wxBitmapBundle rather than wxBitmap
as well to make their API consistent with the tier 1 ports -- even if
there is no actual support for choosing the resolution-appropriate
bitmap in them yet.
src/unix/mediactrl.cpp failed to compile due to a precision-losing cast
(from 'gpointer' {aka 'void*'} to 'window_id_type' {aka 'unsigned int'}).
Tell pkg-config to use 64-bit configs if building for 64-bit on SunOS.
PKG_CONFIG_PATH is set to /usr/lib/64/pkgconfig in that case.
Documented at:
https://docs.oracle.com/cd/E37838_01/html/E66175/gplhi.htmlCloses#22143.
This fixes compilation after the changes of 148f56085f (Define
NIIF_LARGE_ICON ourselves if it's not defined, 2022-02-19) for the
compilers using headers so old that they don't even define NIIF_USER
which exists since XP SP2 -- which includes MinGW 5.3 used for our
AppVeyor CI builds.
Instead of getting the bitmap of the default size and then possibly
rescaling it, get the icon of the correct size directly from the bitmap
bundle, which could yield better results (and at worst would do the same
thing as before, but with less code).
See #22138.
It should actually always be defined when using any of the currently
supported compilers, but define it ourselves just in case it isn't.
At least get rid of #ifdef checking for it in the code.
See #22138.
