Paul Cornett [Thu, 2 Aug 2012 16:25:14 +0000 (16:25 +0000)]
Fix wxGCDC::Clear() for Cairo, and possibly MSW.
Maximum positive coordinate Cairo can handle is 2^23 - 1.
Also convert coordinates to logical so it works right with modified origin or scale.
See #14529
Fix comparisons of char and wchar_t with wxUniChar[Ref].
Comparisons didn't work correctly in the other direction as they were not
reversed as needed. Fix this by adding wxDEFINE_COMPARISON_REV() macro which
defines comparisons in terms of the reverse operations and use it for both
wxUniChar and wxUniCharRef.
Default user-locale-dependent LCID may be inappropriate for some situations,
notably Microsoft Excel uses localized formula names for non-English LCIDs.
So add a way to change the LCID to use at wxAutomationObject level while
preserving the old behaviour by default.
The text data should use CR LF EOLs under Windows but the data inside the
program typically has only LF EOLs, so translate between them automatically in
wxTextDataObject.
No real changes, just use accessors in wxTextDataObject.
Use {Set,Get}Text() instead of using m_text directly in wxTextDataObject as
these virtual methods overridable and should be called in case the class
doesn't actually use m_text for its storage at all.
Fix bug in wxNumberFormatter::ToString() for negative numbers.
Don't include the possible leading sign in the span of digits to be grouped as
this gave nonsensical strings such as "-,123" when adding thousands separators
to "123".
Warn if key presses occur when the input window doesn't have focus as the
usual key events are not generated then which could be understood as a bug in
the sample instead of the correct result.
Generate clipboard events for wxComboBox in wxGTK too.
These events were only generated for wxTextCtrl but should be sent for
non-readonly wxComboBox too, so refactor the code to allow its reuse from
wxComboBox.
Also add EVT_TEXT_PASTE handlers for both controls to the widgets sample for
testing.
Revert HasModifiers() change in behaviour, add HasAnyModifiers().
In 2.8 wxKeyEvent::HasModifiers() returned false if (only) Shift was pressed
as it tested for Control and Alt only but when it was moved to wxKeyboardState
in r55745 it started checking for all modifiers as this made more sense now
that it was used by wxMouseEvent. However it broke existing code using it,
including in wxWidgets itself (in wxTreeCtrl), so revert it now and add
HasAnyModifiers() that does check for all modifiers, including Shift.
Add a wxGTK-specific function to set wxNotificationMessage icon name.
It's trivial to use a stock icon with the given name in wxGTK with libnotify,
so provide a way to do it. However this is not as simple as that in other
implementations (notably Windows), so make it private to this port for now.
In the future we should try to support arbitrary wxIcons as well as extend
wxIconLocation to support FreeDesktop stock icon names.
Václav Slavík [Sun, 29 Jul 2012 08:52:52 +0000 (08:52 +0000)]
Fix installed wx-config broken by r72205.
In an attempt to fix DESTDIR support on platforms without symbolic
links, this commit completely broke the wx-config symlink created by
"make install" everywhere (unless you were insane enough to use
"--prefix=/", that is): $(libdir) is absolute path and so using
../$(libdir) has no chance of working.
Fixed by using `basename $(libdir)`. This is still unsatisfactory,
because it assumes that $(libdir) and $(bindir) are at the same level in
the filesystem -- which, while usually true, is by no means guaranteed.
But at least this works most of the time.
Implement wxNotificationMessage using libnotify in wxGTK.
Use libnotify -- if detected by configure -- to provide native notifications
in wxGTK. Our API maps to libnotify one in rather straightforward way, we
might consider extending it to cover more of libnotify functionality
(categories, user-defined icons and, especially, actions) later.
Also update the dialogs sample to show another kind of notification and the
documentation to clarify the behaviour of various methods.
Add all extra GTK libraries to GUI_TK_LIBRARY in configure.
Some libraries (e.g. Pango) were added to GUI_TK_LIBRARY already while others
(e.g. Hildon) had their own EXTRALIBS_HILDON variables that were then added to
EXTRALIBS_GUI separately. Finally, some others were simply broken because
neither EXTRALIBS_GNOMEVFS nor EXTRALIBS_GTKPRINT were set anywhere.
Standardize on using GUI_TK_LIBRARY for everything.
Václav Slavík [Wed, 25 Jul 2012 16:31:58 +0000 (16:31 +0000)]
wxMSW: Use TBSTYLE_AUTOSIZE for toolbar buttons with horizontal text.
Without this style, all buttons share the same width and look ugly when
their labels differ even a little in their lengths. With
TBSTYLE_AUTOSIZE, toolbar buttons use the size of their image plus
label's width and look better, even though their sizes differ.
This results in a noticeable delay when using wxGTK via a remote X11
connection for every key event as a round trip to server is needed to get the
mouse pointer position every time a key is pressed or released.
Only provide the position on demand. And explain that it's actually not very
useful as it's simply the same as the current mouse position.
Use relative path for wx-config symlink in "make install".
This fixes "make DESTDIR=..." Broken by r71050 but still allows "make
DESTDIR=... install" to work even under MinGW where "ln -s" is mapped to "cp"
and so doesn't work with non-existing first argument.
No real changes, just avoid overloaded virtual wxWebView::SetPage().
Instead, have two public non-virtual SetPage() methods forwarding to a private
DoSetPage(), as usual.
This avoids the need for "using wxWebView::SetPage" which is needed to avoid
warnings about hiding the other base class virtual when implementing one of
them and which was forgotten in wxMSW version resulting in warnings when using
g++ to compile it.
Handle accelerators earlier in keyboard processing code in wxGTK.
Translate key presses to accelerators before sending wxEVT_KEY_DOWN.
Also check for accelerators even for the key combinations handled by IM,
normally IM should take precedence but IM seems to intercept common keys such
as Shift+anything which it makes sense to use as accelerators.
Fix assert in webview sample when viewing pages without a title.
Trying to append a menu item with empty label resulted in an assert (at least
under wxGTK but probably elsewhere too), and in any case wasn't very useful,
so use "(untitled)" instead in this case.
We were getting crashes due to using methods of already half-destroyed object
when wxWebView window was destroyed in wxGTK, with the following abbreviated
stack:
#0 wxgtk_webview_webkit_load_status (widget=0x12eb380, webKitCtrl=0x13ee040) at src/gtk/webview_webkit.cpp:38
...
#8 0x00007ffff05ca281 in dispatchDidFailLoad (error=..., this=0x7fffe32db900) at ../Source/WebKit/gtk/WebCoreSupport/FrameLoaderClientGtk.cpp:1054
#9 WebKit::FrameLoaderClient::dispatchDidFailLoad (this=0x7fffe32db900, error=...) at ../Source/WebKit/gtk/WebCoreSupport/FrameLoaderClientGtk.cpp:1049
#10 0x00007ffff0b0de54 in WebCore::FrameLoader::checkLoadCompleteForThisFrame (this=0x7fffe3221b70) at ../Source/WebCore/loader/FrameLoader.cpp:2128
#11 0x00007ffff0b0cfde in WebCore::FrameLoader::checkLoadComplete (this=<optimized out>) at ../Source/WebCore/loader/FrameLoader.cpp:2370
#12 0x00007ffff0b0a3b2 in WebCore::FrameLoader::receivedMainResourceError (this=0x7fffe3221b70, error=..., isComplete=true) at ../Source/WebCore/loader/FrameLoader.cpp:2640
#13 0x00007ffff0b423b7 in cancel (error=..., this=0x7fffe3279680) at ../Source/WebCore/loader/ResourceLoader.cpp:399
#14 WebCore::ResourceLoader::cancel (this=0x7fffe3279680, error=...) at ../Source/WebCore/loader/ResourceLoader.cpp:353
#15 0x00007ffff0b420c0 in WebCore::ResourceLoader::cancel (this=<optimized out>) at ../Source/WebCore/loader/ResourceLoader.cpp:350
#16 0x00007ffff0afda9e in WebCore::DocumentLoader::stopLoading (this=0x7fffa081d800) at ../Source/WebCore/loader/DocumentLoader.cpp:258
#17 0x00007ffff0b09561 in stopAllLoaders (clearProvisionalItemPolicy=WebCore::ShouldClearProvisionalItem, this=0x7fffe3221b70) at ../Source/WebCore/loader/FrameLoader.cpp:1523
#18 WebCore::FrameLoader::stopAllLoaders (this=0x7fffe3221b70, clearProvisionalItemPolicy=WebCore::ShouldClearProvisionalItem) at ../Source/WebCore/loader/FrameLoader.cpp:1501
#19 0x00007ffff0b0d17a in WebCore::FrameLoader::stopForUserCancel (this=0x7fffe3221b70, deferCheckLoadComplete=false) at ../Source/WebCore/loader/FrameLoader.cpp:1541
#20 0x00007ffff05f3a36 in webkit_web_view_dispose (object=0x12eb380) at ../Source/WebKit/gtk/webkit/webkitwebview.cpp:1327
#21 0x00007ffff27ff7a0 in g_object_run_dispose (object=0x12eb380) at /tmp/buildd/glib2.0-2.32.3/./gobject/gobject.c:1061
#22 0x00007ffff4476c1e in gtk_scrolled_window_forall (container=0x7fff9c007d10, include_internals=0, callback=0x7ffff452cd20 <IA__gtk_widget_destroy>, callback_data=0x0)
at /tmp/buildd/gtk+2.0-2.24.10/gtk/gtkscrolledwindow.c:1085
#23 0x00007ffff439872f in gtk_container_destroy (object=0x7fff9c007d10) at /tmp/buildd/gtk+2.0-2.24.10/gtk/gtkcontainer.c:1073
...
#28 0x00007ffff443cd60 in gtk_object_dispose (gobject=0x7fff9c007d10) at /tmp/buildd/gtk+2.0-2.24.10/gtk/gtkobject.c:421
#29 0x00007ffff27ff7a0 in g_object_run_dispose (object=0x7fff9c007d10) at /tmp/buildd/glib2.0-2.32.3/./gobject/gobject.c:1061
#30 0x00007ffff6cb65ac in wxWindow::~wxWindow (this=0x13ee040, __in_chrg=<optimized out>) at src/gtk/window.cpp:2367
#31 0x00007ffff6d71241 in wxControlBase::~wxControlBase (this=0x13ee040, __in_chrg=<optimized out>) at src/common/ctrlcmn.cpp:49
#32 0x00000000005cf761 in wxControl::~wxControl (this=0x13ee040, __in_chrg=<optimized out>) at include/wx/gtk/control.h:27
#33 0x00007ffff7bd1eae in wxWebView::~wxWebView (this=0x13ee040, __in_chrg=<optimized out>) at include/wx/webview.h:96
#34 0x00007ffff7bd3443 in wxWebViewWebKit::~wxWebViewWebKit (this=0x13ee040, __in_chrg=<optimized out>) at include/wx/gtk/webview_webkit.h:26
#35 0x00007ffff7bd34ae in wxWebViewWebKit::~wxWebViewWebKit (this=0x13ee040, __in_chrg=<optimized out>) at include/wx/gtk/webview_webkit.h:26
Fix this by setting m_isBeingDeleted flag for wxWebViewWebKit early and
checking it in the callback.
Don't change client size of TLW when centering it in wxGTK.
Setting the frame client size and then calling Centre() on it resulted in the
frame coming up with a wrong client size in wxGTK because the call to Centre()
resulted in calling wxTopLevelWindow::DoSetSize() which unconditionally reset
m_deferShowAllowed to true, undoing setting it to false in DoSetClientSize().
Only reset m_deferShowAllowed if the size was really changed but not if the
window was only moved, as happens when centering it.
Václav Slavík [Fri, 20 Jul 2012 20:23:45 +0000 (20:23 +0000)]
Call Realize() later in XRC toolbar handler (patch #13888).
This is a workaround for a deeper compatibility problem in Cocoa
implementation (see the bug for detailed discussion), but for now, this
simple workaround is much better than not doing nothing.
Use both URL-specific and plain text formats in wxGTK wxURLDataObject.
Just as in wxMSW, it makes sense to put URLs on clipboard (or drag them) in
both URL-specific and plain text formats to facilitate pasting (or dropping)
them into other applications.
So make wxURLDataObject in wxGTK a composite data object containing both its
old data object and wxTextDataObject.
Use proper format in wxGTK clipboard selection handler.
For some incomprehensible reason, GDK_SELECTION_TYPE_STRING was hard coded for
all non-text formats even though it was clearly never the right format to use.
Use the correct value specified by the format instead.
Use text/uri-list instead of x-moz-url in wxGTK wxURLDataObject.
The standard exchange format for URLs is text/uri-list and not the deprecated
and Firefox-specific x-moz-url, support for which was moreover implemented
incorrectly anyhow.
Also add an example of copying URLs to the dnd sample.
See https://developer.mozilla.org/En/DragDrop/Recommended_Drag_Types for more
information.
Don't use delayed destruction for sockets in other threads.
The delayed destruction mechanism is not MT-safe, so using it for wxSocket
objects destroyed from threads other than main resulted in crashes. Luckily,
it is not necessary to use it for such sockets anyhow as they don't risk
receiving any events -- which are only dispatched in the main thread -- and so
can be destroyed immediately.
So do destroy them directly instead of just scheduling for later destruction
when wxSocket::Destroy() is called.