views:

993

answers:

7

When the user grabs a corner of a resizable window, and then moves it, windows first moves the contents of the window around, then issues a WM_SIZE to the window being resized.

Thus, in a dialog where I want to control the movement of various child controls, and I want to eliminate flickering, the user first sees what windows OS thinks the window will look like (because, AFAICT, the OS uses a bitblt approach to moving things around inside the window before sending the WM_SIZE) - and only then does my dialog get to handle moving its child controls around, or resize them, etc., after which it must force things to repaint, which now causes flicker (at the very least).

My main question is: Is there a way to force windows NOT to do this stupid bitblt thing? Its definitely going to be wrong in the case of a window with controls that move as the window is resized, or that resize themselves as their parent is resized. Either way, having the OS do a pre-paint just screws the works.

I thought for a time that it might be related to CS_HREDRAW and CSVREDRAW class flags. However, the reality is that I don't want the OS to ask me to erase the window - I just want to do the repainting myself without the OS first changing the contents of my window (i.e. I want the display to be what it was before the user started resizing - without any bitblit'ing from the OS). And I don't want the OS to tell every control that it needs to be redrawn either (unless it happened to be one that was in fact obscured or revealed by the resize.

What I really want:

  1. To move & resize child controls before anything gets updated onscreen.
  2. Draw all of the moved or resized child controls completely so that they appear without artifacts at their new size & location.
  3. Draw the spaces inbetween the child controls without impacting the child controls themselves.

NOTE: Steps 2 and 3 could be reversed.

The above three things appear to happen correctly when I use DeferSetWindowPos() in combination with the dialog resource marked as WS_CLIPCHILDREN.

I'd get an additional small benefit if I could do the above to a memory DC, and then only do a single bitblt at the end of the WM_SIZE handler.

I have played with this for a while now, and I cannot escape two things:

  1. I still am unable to suppress Windows from doing a 'predictive bitblt'. Answer: See below for a solution that overrides WM_NCCALCSIZE to disable this behavior.

  2. I cannot see how one can build a dialog where its child controls draw to a double buffer. Answer: See John's answer (marked as answer) below for how to ask Windows OS to double buffer your dialog (note: this disallows any GetDC() in-between paint operations, according to the docs).


My Final Solution (Thank you everyone who contributed, esp. John K.):

After much sweat and tears, I have found that the following technique works flawlessly, both in Aero and in XP or with Aero disabled. Flicking is non-existent(1).

  1. Hook the dialog proc.
  2. Override WM_NCCALCSIZE to force Windows to validate the entire client area, and not bitblt anything.
  3. Override WM_SIZE to do all of your moves & resizes using BeginDeferWindowPos/DeferWindowPos/EndDeferWindowPos for all visible windows.
  4. Ensure that the dialog window has the WS_CLIPCHILDREN style.
  5. Do NOT use CS_HREDRAW|CS_VREDRAW (dialogs don't, so generally not an issue).

The layout code is up to you - its easy enough to find examples on CodeGuru or CodeProject of layout managers, or to roll your own.

Here are some code excerpts that should get you most of the way:

LRESULT ResizeManager::WinProc(HWND hwnd, UINT msg, WPARAM wparam, LPARAM lparam)
{
    switch (msg)
    {
    case WM_ENTERSIZEMOVE:
        m_bResizeOrMove = true;
        break;

    case WM_NCCALCSIZE:
        // The WM_NCCALCSIZE idea was given to me by John Knoeller: 
        // see: http://stackoverflow.com/questions/2165759/how-do-i-force-windows-not-to-redraw-anything-in-my-dialog-when-the-user-is-resiz
        // 
        // The default implementation is to simply return zero (0).
        //
        // The MSDN docs indicate that this causes Windows to automatically move all of the child controls to follow the client's origin
        // and experience shows that it bitblts the window's contents before we get a WM_SIZE.
        // Hence, our child controls have been moved, everything has been painted at its new position, then we get a WM_SIZE.
        //
        // Instead, we calculate the correct client rect for our new size or position, and simply tell windows to preserve this (don't repaint it)
        // and then we execute a new layout of our child controls during the WM_SIZE handler, using DeferWindowPos to ensure that everything
        // is moved, sized, and drawn in one go, minimizing any potential flicker (it has to be drawn once, over the top at its new layout, at a minimum).
        //
        // It is important to note that we must move all controls.  We short-circuit the normal Windows logic that moves our child controls for us.
        //
        // Other notes:
        //  Simply zeroing out the source and destination client rectangles (rgrc[1] and rgrc[2]) simply causes Windows 
        //  to invalidate the entire client area, exacerbating the flicker problem.
        //
        //  If we return anything but zero (0), we absolutely must have set up rgrc[0] to be the correct client rect for the new size / location
        //  otherwise Windows sees our client rect as being equal to our proposed window rect, and from that point forward we're missing our non-client frame

        // only override this if we're handling a resize or move (I am currently unaware of how to distinguish between them)
        // though it may be adequate to test for wparam != 0, as we are
        if (bool bCalcValidRects = wparam && m_bResizeOrMove)
        {
            NCCALCSIZE_PARAMS * nccs_params = (NCCALCSIZE_PARAMS *)lparam;

            // ask the base implementation to compute the client coordinates from the window coordinates (destination rect)
            m_ResizeHook.BaseProc(hwnd, msg, FALSE, (LPARAM)&nccs_params->rgrc[0]);

            // make the source & target the same (don't bitblt anything)
            // NOTE: we need the target to be the entire new client rectangle, because we want windows to perceive it as being valid (not in need of painting)
            nccs_params->rgrc[1] = nccs_params->rgrc[2];

            // we need to ensure that we tell windows to preserve the client area we specified
            // if I read the docs correctly, then no bitblt should occur (at the very least, its a benign bitblt since it is from/to the same place)
            return WVR_ALIGNLEFT|WVR_ALIGNTOP;
        }
        break;

    case WM_SIZE:
        ASSERT(m_bResizeOrMove);
        Resize(hwnd, LOWORD(lparam), HIWORD(lparam));
        break;

    case WM_EXITSIZEMOVE:
        m_bResizeOrMove = false;
        break;
    }

    return m_ResizeHook.BaseProc(hwnd, msg, wparam, lparam);
}

The resizing is really done by the Resize() member, like so:

// execute the resizing of all controls
void ResizeManager::Resize(HWND hwnd, long cx, long cy)
{
    // defer the moves & resizes for all visible controls
    HDWP hdwp = BeginDeferWindowPos(m_resizables.size());
    ASSERT(hdwp);

    // reposition everything without doing any drawing!
    for (ResizeAgentVector::const_iterator it = m_resizables.begin(), end = m_resizables.end(); it != end; ++it)
        VERIFY(hdwp == it->Reposition(hdwp, cx, cy));

    // now, do all of the moves & resizes at once
    VERIFY(EndDeferWindowPos(hdwp));
}

And perhaps the final tricky bit can be seen in the ResizeAgent's Reposition() handler:

HDWP ResizeManager::ResizeAgent::Reposition(HDWP hdwp, long cx, long cy) const
{
    // can't very well move things that no longer exist
    if (!IsWindow(hwndControl))
        return hdwp;

    // calculate our new rect
    const long left   = IsFloatLeft()   ? cx - offset.left    : offset.left;
    const long right  = IsFloatRight()  ? cx - offset.right   : offset.right;
    const long top    = IsFloatTop()    ? cy - offset.top     : offset.top;
    const long bottom = IsFloatBottom() ? cy - offset.bottom  : offset.bottom;

    // compute height & width
    const long width = right - left;
    const long height = bottom - top;

    // we can defer it only if it is visible
    if (IsWindowVisible(hwndControl))
        return ::DeferWindowPos(hdwp, hwndControl, NULL, left, top, width, height, SWP_NOZORDER|SWP_NOACTIVATE);

    // do it immediately for an invisible window
    MoveWindow(hwndControl, left, top, width, height, FALSE);

    // indicate that the defer operation should still be valid
    return hdwp;
}

The 'tricky' being that we avoid trying to mess with any windows that have been destroyed, and we don't try to defer a SetWindowPos against a window that is not visible (as this is documented as "will fail".

I've tested the above in a real project that hides some controls, and makes use of fairly complex layouts with excellent success. There is zero flickering(1) even without Aero, even when you resize using the upper left corner of the dialog window (most resizable windows will show the most flickering and problems when you grab that handle - IE, FireFox, etc.).

If there is interest enough, I could be persuaded to edit my findings with a real example implementation for CodeProject.com or somewhere similar. Message me.

(1) Please note that it is impossible to avoid one draw over the top of whatever used to be there. For every part of the dialog that has not changed, the user can see nothing (no flicker whatsoever). But where things have changed, there is a change visible to the user - this is impossible to avoid, and is a 100% solution.

A: 

If you can find a place to plug it in, CWnd::LockWindowUpdates() will prevent any drawing from occuring until after you unlock the updates.

But keep in mind this is a hack, and a fairly ugly one at that. Your window will look terrible during resizes. If the problem you are having is flickering during resizes, then the best thing to do is diagnose the flickering, rather than hiding the flickering by blocking paints.

One thing to look for are redraw commands that get called too often during the resize. If you r window's controls are calling RedrawWindow() with the RDW_UPDATENOW flag specified, it is going to repaint then and there. But you can strip out that flag and specify RDW_INVALIDATE instead, which tells the control to invalidate the window without repainting. It will repaint at idle time, keeping the display fresh without spazzing out.

John Dibling
LockWindowUpdate helps against repeated paints of the same control, but not agaisnt overlapping paints of different controls (they all get a WM_PAINT when unlocked), so it helps in some but not all scenarios. See also here: http://blogs.msdn.com/oldnewthing/archive/2007/02/19/1716211.aspx
peterchen
If I LockWindowUpdate() at the start of the resize/move, and try to unlock it for each WM_SIZE iteration - it causes it to be impossible to see the window moving, nor to see even a single iteration of a resize. It locks the whole window, including its frame, not just its contents. so this is effectively a dead end (for my purposes).
Mordachai
A: 

There are various approaches, but I found the only one that can be used generally is double buffering: draw to an offscreen buffer, then blit the entire buffer to screen.

That comes for free in Vista Aero and above, so your pain might be shortlived.

I am not aware of a general double-buffering implementation for windows and system controls under XP, However, here are some things to explore:

Keith Rule's CMemDC for double-buffering anything you draw yourself with GDI
WS_EX_COMPOSITED Window style (see the remarks section, and something here on stackoverflow)

peterchen
I am unclear as to how a parent window can force its child windows to draw themselves to its double buffer?
Mordachai
@Mordachai: actually this might just work. _Paints all descendants of a window in bottom-to-top painting order using double-buffering_ descendants should mean child windows. No need to do any explicit double buffering, Windows does it for you.
John Knoeller
+3  A: 

You can't prevent painting during resizing, but you can (with care) prevent repainting which is where flicker comes from. first, the bitblt.

There a two ways to stop the bitblt thing.

If you own the class of the top level window, then just register it with the CS_HREDRAW | CS_VREDRAW styles. This will cause a resize of your window to invalidate the entire client area, rather than trying to guess which bits are not going to change and bitblting.

If you don't own the class, but do have the ability to control message handling (true for most dialog boxes). The default processing of WM_NCCALCSIZE is where the class styles CS_HREDRAW and CS_VREDRAW are handled, The default behavior is to return WVR_HREDRAW | WVR_VREDRAW from processing WM_NCCALCSIZE when the class has CS_HREDRAW | CS_VREDRAW.

So if you can intercept WM_NCCALCSIZE, you can force the return of these values after calling DefWindowProc to do the other normal processing.

You can listen to WM_ENTERSIZEMOVE and WM_EXITSIZEMOVE to know when resizing of your window starts and stops, and use that to temporarily disable or modify the way your drawing and/or layout code works to minimize the flashing. What exactly you want to do to modify this code will depend on what your normal code normally does in WM_SIZE WM_PAINT and WM_ERASEBKGND.

When you paint the background of your dialog box, you need to not paint behind any of the child windows. making sure that the dialog has WS_CLIPCHILDREN solves this, so you have this handled already.

When you do move the child windows, Make sure that you use BeginDeferWindowPos / EndDefwindowPos so that all of the repainting happens at once. Otherwise you will get a bunch of flashing as each window redraws their nonclient area on each SetWindowPos call.

John Knoeller
Is there a technique for choosing a windows class when you create a modal dialog box, instead of having the OS choose the standard one? I can certainly hook the dialog box's wndproc, but is there a better way?
Mordachai
Okay, I'm not sure that this is going to help. The options available to me seem to be "preserve the window thusly (there are several alignment options), or redraw it all". I don't want windows to redraw it at all, but I don't want it to necessarily redraw it all either - just let me do it (maybe I can validate things in WM_SIZE handler to counteract the WVR_REDRAW.
Mordachai
@Mordachai: You can create a window class, and use DefDialogProc instead of DefWindowProc, but its something no-one ever does, I'm not sure why but I suspect that when you do that, subtle things go wrong.
John Knoeller
@Mordachai: I think trying to prevent redrawing is going to be doomed to fail. But you _can_ create a child window on WM_ENTERSIZEMOVE that covers all of the others, it will then be the only one that draws until you destroy it on WM_EXITSIZEMOVE.
John Knoeller
Well, I'm not sure what to make of it, but I seem to have a solution to my initial motivation: flicker free resizing. If I use DeferSetWindowPos when moving the child controls during the WM_SIZE handler, then the window resizes almost perfectly (I also force the dialog itself to have WS_CLIPCHILDREN). I still find all of this to be a bit of a black art. Why is it so difficult to manage the order that child windows redraw? Why does the DeferSetWindowPos loop succeed where manually moving the controls, then invalidating them fail? Sigh...
Mordachai
@John - I hooked WM_NCCALCSIZE, and looked at what the base proc returned - and it always returns 0 for a dialog. If I override it to return WVR_REDRAW during a resize, then the controls on the dialog "drift away"... its bizarre.
Mordachai
@Mordachai: Yes DeferWindowPos is key. I should have mentioned it. I saw the window drift thing once, I can't remember what I did to fix it. I think it has something to do with the values that are set into the buffer pointed to by lParam, but I never really investigated.
John Knoeller
@Mordachai: SetWindowPos causes non-client painting of a moved window _immediately_, while DeferWindowPos waits until all have been moved before any painting.
John Knoeller
Thanks John, that helps clarify why SetWindowPos works better than MoveWindow(...,FALSE). However, I don't understand how to suppress the bitblt behavior still. And although you seem to think that this is not possible, and its okay for the bitblt to happen, for a dialog with resizing controls - that bitblt is going to always be wrong. Its right only if the child windows don't resize, and move their origins in lock step with the parent dialog's origin, which isn't necessarily true (they may be anchored to the lower right, upper right, etc., by my layout manager).
Mordachai
@Mordachai: the bitblt is controlled by the WM_NCCALCSIZE message, which Microsoft doesn't really document sufficiently well to use fully. But you _can_ have a different origin for the bitblt by filling out the 3 rectangles in WM_NCCALCSIZE correctly. It's something I've never done personally.
John Knoeller
@Mordachai: by the way. I admire your tenacity. I think I'd enjoy working with you. :)
John Knoeller
@John - The feeling's mutual. Thanks for the help :D
Mordachai
I really didn't want to have to figure out this WM_NCCALCSIZE thing, but I seem to have stumbled on to what is almost a solution. I will post it here tomorrow in detail - but essentially if I construct the two rects nccs_params->rgrc[0] and [1], and explicitly use WVR_ALIGNLEFT|WVR_ALIGNTOP, then I get NO bitbltting (or at least its benign since its a bitblt over the top of itself, and therefor a visual no-op). But then I have to handle explicitly moving all child windows, rather than just those that aren't anchored to upper left. This looks extremely exciting! ;)
Mordachai
@Mordachai: cool. I'd love to see your final solution.
John Knoeller
+2  A: 

If I understood the question properly, it's exactly the question Raymond addressed today.

GSerg
Unfortunately, The technique Raymond describes assumes that the window has no child windows.
John Knoeller
@John: Couldn't it be applied even if there are child windows? Windows doesn't automatically know how to resize child windows. The application could choose to defer those resizes.
jamesdlin
@jamesdlin: The painting of child windows is not under control of the parent, so no, the technique doesn't work if there are child windows unless you do tricky things like hiding them when resize starts.
John Knoeller
@Jamesdlin: to be clear, you would need to defer the resize _and the repaint_ of the children to avoid flicker and/or drawing wierdness on a resize. Deferring resize only gets you part way there.
John Knoeller
A: 

there is only one way to effectively diagnose repainting problems - remote debugging.

Get a 2nd PC. Install MSVSMON on it. Add a post build step or utility project that copies your build products to the remote PC.

Now you should be able to place breakpoints in WM_PAINT handlers, WM_SIZE handlers and so on and actually trace through your dialog code as it performs the size and redraw. If you download symbols from the MS symbol servers you will be able to see full call stacks.

Some well placed breakpoints - in your WM_PAINT, WM_ERAGEBKGND handlers and you should have a good idea of why your window is being synchronously repainted early during the WM_SIZE cycle.

There are a LOT of windows in the system that consist of a parent window with layered child controls - explorer windows are massivly complicated with listviews, treeviews preview panels etc. Explorer does not have a flicker problem on resizing, so It is celarly possible to get flicker free resizing of parent windows :- what you need to do is catch the repaints, figure out what caused them, and, well, ensure that the cause is removed.

Chris Becke
Have you found that debugging drawing code is improved by having a remote machine over simply having a 2nd monitor, and running the debugger on one monitor, and the app on another? I have found them to be equivalent.
Mordachai
Also, my observations, running Windows 7 with Aero, is that many applications do the draw first as if all of the controls moved, then redraw with them in their correct position. Windows Explorer is a notable exception, showing almost no redraw lag.
Mordachai
If you try to debug window painting code on a single PC you get effects caused by the debugger taking activation each time you single step. This had a tendency to panic the window manager that would start drawing non client area's especially out of process because it thought that the process was not responding (not unjustified when its being single stepped).
Chris Becke
A: 

What appears to work:

  1. Use the WS_CLIPCHILDREN on the parent dialog (can be set in WM_INITDIALOG)
  2. During WM_SIZE, loop through the child controls moving and resizing them using DeferSetWindowPos().

This is very close to perfect, in my testing under Windows 7 with Aero.

Mordachai
Aero effectively does a lot of double-buffering for you. If you turn off Aero or try it on XP, you'll likely still have a lot of flicker.
Adrian McCarthy
Yes, it is somewhat flashier with Aero turned off. I can see more of the behavior I originally complained about, and would like to see eliminated. When I drag the resize corner of the dialog, I can see a ghosting of the OS trying to predict where things will appear (what seems most likely to be a bitblt of the former contents of my dialog), quickly overwritten by the corrected draw during my WM_SIZE handler. Its not terrible, but it still irks me that I don't seem to have a way to force the OS to NOT do the errant bitblt on my behalf. :(
Mordachai
+1  A: 

For some controls, you can use WM_PRINT message to make the control draw into a DC. But that doesn't really solve your primary problem, which is that you want Windows to NOT draw anything during resize, but to let you do it all.

And the answer is that you just can't do what you want as long as you have child windows.

The way I ended up solving this eventually in my own code is to switch to using Windowless Controls. Since they have no window of their own, they always draw at the same time (and into the same DC) as their parent window. This allows me to use simple double buffering to completely remove flicker. I can even trivially suppress painting of the children when I need to just by not calling their draw routine inside the parent's draw routine.

This is the only way I know of to completely get rid of flicker and tearing during resize operations.

John Knoeller
This is unfortunate if true. It is a ton of work - and an ongoing workload, to try to re-implement every control in Windows. "F!@#$ ton" is a reasonable label, IMO ;) Do you have a 3rd party library vendor that supplies such a thing? IMX of programming and using early Java implementations they sucked. Hard. And mostly because their windowless controls didn't get it right. They had a non-Windows feel to them, and behaved oddly enough to be a significant annoyance to us and to our users. But it blows my mind that Windows OS doesn't have a class or window flag for "don't predictive draw".
Mordachai
@Mordachai: We had already rolled our own controls for things we cared about, so it was just a matter of adding a windowless mode to them. The only control that I think is truly _hard_ to replicate is the edit control, but it was still about a programmer-years worth of control code that we already had. Have a look a the answer about WS_EX_COMPOSITED. This flag didn't exist back when I did the windowless controls so I can't say whether it will solve the problem or not, but it _sounds_ promising.
John Knoeller
The biggest downside of WS_EX_COMPOSITED that I am aware of is that you can't draw directly to the window's DC (only during WM_PAINT). Also, the solution I came up with using your other suggestion avoids the need for this for all intents and purposes anyway :)
Mordachai