No real changes, just some cleanup to ensure that the item being edited stored
in wxDataViewRendererBase never refers to an item which is not, actually,
being edited any longer.
Call FinishEditing() when the cell is done being edited instead of sending
wxEVT_DATAVIEW_ITEM_EDITING_DONE event manually from wxGTK code.
This fixes the bug with not being able to edit an item the second time if the
editor was dismissed by GTK+ itself and not by our own code (which already did
call FinishEditing()) as the old editor still remained alive because
DestroyEditControl(), usually called from FinishEditing(), wasn't executed.
It also removes code duplication and avoids the need to keep a global
s_user_data pointer as the item currently being edited is already stored in
wxDataViewRenderer anyhow.
When using a custom renderer, wxEVT_DATAVIEW_ITEM_EDITING_STARTED was sent
twice: once from the generic base class StartEditing() and another time from
the GTK-specific "editing_started" signal handler. And we must send it from
the latter, because otherwise no event would be generated at all for the
standard renderers (i.e. text cells) for which we don't call StartEditing()
ourselves, so don't call it from the former and instead generate the event by
explicitly calling NotifyEditingStarted() after calling StartEditing() in the
generic version (as for wxOSX version, it doesn't use StartEditing() at all
and so doesn't need to be changed).
Reuse the same code from the generic and native GTK and OS X implementations
of wxDataViewCtrl instead of triplicating it.
This fixes a small discrepancy between the wxOSX version, which didn't see the
model pointer correctly in the generated event, and all the others, but mainly
paves way for future improvements.
We need to delete the editor control even if GetValueFromEditorCtrl() failed
as otherwise we're never going to be able to edit another item again, at least
with the GTK native implementation, which does nothing when starting to edit
an item if an editor control already exists.
Use the existing wxRect::Get{Position,Size}() methods instead of explicitly
creating the objects from the wxRect components, this is simpler and more
readable.
No real changes.
Skipping the event is supposed to have the same effect as not handling the
event at all, but in wxMSW wxTE_PROCESS_ENTER style must be specified for a
wxEVT_TEXT_ENTER handler to be executed at all and if this style is used, then
the default handling in MSWProcessMessage() which normally happens before
calling the handler doesn't take place at all.
Work around this by explicitly performing the default "Enter" key action if
the event generated by it wasn't handled to make wxMSW behaviour more
intuitive.
This will make it possible to reproduce the default "Enter" key functionality
from elsewhere.
Almost no changes yet, the only minor change is that we now wouldn't try to
"click" any windows using DLGC_DEFPUSHBUTTON dialog code but which are not
really buttons -- but then this shouldn't ever happen anyhow.
Override newer and more flexible virtual method: this doesn't change anything
yet but will allow to provide default handling for some messages in a single
overridden method in the future commits, when it would have required to also
override MSWDefWindowProc() with the old method.
While Enter and Tab on their own should be used to finish cell editing, the
cell editor itself may want to process key combinations involving these keys
with modifiers, e.g. Shift-Enter, so don't intercept those in at least the
generic version of wxDataViewCtrl to allow catching them in the editor.
This is never going to work, so complain about trying to do it to help with
catching this bug.
This is possible thanks to the new OnDynamicBind() method invoked whenever a
dynamic event handler is bound to a control, so this doesn't detect all
possible occurrences of the bug (such as specifying the handler in a static
event table or in a validator), but it's still better than nothing.
In the future OnDynamicBind() should be extended for other invalid calls, e.g.
binding a handler for wxEVT_TEXT_ENTER to a non-text control shouldn't work
neither, ideally.
Use the same GetParentForModalDialog() method as for the normal dialogs to
find the parent to use for this native dialog and ensure that it is shown
modally even if no parent is explicitly specified when constructing it.
Closes#17384.
Use wxConvWhateverWorks when converting the command line given as a string to
individual arguments: we already used wxSafeConvertWX2MB() when converting the
arguments specified as an array, but not here.
Closes#16206.
These functions were almost but not quite identical to it:
wxSafeConvertMB2WX() tried the current locale encoding before UTF-8 while
wxConvWhateverWorks tries UTF-8 first and then the current locale encoding.
The latter behaviour is more correct as valid UTF-8 could be misinterpreted as
some legacy multibyte encoding otherwise, so get rid of this difference and
just forward these functions to wxConvWhateverWorks.
This ensures that we can create output files with Unicode names even when
they're not representable in the current locale encoding, notably when the
current locale has never been changed and is still the default "C" one, not
supporting anything else other than 7 bit ASCII.
Credits for the new class name go to Woody Allen.
Fall back to UTF-8 rather than not outputting anything at all if the string is
not representable in the current locale encoding.
Even if we did try to handle this error by setting failbit, chances of anybody
checking for it (especially on e.g. std::cout) were very low and the only
possible workaround in practice would have been attempting to output the
string in UTF-8 anyhow, so just do it ourselves.
See #17358.
Ensure that we do output the string contents even if we have to encode it in
UTF-8 instead of the current locale encoding -- this is still better than not
outputting anything at all.
Closes#17358.
Disable more libraries and use --disable-optimize instead of --enable-debug
(also for MinGW) as we don't really need debug information in these builds and
not generating it might make things faster.
Include wx/msw/uxtheme.h even when wxUSE_UXTHEME == 0 because
wxUxThemeEngine::GetIfActive() and MARGINS struct are still used even then in
anybutton.cpp and menuitem.cpp respectively.
If a class not using "override" for its other, not wx-related, virtual methods
included wxDECLARE_EVENT_TABLE() with wxOVERRIDE inside it, it resulted in a
clang -Winconsistent-missing-override warning per each virtual method without
it which was very annoying.
Avoid it by not using wxOVERRIDE in this macro and explicitly disabling the
-Winconsistent-missing-override for the methods inside it in case the rest of
the class does use "override".
Notice that this also required rearranging the order of the declarations in
this macro to ensure that a semicolon is still required after it.
Closes https://github.com/wxWidgets/wxWidgets/pull/217
For ComCtl32 prior to 6.0 all text labels in wxListCtrl need to be re-assigned when new wxImageList is set in order to position them correctly within the control.
ComCtl32 prior to 6.0 doesn't support images with alpha channel so if we have 32-bit icon with transparency we need to convert it to a wxBitmap and then add this bitmap to wxImageList. If required, bitmap with alpha channel will be converted to the mask in wxImageList::Add.
This symbol is tested using "#if", so it should be always defined and there is
no need for testing whether this is the case.
Moreover, doing this useless check triggers warning C4574 (which is disabled
by default, but it's useful enough to enable it explicitly) with VC14 about
using "#ifdef" with a symbol defined as 0.
1. Override wxWindow::GetContentScaleFactor() to use gdk_window_get_scale_factor()
when available, and to use correct scale (1.0) otherwise, as wxDC::GetPPI()
(used by overridden method) is not properly implemented for wxGTK
2. Record scale in wxBitmap(wxImage&) ctor and wxBitmap::CreateScaled()
3. Adjust cairo scale for drawing bitmap, and (inversely) for drawing on bitmap
These tests don't seem to be sound, it's possible for one second or more to
pass between the two calls to GetTimes() and it's apparently possible for the
access time on the symlink to not change (due to /tmp being mounted with
"noatime" option perhaps?), resulting in Travis CI build failures.
