Related
I have a call to a dialog box that comes up ok. But now I want to load an .png image onto the dialog box screen at design time.
I cannot seem to get it all together, and I do not have a solid understanding of how to do all this. Any help, sure be helpful.
Somewhere I read that I could move a copy of the png file onto the dialog box screen, all that appeared, when I did, was a large box with a little flower in the middle appeared.
Below is a copy of my dialog box function and function call. All I want to do is just add a png image file onto the dialog screen and have it be shown; at design time and also have it during run time. Nothing fancy, but right to the point.
INT_PTR CALLBACK DialogXProcedure(HWND hwnd, UINT uMsg, WPARAM wParam, LPARAM lParam) {
switch (uMsg) {
case WM_COMMAND: {
switch (LOWORD(wParam)) {
case IDOK:
//// just falls thru...
case IDCANCEL:
EndDialog(hwnd, LOWORD(wParam));
}
return 1;
}
case WM_INITDIALOG: {
SendMessage(hwnd, WM_SETICON, ICON_SMALL,
reinterpret_cast<LPARAM>(LoadImage(0, IDI_APPLICATION, IMAGE_ICON, 0, 0, LR_SHARED)));
///ensure focus rectangle is properly draw around control with focus
PostMessage(hwnd, WM_KEYDOWN, 0, 0);
return TRUE;
}
case WM_CLOSE:
break;
default:
return FALSE;
}
}
INT_PTR returnval = DialogBox(hInstance, MAKEINTRESOURCE(IDD_DIALOGKEPT), 0,
(DLGPROC)DialogXProcedure);
case WM_NCHITTEST: {
LRESULT hit = DefWindowProc( hWnd, uMsg, wParam, lParam );
if ( hit == HTCLIENT ) {
hit = HTCAPTION;
SendNotifyMessageA( hWnd, WM_MOUSEMOVE, NULL, MAKELPARAM( LOWORD(lParam), HIWORD(lParam) ) );
pts.x = pts.y = 0;
ClientToScreen( hWnd, &pts );
SendNotifyMessageA( hWnd, WM_EXITSIZEMOVE, NULL,NULL );
return hit;
} } break;
case WM_NCLBUTTONDOWN: { LMB_Dyn[h2] = true; } break;
I'm using this code while I try to move a window by clicking in the client area (anywhere in window). This works fine.
But I have problem correctly receiving the state of the mouse left button and while this code kind of works the problem is that when I do a simple click instead of a "click and move" the LMB_Dyn variable (mouse left button) has some delay (does not immediately get notified that I pressed the mouse button.
How to correctly get the state of the mouse left button while using WM_NCHITTEST?
Mistake: } } break; should read } return hit; }
However, there's something more radical going on here. WM_NCHITTEST is not meant to be used like that and is allowed to be called with coordinates other than the current mouse position. This means your SendNotifyMessageA can send garbage.
I see you have a handler for WM_NCLBUTTONDOWN. Maybe that's all you need. Or, if you're trying to make the window draggable by anywhere on it, that should be like so:
case WM_NCHITTEST: {
LRESULT hit = DefWindowProc( hWnd, uMsg, wParam, lParam );
if ( hit == HTCLIENT ) {
return HTCAPTION;
}
return hit;
}
The documentation for WM_NCLBUTTONDOWN and WM_NCLBUTTONUP says they fire immediately. It's really easy to forget your InvalidateRect call and so not have your screen redraw for you so it appears slow.
As the title says, I would like to move the window only when the user will drag it from a portion of the client area. This will be an imitation of the normal caption bar movement and it's because my form is custom and it doesn't have any title or caption bars. At the moment, I use the code as follows:
...
case WM_NCHITTEST:
return HTCAPTION;
and that works fine for making the user able to move the window no matter where he drags from. I would like to limit this possibility (only the top of the window will allow movement). I haven't tried checking the position of the mouse pressed because I don't know how to do it in the WM_NCHITTEST message.
I use plain Win32 (winapi) C code (no MFC or anything else at the moment) in Visual Studio 2015.
You will run into trouble if you just return HTCAPTION in response to all WM_NCHITTEST messages. You will break things like scrollbars, close buttons, resizing borders, etc. that are all implemented via different HT* values.
You have the right idea, though. You want to make the client area of your window draggable, so you need to trick Windows into thinking that your client area is actually the caption area (which, as you know, is draggable). That code looks like this:
LRESULT CALLBACK WindowProc(HWND hWnd, UINT uMsg, WPARAM wParam, LPARAM lParam)
{
// ...
case WM_NCHITTEST:
{
// Call the default window procedure for default handling.
const LRESULT result = ::DefWindowProc(hWnd, uMsg, wParam, lParam);
// You want to change HTCLIENT into HTCAPTION.
// Everything else should be left alone.
return (result == HTCLIENT) ? HTCAPTION : result;
}
// ...
}
However, based on the image in your question, you appear to want to restrict this to only a certain region of your window. You will need to define exactly what that area is, and then hit-test to see if the user has clicked in that area. Assuming that rcDraggable is a RECT structure that contains the bounds of the red box shown in your image (in screen coordinates), you can use the following code:
static RECT rcDraggable = ...
LRESULT CALLBACK WindowProc(HWND hWnd, UINT uMsg, WPARAM wParam, LPARAM lParam)
{
// ...
case WM_NCHITTEST:
{
// Call the default window procedure for default handling.
const LRESULT result = ::DefWindowProc(hWnd, uMsg, wParam, lParam);
// Get the location of the mouse click, which is packed into lParam.
POINT pt;
pt.x = GET_X_LPARAM(lParam);
pt.y = GET_Y_LPARAM(lParam);
// You want to change HTCLIENT into HTCAPTION for a certain rectangle, rcDraggable.
// Everything else should be left alone.
if ((result == HTCLIENT) && (PtInRect(&rcDraggable, pt))
{
return HTCAPTION;
}
return result;
}
// ...
}
If you define rcDraggable in terms of client coordinates, you will need to convert it to screen coordinates before doing the hit-testing in response to WM_NCHITTEST. To do that, call the MapWindowPoints function, like so:
RECT rc = rcDraggable;
MapWindowPoints(hWnd, /* a handle to your window */
NULL, /* convert to screen coordinates */
reinterpret_cast<POINT*>(&rc),
(sizeof(RECT) / sizeof(POINT)));
You can call some magic code in WM_LBUTTONDOWN handler, AFAIR this:
ReleaseCapture();
SendMessage(yourWindowHandle, WM_SYSCOMMAND, 0xf012, 0) ;
I used this method a few years ago in Delphi and Windows XP. I think it must be similar for c++. Of course, you can check x and y before doing this.
I wrote this small program which displays a list view and makes items and subitems editable.
I want to change this to make only the subitems editable. And I would like to make the list view window procedure stand on itself, that I don't have to forward WM_NOTIFY messages every time as I'm doing now in WndProcMain. And the purpose is that I don't use only one list view with editable subitems in my program, I'm going to use it in many different windows.
The LVN_ENDLABELEDIT notification is processed by WndProcList because the bEditing has to be changed. This flag is used for WM_PAINT when subitems have to be edited. This is a fix, otherwise the text in the first subitem disappears because it thinks the first item is being edited. However, I would like to also receive a message like LVN_ENDLABELEDIT in the window procedure of the list view owner window (in this case WndProcMain), because I want to manipulate the user input also.
Please ask if you have questions.
Thanks in advance
Midas
WNDPROC wpOrigEditProc;
RECT rcSubItem;
LRESULT CALLBACK WndProcEditList(HWND hWnd, UINT uMsg, WPARAM wParam, LPARAM lParam) {
switch (uMsg) {
case WM_WINDOWPOSCHANGING:
{
WINDOWPOS *pos = (WINDOWPOS*) lParam;
pos->x = rcSubItem.left;
pos->cx = rcSubItem.right;
}
break;
default:
return CallWindowProc(wpOrigEditProc, hWnd, uMsg, wParam, lParam);
}
return 1;
}
LRESULT CALLBACK WndProcList(HWND hWnd, UINT uMsg, WPARAM wParam, LPARAM lParam) {
static HWND hEdit;
static RECT rc;
static LVITEM lvI;
static unsigned char bEditing = 0;
switch (uMsg) {
case WM_NOTIFY:
switch (((NMHDR*) lParam)->code) {
case NM_CLICK:
lvI.iItem = ((NMITEMACTIVATE*) lParam)->iItem;
lvI.iSubItem = ((NMITEMACTIVATE*) lParam)->iSubItem;
break;
case NM_DBLCLK:
SendMessage(hWnd, LVM_EDITLABEL, lvI.iItem, 0);
break;
case LVN_BEGINLABELEDIT:
{
char text[32];
bEditing = 1;
hEdit = (HWND) SendMessage(hWnd, LVM_GETEDITCONTROL, 0, 0);
rcSubItem.top = lvI.iSubItem;
rcSubItem.left = LVIR_LABEL;
SendMessage(hWnd, LVM_GETSUBITEMRECT, lvI.iItem, (long) &rcSubItem);
rcSubItem.right = SendMessage(hWnd, LVM_GETCOLUMNWIDTH, lvI.iSubItem, 0);
wpOrigEditProc = (WNDPROC) SetWindowLong(hEdit, GWL_WNDPROC, (long) WndProcEditList);
lvI.pszText = text;
lvI.cchTextMax = 32;
SendMessage(hWnd, LVM_GETITEMTEXT, lvI.iItem, (long) &lvI);
SetWindowText(hEdit, lvI.pszText);
}
break;
case LVN_ENDLABELEDIT:
bEditing = 0;
SetWindowLong(hEdit, GWL_WNDPROC, (long) wpOrigEditProc);
if (!lvI.iSubItem) return 1;
lvI.pszText = ((NMLVDISPINFO*) lParam)->item.pszText;
if (!lvI.pszText) return 1;
SendMessage(hWnd, LVM_SETITEMTEXT, lvI.iItem, (long) &lvI);
break;
default:
return CallWindowProc((WNDPROC) GetClassLong(hWnd, GCL_WNDPROC), hWnd, uMsg, wParam, lParam);
}
break;
case WM_PAINT:
if (bEditing) {
RECT rcItem;
if (lvI.iSubItem > 0) {
rcItem.left = LVIR_LABEL;
if (SendMessage(hWnd, LVM_GETITEMRECT, lvI.iItem, (long) &rcItem))
ValidateRect(hWnd, &rcItem);
}
}
default:
return CallWindowProc((WNDPROC) GetClassLong(hWnd, GCL_WNDPROC), hWnd, uMsg, wParam, lParam);
}
return 0;
}
LRESULT CALLBACK WndProcMain(HWND hWnd, UINT uMsg, WPARAM wParam, LPARAM lParam) {
static HWND hList;
static RECT rc;
switch (uMsg) {
case WM_NOTIFY:
switch (((NMHDR*) lParam)->code) {
case NM_CLICK:
case NM_DBLCLK:
case LVN_BEGINLABELEDIT:
case LVN_ENDLABELEDIT:
return CallWindowProc(WndProcList, ((NMHDR*) lParam)->hwndFrom, uMsg, wParam, lParam);
}
break;
case WM_CREATE:
{
LVCOLUMN lvc;
LVITEM lvI;
unsigned int i;
float vertex;
char text[32];
hList = CreateWindow(WC_LISTVIEW, 0, WS_CHILD | WS_VISIBLE | LVS_REPORT | LVS_EDITLABELS, rc.left, rc.top, rc.right, rc.bottom, hWnd, 0, hInstance, 0);
SendMessage(hList, LVM_SETEXTENDEDLISTVIEWSTYLE, 0, LVS_EX_FULLROWSELECT | LVS_EX_GRIDLINES);
SetWindowLong(hList, GWL_WNDPROC, (long) WndProcList);
lvc.mask = LVCF_WIDTH;
lvc.cx = 30;
SendMessage(hList, LVM_INSERTCOLUMN, 0, (LPARAM) &lvc);
lvc.mask = LVCF_TEXT;
lvc.pszText = "Vertex";
SendMessage(hList, LVM_INSERTCOLUMN, 1, (LPARAM) &lvc);
SendMessage(hList, LVM_SETCOLUMNWIDTH, 1, LVSCW_AUTOSIZE_USEHEADER);
lvI.mask = LVIF_TEXT;
lvI.pszText = text;
for (i = 0; i < 10; i++) {
vertex = (float) i;
lvI.iItem = i;
lvI.iSubItem = 0;
sprintf(text, "%d", i);
SendMessage(hList, LVM_INSERTITEM, 0, (LPARAM) &lvI);
lvI.iSubItem = 1;
sprintf(text, "%f, %f, %f", vertex - 1, vertex, vertex + 1);
SendMessage(hList, LVM_SETITEM, 0, (LPARAM) &lvI);
}
}
break;
case WM_SIZE:
GetClientRect(hWnd, &rc);
MoveWindow(hList, rc.left, rc.top, rc.right, rc.bottom, 1);
SendMessage(hList, LVM_SETCOLUMNWIDTH, 1, LVSCW_AUTOSIZE_USEHEADER);
break;
case WM_DESTROY:
PostQuitMessage(0);
break;
default:
return DefWindowProc(hWnd, uMsg, wParam, lParam);
}
return 0;
}
I posted an answer related to the question but it was in C#. Never had much of expertise in winapi rather than playing with it as a hobbyist long ago, The answer at the end of this post looks promising -- http://cboard.cprogramming.com/windows-programming/122733-%5Bc%5D-editing-subitems-listview-win32-api.html
The first problem that cropped up for me in trying to compile your code is the fact that you either aren't compiling for Unicode, or you're using the non-Unicode sprintf function to format text. That's the first thing that needs to be fixed: Windows applications have been fully Unicode for over a decade. Replace every instance of char variable declarations with wchar_t (or TCHAR), prefix your string literals with L (or surround them with the TEXT() macro), and quickly replace the calls to sprintf with calls to wsprintf. As the documentation indicates, there are certainly better functions to use than wsprintf, but the same is true for sprintf, and this gets the code to compile with minimum effort.
The other thing that looks non-idiomatic to me is your use of the Get/SetClassLong and Get/SetWindowLong functions. Nowadays, I always write code with 64-bit portability in mind, and therefore I'd replace those with the Get/SetClassLongPtr and Get/SetWindowLongPtr macros, which automatically resolve to the correct function call, depending on if you're compiling for x86 or x64. This isn't a deal-breaker, though.
You first asked if there was a way to handle the WM_NOTIFY message directly from the sub-classed control by forwarding them automatically. Unfortunately, this is not possible. The Win32 model is such that parents always own their children, and it is thus their responsibility to handle events. I agree with your intuition on the separation of concerns, but the only way to make this happen is to explicitly forward the message from the parent to the appropriate child control yourself. Frameworks like MFC (which encapsulate the Win32 API) do this for you apparently automatically, still have to forward the notification messages from the parent to the child. They do it using something called "message reflection", which you can read about here. There's nothing stopping you from implementing something similar in your own application, but at a certain point you have to stop and ask yourself if it isn't worth using one of the many available GUI frameworks just for this sort of thing.
Anyway, as I understand it, the main thrust of your question is this:
I want to change this to make only the subitems editable.
That seems like a pretty simple fix. All you have to do is check in the LVN_BEGINLABELEDIT notification message handler that the user has, in fact, requested to edit a sub-item. Since you've used it elsewhere in your code, you know that the LVITEM.iSubItem member gives you either the one-based index of the sub-item, or 0 if the structure refers to an item rather than a sub-item.
So insert this line to ensure that lvI.iSubItem is not equal to 0 at the top of the LVN_BEGINLABELEDIT handler:
if (lvI.iSubItem == 0) return TRUE; // prevent editing
As the documentation for the LVN_BEGINLABELEDIT message indicates, returning FALSE allows the user to edit the label, and returning TRUE prevents them from editing. Since we return TRUE, we prevent the edit of anything but sub-items before the edit even starts.
It looks to me like you already tried to do something similar in the LVN_ENDLABELEDIT notification message handler with this line:
if (!lvI.iSubItem) return 1;
but that's too late! If the edit is already ending, then you've already given the user the impression that they were able to edit the main item, which you don't want to do. Take that line out, and it should work as expected.
Note that your implementation does have at least one glaring bug: you don't prevent the user from modifying the contents of a sub-item to a string longer than 32 characters, but your code populating the editing control only accepts a string up to 32 characters long:
TCHAR text[32];
// ... snip ...
lvI.pszText = text;
lvI.cchTextMax = 32;
SendMessage(hWnd, LVM_GETITEMTEXT, lvI.iItem, (long) &lvI);
SetWindowText(hEdit, lvI.pszText);
Writing that code the correct way would be (and I suspect that's why you haven't done it the right way) a giant pain in the ass. Typically, I will create a string buffer that I think is long enough, try to get the text of the sub-item, and check the return value of the LVM_GETITEMTEXT message. The return value tells me how many characters were copied into the string buffer. If the number of characters copied indicates that it entirely filled the available space in the string buffer, I'll make the buffer larger (perhaps doubling the size), and then try to send the LVM_GETITEMTEXT message again. As I recall, MFC does something similar. Told you it was a pain, but it's worth getting these things right.
A simpler solution (although more limiting) would be to prevent the user from ever setting the length of one of the sub-items to a string of text longer than 32 characters. Then you wouldn't have to worry about handling long input because you'd know it would never be there, and the user would never be confused about the behavior of your control. To do that, send the edit control an EM_LIMITTEXT message at the end of your LVN_BEGINLABELEDIT handler:
case LVN_BEGINLABELEDIT:
{
// ... snip ...
lvI.cchTextMax = 32;
SendMessage(hWnd, LVM_GETITEMTEXT, lvI.iItem, (long) &lvI);
SetWindowText(hEdit, lvI.pszText);
// (begin new code)
SendMessage(hEdit, EM_LIMITTEXT, lvI.cchTextMax, 0);
}
Now the user can't enter more than the allowed number of characters, so your code knows that you will never have to deal with any more than that being there (unless you write code to place them there yourself, in which case...).
All of that said, I think I agree with Hans:
Ugh, you'll fight glitches forever. There's little point with grid controls universally available.
I've got a sizable dialog with one child window - a list control. When the dialog is re-sized, I re-size the list control appropriately; it is basically anchored to all 4 edges of the dialog. The problem is that during sizing there is noticeable flicker around the edges of the list control, especially when the scroll bars are present. I am a novice in Win32 GUI stuff, so I don't really know how to handle this. I've seen a lot of articles about flicker-free drawing, but they are all about individual custom-drawn controls and none of them deal with flicker-free drawing of a dialog as a whole. How can I get this to work without flickering so much?
My actual dialog has multiple controls obviously, but here is a minimal code example that reproduces the issue (IDC_LIST1 is a list control in Report view, IDD_DIALOG2 has WS_CLIPCHILDREN style set).
#define NUM_COLUMNS 8
#define NUM_ROWS 32
RECT rcDialog2WindowOriginal;
RECT rcDialog2ClientOriginal;
RECT rcList1ClientOriginal;
INT_PTR Dialog2_OnInitDialog(HWND hDlg, WPARAM wParam, LPARAM lParam)
{
GetWindowRect(hDlg, &rcDialog2WindowOriginal);
GetClientRect(hDlg, &rcDialog2ClientOriginal);
GetWindowRect(GetDlgItem(hDlg, IDC_LIST1), &rcList1ClientOriginal);
ScreenToClient(hDlg, ((LPPOINT)&rcList1ClientOriginal));
ScreenToClient(hDlg, ((LPPOINT)&rcList1ClientOriginal) + 1);
SendDlgItemMessage(hDlg, IDC_LIST1, LVM_SETEXTENDEDLISTVIEWSTYLE, LVS_EX_FULLROWSELECT, LVS_EX_FULLROWSELECT);
TCHAR szText[32];
// add some columns
LVCOLUMN col;
ZeroMemory(&col, sizeof(LVCOLUMN));
col.mask = LVCF_SUBITEM | LVCF_TEXT | LVCF_WIDTH;
col.cx = 60;
col.pszText = szText;
for(int i = 0; i < NUM_COLUMNS; i++)
{
col.iSubItem = i;
_stprintf_s(szText, 32, _T("Column %d"), col.iSubItem);
SendDlgItemMessage(hDlg, IDC_LIST1, LVM_INSERTCOLUMN, col.iSubItem, LPARAM)&col);
}
// add some items
LVITEM item;
ZeroMemory(&item, sizeof(LVITEM));
item.mask = LVIF_TEXT;
item.pszText = szText;
for(int i = 0; i < NUM_ROWS; i++)
{
item.iItem = i;
for(int j = 0; j < NUM_COLUMNS; j++)
{
item.iSubItem = j;
_stprintf_s(szText, 32, _T("Item %d, SubItem %d"), i, j);
if(j == 0)
{
SendDlgItemMessage(hDlg, IDC_LIST1, LVM_INSERTITEM, 0, (LPARAM)&item);
}
else
{
SendDlgItemMessage(hDlg, IDC_LIST1, LVM_SETITEM, 0, (LPARAM)&item);
}
}
}
// autosize the columns
for(int i = 0; i < NUM_COLUMNS; i++)
{
SendDlgItemMessage(hDlg, IDC_LIST1, LVM_SETCOLUMNWIDTH, i, LVSCW_AUTOSIZE);
}
return TRUE;
}
INT_PTR Dialog2_OnGetMinMaxInfo(HWND hDlg, WPARAM wParam, LPARAM lParam)
{
LPMINMAXINFO lpMinMaxInfo = (LPMINMAXINFO)lParam;
// don't allow dialog to be resized smaller than original size
lpMinMaxInfo->ptMinTrackSize.x = rcDialog2WindowOriginal.right - rcDialog2WindowOriginal.left;
lpMinMaxInfo->ptMinTrackSize.y = rcDialog2WindowOriginal.bottom - rcDialog2WindowOriginal.top;
return TRUE;
}
INT_PTR Dialog2_OnSize(HWND hDlg, WPARAM wParam, LPARAM lParam)
{
int cx = LOWORD(lParam);
int cy = HIWORD(lParam);
// anchor the list control to all edges of the dialog
int left_delta = rcList1ClientOriginal.left - rcDialog2ClientOriginal.left;
int right_delta = rcDialog2ClientOriginal.right - rcList1ClientOriginal.right;
int top_delta = rcList1ClientOriginal.top - rcDialog2ClientOriginal.top;
int bottom_delta = rcDialog2ClientOriginal.bottom - rcList1ClientOriginal.bottom;
int left = left_delta;
int top = top_delta;
int width = cx - left_delta - right_delta;
int height = cy - top_delta - bottom_delta;
HWND hWndList1 = GetDlgItem(hDlg, IDC_LIST1);
SetWindowPos(hWndList1, NULL, left, top, width, height, SWP_NOZORDER);
return TRUE;
}
INT_PTR Dialog2_OnClose(HWND hDlg, WPARAM wParam, LPARAM lParam)
{
EndDialog(hDlg, IDOK);
return TRUE;
}
INT_PTR CALLBACK Dialog2_DialogProc(HWND hDlg, UINT nMsg, WPARAM wParam, LPARAM lParam)
{
switch(nMsg)
{
case WM_INITDIALOG:
return Dialog2_OnInitDialog(hDlg, wParam, lParam);
case WM_GETMINMAXINFO:
return Dialog2_OnGetMinMaxInfo(hDlg, wParam, lParam);
case WM_SIZE:
return Dialog2_OnSize(hDlg, wParam, lParam);
case WM_CLOSE:
return Dialog2_OnClose(hDlg, wParam, lParam);
}
return FALSE;
}
Update
After taking a look at many other Windows applications (even ones written by Microsoft), every single one of them suffers the same flickering issues. It is particularly evident when resizing a window with both a status bar and scroll bar from the top left. I guess I will just have to live with it.
A lot of flicker comes from WM_ERASEBKGRD, handle this message and just return TRUE;
You can do several things.
Handle WM_ERASE yourself. You may still need to erase some of the background, but you can draw 4 rectangles around the edges of the window to fill the background area of the dialog without overdrawing the rectangle in the middle where your child control sits.
After calling SetWindowPos, try calling UpdateWindow() to immediately repaint the window (try it on both the list view control and your own window). This minimises the time between you making the size change and the redraw being completed.
Another approach is double buffering (you draw the entire window surface to an offscreen bitmap, and only draw it to the screen when it's all finished. This minimises the time between the start and end of the update proicess on screen, so reduces flicker. This is harder to achieve with a dialog and a windows control though, so not really applicable in your case).
In general, don't be afraid to experiment with things. Try doing things in different orders, etc and see what the results are. If something (like UpdateWindow()) doesn't give a noticeable improvement, then it's easy to remove the code again, and try something else until you get the best results.
edit - additional ideas
see also this SO question
When updating controls you can also suspend and resume repainting (BeginUpdate() and EndUpdate()) to stop them drawing more than once when you are adding many items, etc. THis is not likely to help with window resizing though.
After taking a look at many other Windows applications (even ones written by Microsoft), every single one of them suffers the same flickering issues. It is particularly evident when resizing a window with both a status bar and scroll bar from the top left. I guess I will just have to live with it.
If you try to replace any window's default windowsproc with the following code, the window will basically be updating zero pixels on the screen (besides the border and the title bar; see WM_NCPAINT and other NC-messages for that).
LRESULT CALLBACK WindowProc(HWND hwnd, UINT uMsg, WPARAM wParam, LPARAM lParam)
{
switch (uMsg)
{
case WM_ERASEBKGND:
// Lie and tell Windows that this window has cleared its background.
return (-1);
case WM_PAINT:
// Lie and tell Windows that this window has completed all its painting.
ValidateRect(hwnd, nullptr);
return 0;
default:
return DefWindowProc(hwnd, uMsg, wParam, lParam);
}
}
This tells us that you are free to paint the pixels however you like. With or without flickering. It's up to you.
How?
Create a compatible bitmap (CreateCompatibleBitmap) and select it into the device context of the window within WM_PAINT. When you're done drawing, BitBlt() that bitmap to the actual window DC and voila; flicker free rendering.