Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
branch: mwf-handle-bra…
Commits on Mar 5, 2007
  1. In Test/System.Windows.Forms:

    Chris Toshok authored
    2007-03-05  Chris Toshok  <toshok@ximian.com>
    
    	* ControlHandleTest.cs: make this build on both 1.1 and 2.0
    	profiles.
    
    In System.Windows.Forms:
    2007-03-05  Chris Toshok  <toshok@ximian.com>
    
    	* Control.cs (AccessibilityNotifyClients): turns out in 1.1 this
    	causes handle creation.
    
    
    svn path=/branches/mwf-handle-branch/; revision=73757
Commits on Feb 28, 2007
  1. In Test/System.Windows.Forms:

    Chris Toshok authored
    2007-02-28  Chris Toshok  <toshok@ximian.com>
    
    	* ControlTest.cs: Flesh out ControlInvokeTest to make sure that
    	the Control thread dies (that Application.Run returns, basically).
    	This works in isolation, but for some reason it isn't exiting when
    	run in concert with other tests (this test is the cause of the
    	hangs running the unit tests.)  Marking NotWorking for the time
    	being.  This needs to be fixed before merging back to the
    	mainline, imo.
    
    In System.Windows.Forms:
    2007-02-28  Chris Toshok  <toshok@ximian.com>
    
    	* ApplicationContext.cs: Add a flag to make sure we only raise the
    	ThreadExit event once (ExitThreadCore can be indirectly called
    	from a few places.)  I don't like the additional flag, but it
    	makes the event ordering/count correct.
    
    	* Application.cs (MWFThread.LoopCount): don't use an enumerator
    	without locking the collection.  An enumerator doesn't give us any
    	protection from modification anyway.  Lock the thread hash and
    	replace the complicated enumerator loop with a foreach.
    	(Application.CloseForms): make internal so it can be called from
    	ApplicationContext.  This should probably be moved to MWFThread.
    	(Application.ExitThread): don't call MWFThread.Current.Exit()
    	here.  just call XplatUI.PostQuitMessage.  We'll exit the thread
    	when the runloop exits (in response to WM_QUIT.)
    	(Application.RunLoop): add a comment (and check) for
    	context.MainForm being null after setting context.MainForm.Visible
    	= true.  This is because you're perfectly free to dispose of a
    	form in VisibilityChanged.  Chalk this up to another case where we
    	need to synchronously generate WM_ACTIVATE from Control.Show.
    	Also, add handling for WM_QUIT here so we'll exit the loop.
    	
    	* XplatUIX11.cs: clean up MapWindow and UnmapWindow a bit.  The
    	fact that we don't wait if we're only unmapping the whole_window
    	makes me a bit nervous, but it doesn't seem to cause any problems
    	yet.
    
    	also, add a comment about the stupid, broken and wrong resetting
    	of PostQuitState to false in GetMessage().
    
    	In PostQuitMessage, we need to add a WM_QUIT message to the
    	thread's queue.  We use the FosterParent to get the right
    	handle/hwnd/queue.
    
    	Lastly, in SetVisible, we need to unmap both windows, since the
    	waiting only happens when we're unmapping the client window.  So
    	now, the *only* time we unmap just the whole_window is in the hack
    	for resizing a control to 0,0.
    	
    
    svn path=/branches/mwf-handle-branch/; revision=73571
Commits on Feb 21, 2007
  1. 2007-02-21 Chris Toshok <toshok@ximian.com>

    Chris Toshok authored
    	* Application.cs (CloseForms): rewrite this so that we don't
    	modify the list while we're traversing it.
    
    
    svn path=/branches/mwf-handle-branch/; revision=73265
  2. * ListBox.cs (.ctor): move the Control.AddImplicits here instead

    Chris Toshok authored
    	of OnHandleCreated.
    	(HorizontalScrollEvent): only call XplatUI.ScrollWindow if the
    	handle is created.  otherwise we'll create it here.
    	(VerticalScrollEvent): same here.
    
    	* Application.cs (CloseForms): call Form.Dispose, don't post
    	WM_CLOSE_INTERNAL.
    
    	* Form.cs (WndProc): we don't need to use CLOSE_INTERNAL
    	here. Application should Dispose() of the Form's.
    
    	* XplatUIX11.cs (WaitForHwndMessage): break out of the loop on
    	WM_DESTROY as well.
    	(MapWindow,UnmapWindow): only actually do the waiting for
    	SHOWWINDOW if the control we're dealing with is a Form.
    	(CreateWindow): if the control isn't a form, SendMessage
    	WM_SHOWWINDOW here (if the WS_VISIBLE style is set).
    
    	* Control.cs (SetVisibleCore): always use is_visible here, not
    	value.  If we use value, we can end up re-setting something
    	visible if, for instance, you do Control.Hide() in a delegate
    	attached to VisibleChanged as we do in FormTest.ShowDialogTest.
    
    2007-02-20  Chris Toshok  <toshok@ximian.com>
    
    
    svn path=/branches/mwf-handle-branch/; revision=73231
  3. 2007-02-20 Chris Toshok <toshok@ximian.com>

    Chris Toshok authored
    	* XplatUIX11.cs (WaitForHwndMessage): we need to loop until we get
    	the message we need.  PeekMessage returning false should not be a
    	condition under which we exit the loop.
    
    
    svn path=/branches/mwf-handle-branch/; revision=73226
Commits on Feb 15, 2007
  1. mark some more tests as NotWorking

    Chris Toshok authored
    svn path=/branches/mwf-handle-branch/; revision=72956
  2. 2007-02-15 Chris Toshok <toshok@ximian.com>

    Chris Toshok authored
    	* Control.cs (Refresh): only refresh if we've got a handle and are
    	visible.
    	(CreateAccessibilityInstance): CreateControl() here.
    	(UpdateChildrenZOrder): complicate the code loop even more by
    	taking into account controls that haven't had their handle
    	created, and those that aren't visible.  But on the flip side,
    	simplify the code by splitting it into two loops.  one which
    	builds up the list of child controls we're interested in, and the
    	other that sets the z order of those children.
    
    	* Control.cs: Control.AccessibilityObject causes the control to be
    	created, not just the handle.
    
    2007-02-14  Chris Toshok  <toshok@ximian.com>
    
    
    svn path=/branches/mwf-handle-branch/; revision=72955
Commits on Feb 14, 2007
  1. 2007-02-14 Chris Toshok <toshok@ximian.com>

    Chris Toshok authored
    	* Control.cs: rework UpdateChildrenZOrder to correctly handle the
    	problem on X where a window might have its handle created (and be
    	visible) while the window is unmapped.  calling XConfigureWindow
    	on an unmapped window is bad, and generates X errors.
    
    
    svn path=/branches/mwf-handle-branch/; revision=72854
  2. 2007-02-13 Chris Toshok <toshok@ximian.com>

    Chris Toshok authored
    	* Control.cs (CreateHandle): don't loop over our children setting
    	their parent here.  do it when in WndProc when we're shown.
    	(UpdateChildrenZOrder): make this internal so we can call it from
    	ScrollableControl.
    	(WndProc): for WM_SHOWWINDOW, reparent the child control after
    	creating its handle.  Also, remove the calls to PerformLayout from
    	here.  they're done in ScrollableControl.OnVisibleChanged.  Also,
    	OnVisibleChanged only seems to be called directly here for the
    	toplevel control.  It's propagated down the window hierarchy by
    	calls to child.OnParentVisibleChanged.
    	(OnVisibleChanged): don't do layout here - it's done (oddly
    	enough, according to a glance at stack traces on ms.net..) in
    	ScrollableControl.
    	
    	* ScrollableControl.cs (OnVisibleChanged): make sure we update the
    	z order of our children before calling PerformLayout.
    
    
    svn path=/branches/mwf-handle-branch/; revision=72806
Commits on Feb 13, 2007
  1. 2007-02-12 Chris Toshok <toshok@ximian.com>

    Chris Toshok authored
    	[big change, fixes #80020]
    	
    	* AccessibleObject.cs: we need to make owner internal again to fix
    	some of ControlAccessibleObject.
    
    	* Control.cs: lots of changes here.  add support for WM_CREATE,
    	for which we generate OnHandleCreated.  Remove the OnHandleCreated
    	call from CreateHandle.  Also add support for WM_SHOWWINDOW where
    	we create child controls.  leave the MonoTODO's for the
    	accessibility calls, but fix the exceptions so the tests pass.
    
    	Add the InvalidOperationExceptions to Invoke methods, and remove a
    	couple of InvokeInternal methods we aren't using.
    	
    	Also, add a couple of CreateHandle calls in places where we know
    	the handles are being created but our code doesn't reference
    	.Handle.
    
    	Make SetVisibleCore call OnVisibleChange if the handle isn't
    	created.  If the handle is created, we rely on XplatUI.SetVisible
    	generating the event synchronously.
    	
    	Lastly, make sure we don't use this.Handle inside CreateHandle,
    	because we can call back into client (and that code can dispose of
    	the control).
    
    	* XplatUIStructs.cs: misc/cleanup.
    
    	* XplatUIX11.cs: Map/Unmap X events correspond to WM_SHOWWINDOW,
    	although we don't populate the wParam properly.
    	(CreateWindow): generate WM_CREATE.
    	(MapWindow,UnmapWindow): make these calls synchronous, at great
    	performance expense (particularly in the unmap case), to match
    	win32 behavior.
    
    	* Form.cs (.ctor): remove the call to UpdateBounds. we don't need
    	to call it.
    	(set_MdiParent): don't recreate the handle unless it's been
    	created already.
    	
    	* MdiClient.cs (OnResize): don't InvalidateNC Parent.Handle unless
    	it's created.
    
    	* NativeWindow.cs: this is probably the weirdest part of the
    	patch.  We need a way to link up the window being created to the
    	WM_CREATE message.  Since we can only be creating one window at a
    	time on a given thread, we keep track of a per-thread reference so
    	we can dispatch it properly.  We also need to keep track of the
    	Hwnd currently being created so that the win32 backend doesn't
    	have problems.
    	
    	* XplatUIWin32.cs: a similar change to the one we made in
    	NativeWindow.cs.
    
    
    svn path=/branches/mwf-handle-branch/; revision=72727
Commits on Feb 12, 2007
  1. start mwf handle work

    Chris Toshok authored
    svn path=/branches/mwf-handle-branch/; revision=72718
  2. @robert-j

    2007-02-12 Robert Jordan <robertj@gmx.net>

    robert-j authored
    	* HttpResponseStream.cs (BlockManager.EnsureCapacity):
    	don't call Marshal.ReAllocHGlobal on NULL because, unlike
    	realloc(3) and g_realloc, it doesn't support this semantic.
    	Fixes a part of bug #77075.
    
    
    svn path=/trunk/mcs/; revision=72714
  3. @robert-j

    In .:

    robert-j authored
    2007-02-12  Robert Jordan  <robertj@gmx.net>
    	* marshal.c, marshal.h, icall-def.h: rename mono_marshal_realloc
    	to ves_icall_System_Runtime_InteropServices_Marshal_ReAllocHGlobal
    	and use GlobalReAlloc on WIN32 to be in sync with GlobalAlloc,
    	GlobalFree. Fixes a part of bug #77075.
    
    
    svn path=/trunk/mono/; revision=72713
  4. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72711
  5. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72710
  6. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72709
  7. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72708
  8. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72707
  9. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72706
  10. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72705
  11. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72704
  12. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72703
  13. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72702
  14. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72701
  15. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72700
  16. Remove unneeded directory. Use Mainsoft.Web.AspnetConfig instead

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72699
  17. Remove unneeded directory. Use Mainsoft.Web.AspnetConfig instead

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72698
  18. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72697
  19. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72696
  20. Grasshopper project system now uses csproj extension

    Noam Lampert authored
    svn path=/trunk/mcs/; revision=72695
  21. @migueldeicaza

    2007-02-12 Miguel de Icaza <miguel@novell.com>

    migueldeicaza authored
    	* CStreamWriter.cs (Write with char []): take the lock once for
    	all characters and call manually the InternalWriteChar properly to
    	speed things up.
    
    	(Write with string parameter): same thing, if the driver is not
    	initialized use a fast path.
    
    	If the driver has not been initialized, use a fast path instead. 
    
    svn path=/trunk/mcs/; revision=72694
  22. 2007-02-12 Everaldo Canuto <everaldo@simios.org>

    Everaldo Canuto authored
    	* Label.cs: Fix CalcPreferredHeight for 2.0 that must return only
    	font height when compatible_text_rendering is false. Partially fix #80801.
    
    
    svn path=/trunk/mcs/; revision=72693
  23. @illupus

    Mon Feb 12 21:54:57 CET 2007 Paolo Molaro <lupus@ximian.com>

    illupus authored
    	* MonoType.cs: patch from Cedric Vivier <cedricv@neonux.com> to
    	get correctly non-public fields from generic types.
    
    
    svn path=/trunk/mcs/; revision=72692
  24. revert r72565 and r72581. these will go on a branch and be merged bac…

    Chris Toshok authored
    …k in once they stabilize and don't break everyone.
    
    svn path=/trunk/mcs/; revision=72688
  25. @illupus

    Mon Feb 12 21:10:07 CET 2007 Paolo Molaro <lupus@ximian.com>

    illupus authored
    	* loader.c: implemented typedef parent in field memberref.
    
    
    svn path=/trunk/mono/; revision=72687
Something went wrong with that request. Please try again.