Remove caching of whether any of the children accept focus in
wxControlContainer as it can change at any moment under Mac, due to full
keyboard access being turned on and off (which can be done using
Ctrl-F7, i.e. easily, and so this might be something users actually do
and not just a theoretical edge case). This also incidentally fixes
caching of the wrong focusability state during window initialization,
when its children are not yet shown, as [NSView canBecomeKeyView:] used
for AcceptsFocusFromKeyboard() implementation under Mac, apparently
always returns false for hidden windows.
The behaviour under the other platforms should remain the same, but
AcceptsFocus() is slower now as it always has to query children instead
of being able to avoid doing it when we know that none of them accepts
focus anyhow. OTOH this only happens to the windows that don't accept
focus themselves and, at least at some moment, don't have any children
accepting focus neither, which should be quite rare and optimizing this
case doesn't seem to be worth the extra code complexity due to extra
preprocessor platform checks.
Closes#18089.
The radio box is just a static box and so can't have focus, only its
child radio buttons should have it.
This resolves the problem with getting stuck on wxRadioBox when full
keyboard access is off, as wxControlContainer code tried to give focus
to wxRadioBox because its AcceptsFocusFromKeyboard() returned true, but
none of its radio buttons could be focused without full keyboard access.
However this introduces a new problem with wxRadioBox being skipped when
full keyboard access is on, which will be fixed in the following commit.
See #18089.
Handle these events in the menu itself first, then the window and only
then at the application level instead of falling back on the application
before searching the window for the handler.
This requires using the logic already present in the base class
DoProcessEvent() method, so make it protected to allow reuse.
Closes#4769.
Size grip in generic wxStatusBar is only supported in wxGTK as it's
drawn using GTK functions, but this is not really a problem as other
platforms either use native implementations (MSW, Qt) or shouldn't show
size grip anyhow as it looks non-native (Mac).
So just ensure we don't leave space for the grip if it's not shown,
correcting the change of 6c1b2b23cf.
Closes#18469.
This renders unnecessary the previous workarounds for using NSOpenGLView
as the base class for wxGLCanvas view, so revert them.
See:
- 2ab430965c which introduced the use of
NSOpenGLView (see https://github.com/wxWidgets/wxWidgets/pull/846).
- ea68934b8e for the first workaround due
to the use of NSOpenGLView and its default context creation.
- 6974a6ceaa for the refinement of the
workaround above.
- e6ae83c386 and
b822e9efe8 which tweaked it futher.
This commit renders unnecessary all the changes since the initial switch
to NSOpenGLView by just preventing NSOpenGLView from creating its own
context and thus enforcing the use of wxGLContext associated with
wxGLCanvas.
Closes https://github.com/wxWidgets/wxWidgets/pull/1617
This can happen when using wxGrid and calling UseNativeColHeader()
before adding any columns and shouldn't be fatal, so simply don't do
anything in this case.
Because wxPGRootProperty has no parent, just like an unattached (waiting
for deletion) child wxPGProperty, so we have to first check whether
the property is the root and next whether it has a parent.
Instead of doing basically nothing, this call completely mislaid out the
control as -1 were used used literally instead of meaning "use the
current geometry", as they're supposed to. Fix this by handling
arguments with the value of -1 specially.
Such call was done by wxCompositeWindow::SetLayoutDirection() used by
the widgets sample and explained why the slider couldn't be seen when
switching to its page in this sample.
Closes#18442.
No real changes, just add a simple helper function instead of
duplicating it in 2 different places.
This also provides a unique place to write the comment explaining why
PMPrinterGetID() is being used instead of PMPrinterGetName().
See #16774.
Apparently, the user-readable name retrieved by PMPrinterGetName() is
not necessarily unique when multiple printers are available, so use
PMPrinterGetID() which is guaranteed to return a unique string.
Closes#16774.
This implementation suffers at least from 2 problems:
1. It doesn't support icons at all.
2. It toggles the checkbox when clicking on the text and not just on the
checkbox itself, as would be expected.
but it's still better than the current version which simply doesn't work
at all, i.e. can't be toggled in any way (and also doesn't draw itself
correctly when using dark mode under macOS 10.14+), so use it for now.
A better solution would be to fix the problem with ActivateCell() not
working at all (see #17746) and update the code to respect drawing in
dark mode.
Closes#17473.
Closes https://github.com/wxWidgets/wxWidgets/pull/904
Retain the graphics context for the helper overlay window as otherwise
it gets released leaving us with a dangling pointer.
Unfortunately this still doesn't make the existing code actually work.
See #18399.
Fixes to wxUIActionSimulator allowing the tests using it to work for
wxGrid in wxGTK.
And some improvements and bug fixes to wxGrid itself.
Closes https://github.com/wxWidgets/wxWidgets/pull/1609
This is always 0 for any still supported compiler, so remove the option
and configure checks for it.
Still define it as 0 for compatibility, just in case it's used outside
of the library.
The loop computing "colswidth" (the sum of widths including last column)
actually incorrectly computed "lastColX" (the sum of widths until the
last column), fix this.
See #18295.
Use "relative pixels" (known as DIPs in wx) instead of dialog units, as
the latter ones don't scale correctly due to rounding errors when using
high DPI.
Also take into account the fact that the 1px invisible border around the
visible part of the buttons is not scaled by the standard control, so
don't apply scaling to this part when determining the best size neither.
Closes#18528.
This results in buttons being forced to be too tall when using high DPI
and is useless anyhow, as all callers of this function deal with the
height by increasing it to the minimum acceptable value already (or
discard it entirely in wxMessageDialog::AdjustButtonLabels()).
See #18528.
Fix autosizing broken in 3c72396a36 and
not fully fixed by f7e335c031.
Simplify the code to make it more obviously correct, by separating the
computation of the extent suitable for the label and determining the
size to use taking into account the extents of both the column data and
the its column.
Also add the unit test checking that auto-sizing works correctly in all
the different cases.
Closes https://github.com/wxWidgets/wxWidgets/pull/1600
Co-Authored-By: Ilya Sinitsyn <the_siv@mail.ru>
On DC with a non-rotated coordinate system drawing a filled rectangle with ExtTextOut() Win API is faster than drawing a line with MoveTo()/LineTo() so on such DCs we can use ExtTextOut() to draw horizontal/vertical lines with solid colors and square ends instead of calling MoveTo()/LineTo().
See #18517.
This is unnecessary as TAB navigation is not supposed to work between
wxGrid children and actually harmful as this resulted in SetFocus()
doing nothing, instead of setting focus to wxGridWindow, if the focus
was on wxGrid itself for some reason (this happened at least in the grid
unit tests and resulted in failures because the in-place editor didn't
appear as expected).
Somehow, mouse release events generated immediately after the mouse
press ones are sometimes (not always, but often enough to reliably
prevent the test suite from passing) simply lost, i.e. not received by
the GTK event loop.
The only workaround seems to be to introduce an artificial delay, which
does avoid the problem, at the price of making the tests run longer and,
worse, not really solving the underlying problem, whatever it is. But
it's still better than nothing.
Restore the checks for the model stamp, reverting the changes of
18594afe76: we still need to ignore the
calls to at least iter_children() and iter_nth_child() model methods
that can be called from inside gtk_tree_view_set_model() when we reset
the model, as running these methods crashes when trying to use the
pointers to already deleted items.
For consistency and robustness, add checks for the model stamp to all
the methods and not just those two, just in case other ones end up being
called later in some way.
Also add a unit test checking that DeleteAllItems() doesn't crash and
does delete all items.
Closes#18533.
The previous fix in 5442edbbe9 corrected
the type for the older NetBSD versions, but not NetBSD 10, which now
uses "void*", as all the other platforms, so restrict the original fix
to NetBSD versions < 10 only.
See #18199.
For some reason, names were used for cyan and light grey, even though
all the other ones used RGB values since refactoring the stock objects
to be created on demand in f516d98637.
Use RGB values for these two as well now.
Closes#18530.
For some reason, mouse press events were still received without
flushing, but mouse release ones were not received by GTK itself (and a
fortiori by wxGTK) without it.
