Apparently in at least some gcc versions fpos_t is not the same thing as
long long int, resulting in compilation problems due to using different
types for the ternary operator arguments.
It's unexpected that decreasing the width of the control makes the last
column diminish in size until nothing (at least if it's minimum size was
not set), instead of showing horizontal scrollbar, so prevent this from
happening by considering the initial column width as being "manually
set", which prevents the code from making the column narrower than it
automatically.
This seems to make sense and is consistent with the handling of initial
size, which becomes "best", and hence "minimal", size of the control,
for wxWindow.
Closes#18343.
When running under Wayland and using two-finger scrolling, GTK seems to only
emit "smooth" scrolling events. The code was already setup to handle
GDK_SCROLL_SMOOTH events - we just needed to add GDK_SMOOTH_SCROLL_MASK to the
event mast of the window in order to receive the events.
Fixes#17734.
Although white background was used if the brush was never set, setting
and resetting it nothing unexpectedly (and inconsistently with the other
ports) resulted in using black background.
Fix this by just resetting the brush to white if it's invalid, instead
of not setting it at all.
To preserve layout, old wxComboBox should be replaced with the new one in the same location in the sizer. So we need to create a new combo box first, put in the sizer and only after that the old combo box can be deleted.
Since 848f5e78a6 the background wasn't
cleared at all if the background brush hadn't been explicitly set prior
to calling wxDC::Clear(). This was incompatible with the old behaviour
and even managed to break our own print preview code, so it clearly
wasn't a good idea to change Clear() like this.
Instead, continue to clear the DC using white background by default,
while still not doing anything if a transparent background brush had
been explicitly set.
This fixes print preview background under MSW.
See #10273, #18371.
Checking whether alpha info flag contains one of
the constants representing image/bitmap with alpha
channel is done more than once in the code so this
check can be moved to a dedicated function.
This makes the collapsible pane allocated exactly its best size better
looking, as the header is centered now, instead of having a border on
the left side, but not on the right one.
See #18515.
DoGetBestSize() actually calculated the best client size and not the
full size, as it didn't take the pane border into account.
Fix this in the simplest possible way, by just renaming the function to
DoGetBestClientSize() instead.
This ensures that the pane is actually big enough to show its contents,
without cutting off the text shown in its header.
Closes#18515.
Make it more clear that we do what we do in DoGetBestClientSize()
because of what our OnPaint() does.
Incidentally fix off by 1 (or 2 under non-MSW platforms) mismatch
between the 2 methods: the gap between the button and the text is
actually just 2 DIPs, not 4, but we need to add another 1 DIP for the
focus rectangle under MSW.
Somehow using enum results in wrong comparison being done when using gcc
5.3 (the test run when using this compiler fails because the RHS value
is "true" and not the numeric value of the enum element), so add a cast
to make this work.
This key combination is used for selecting all text, while it's also
relatively common to use it as an accelerator for some menu item.
Resolve the conflict in favour of wxTextCtrl, i.e. let it have this key
when it has focus, while still allowing to use it as an accelerator
otherwise.
