Ensure that m_pIDataObject is reset to avoid assertions from DragEnter()
during the next drag-and-drop operation after the one which resulted in an
exception being thrown.
All exceptions thrown by wxDropTarget::OnXXX() must be caught and handled in
the same way as we already do it for the exceptions in the event handlers as
we can't let the exceptions escape into system/kernel code: while it can work
in some cases, it doesn't work in general, e.g. exceptions simply disappear
when they happen in 32 bit programs under 64 bit Windows 7.
SizeWindows() optimized rendering calls to DrawSash() too aggressively
and wouldn't repaint e.g. slowly moving sash with wxSP_LIVE_UPDATE
enabled, even though child windows were repositioned.
Fix by always painting the sash from SizeWindows().
This overrides the default minimum size of 150+ pixels, avoiding swarms of
GTK3 debug warnings about "attempt to underallocate wxPizza's child GtkEntry"
Use a slightly higher idle priority so callback runs before TLW is deleted,
and ref the widget just to make sure it doesn't disappear. Avoids accessing
de-allocated memory.
Accessible object must be destroyed as soon as possible, because
otherwise a screen reader may try to query it before wxWindow destructor
removed it, but after ~wxDataViewCtrl destructor finished. The
wxACC_EVENT_OBJECT_DESTROY notification causes exactly that under JAWS.
A change in size-allocate handling with GTK+ 3.20 exposed a flaw in our method for
deferring queue-resize requests. Using an idle callback to process the requests
did not work well with the GdkFrameClock-based system used since GTK+ 3.8. Using
the "check-resize" signal works better. Also with GTK+ >= 3.20, it seems necessary
to manually work the queue-resize up to the TLW, as otherwise the resized widgets
don't get updated without an external size-allocate event.
See #17585
Significantly improve the appearance of wxStatusBar on macOS, both the
modern flat look since 10.10 and older versions. Increase the size to
match native bottom bars, center the text inside it, use appropriate
background and border colors and the same gray for the text as Finder
uses, correctly change the appearance for inactive windows.
This is still far from ideal - that would be using
setContentBorderThickness:forEdge: and rendering the text atop it. But
that seems to be much easier said than done due to interference from
other parts of wx. This is much better than the previous state.
It is not forbidden to have function calls in static initializers, and
such code compiles fine, contrary to claims made by the text. Explain the
real reason why wxTRANSLATE is necessary.
This is a default description of the renderer content (for accessibility purposes).
Thanks to this implementation there is not necessary to override GetAccessibleDescription() in the renderers derived from wxDataViewCustomRenderer.
Calling wxDataViewTreeNode::ToggleOpen() is not sufficient to actually expand/collapse the item.
Calls to wxDataViewMainWindow::Expand()/Collapse() are necessary to do so.
Since wxDataViewCtrlAccessible::GetName() calls to wxDataViewIconTextRenderer::GetAccessibleDescription() to directly retrieve the content of the renderer, there is no need to override wxDataViewCtrlAccessible::GetName() in order to get a proper content of the renderer in wxDataViewTreeCtrl.
Call wxDataViewRenderer::GetAccessibleDescription() to retrieve the content of the renderer instead of using a raw item value taken from wxDataViewModel. GetAccessibleDescription() returns a renderer-aware text dedicated for accessibility purposes and hence text presented in GetName() and GetDescription() is accurate in contrary to the text deduced from the item value.
The purpose of this method is to provide a textual description of the renderer's content to the class implementing accessibility framework in wxDVC (wxDataViewCtrlAccessible).
It is exposed if wxUSE_ACCESSIBILITY is set to 1.
Blind fix for compilation failures in MinGW builds after the changes of
92a1f643ba which started using HDS_NOSIZING
apparently not present in this compiler SDK headers.
Don't apply alignment to native controls in SetAlignment() method, where
it may not be known yet due to column-to-renderer inheritance if
wxDVR_DEFAULT_ALIGNMENT is used (the default). Move such code to
GtkUpdateAlignment() (which was made virtual) in all renderers.
This fixes unintended right-aligning of columns with GTK+ 2 when default
alignment was used.
Change the animation delay from ridiculously long (half a second) to
somewhat more reasonable 200ms to bring it in line with Windows per the
comment in this code. Feels more similar to native macOS animations now
too.
Fix the logic for restoring previous span's colors to account for the
possibility of spans that don't change the color, such as in "<i><span
color='red'>...</span></i>foo". Previously, "foo" would always be
rendered black, because unwinding the attributes stack would encounter
an invalid color (which has r=g=b=0) and set it, disregarding and
preexisting attributes.
Because there's code in there that checks whether the attributes are
valid, we need to keep track of both the actually specified attributes
and the currently effective ones, and use the latter for restoration.
Don't use custom color attributes or color markup for rendering of
selected items' text in wxOSX's wxDataViewCtrl and use the selection
text color instead. This is consistent with the other ports and should
be done to improve readability (in absence of selection-specific
attributes).
Fix wxDataViewTextRenderer to at least partially respect ellipsize mode
when using markup text. Generic implementation only supports
wxELLIPSIZE_END and wxELLIPSIZE_NONE at the moment, but the wxOSX and
wxGTK ones have full support.
Native Windows header control doesn't have the ability to show
per-column resizing cursor and it's only possible to enable or disable
it for the entire control, so we can't fully support the wxHeaderColumn
API. But we can at least hide the resizing cursor if none of the columns
are resizable.
By default double dashes are converted to en-dashes in the output (and triple
ones -- to em-dashes), but this is undesirable when double dashes are used not
as a punctuation mark but in command line options or as C++ decrement
operator, so escape them to avoid such conversion in this case.
