Win32 C++ Alphablend a Bitmap Partially Transparent - c++

I've googled, seen examples, other questions here, MSDN and Downloaded Example code. I cannot figure out what is wrong with this.
// setting up the memory DC and selecting in the bitmap
HDC hdc = GetDC(hWnd);
HDC hdcMem = CreateCompatibleDC(hdc);
ReleaseDC(hWnd, hdc);
HBITMAP hbmOld = (HBITMAP)SelectObject(hdcMem, bitmap.hbmLogo);
// setting up the blend function
BLENDFUNCTION bStruct;
bStruct.BlendOp = AC_SRC_OVER;
bStruct.BlendFlags = 0;
bStruct.SourceConstantAlpha = 255;
bStruct.AlphaFormat = AC_SRC_ALPHA;
// try
BOOL check = AlphaBlend(buffer.getBufferDC(), 0, 0, bitmap.bmLogo.bmWidth, bitmap.bmLogo.bmHeight, hdcMem, 0, 0, bitmap.bmLogo.bmWidth, bitmap.bmLogo.bmHeight, bStruct);
if (check == FALSE) MessageBox(0,0,0,0);
// this is how I load the bitmap, it is a resource.
bitmap.hbmLogo = LoadBitmap(GetModuleHandle(NULL), MAKEINTRESOURCE(IDB_LOGO_0));
if (bitmap.hbmLogo == NULL) { MessageBox(NULL, "Could not read the logo bitmap.", "Error", MB_OK); return false; }
GetObject(bitmap.hbmLogo, sizeof(bitmap.bmLogo), &bitmap.bmLogo);
I use the message box to quickly check the result. Check always returns TRUE. The bitmap and its dimensions are correct.
I've tried it over different background colors, alpha values, and still nothing, replacing that with BitBlt or TransparentBitBlt, no problem, the logo displays. All my attempts with the AlphaBlend function has resulted in no change. The logo does not appear, even for a second, on the screen.
Any ideas?
Thanks.

Found the solution after looking closer at an example.
I set the BLENDFUNCTION as a global, and in the WM_CREATE message I used:
m_bf.BlendOp = AC_SRC_OVER;
m_bf.BlendFlags = 0;
m_bf.SourceConstantAlpha = 100; // any 0 to 255
m_bf.AlphaFormat = 0;
LoadBitmapsFromResource();
and it is now working.

Related

How to draw text with transparency using GDI?

My goal is to dynamically put some arbitrary text into an HICON image (at runtime.) I'm using the following code:
//Error checks are omitted for brevity
//First create font
LOGFONT lf = {0};
lf.lfHeight = -58;
lf.lfWeight = FW_NORMAL;
lf.lfOutPrecision = OUT_TT_PRECIS; //Use TrueType fonts for anti-alliasing
lf.lfQuality = CLEARTYPE_QUALITY;
lstrcpy(lf.lfFaceName, L"Segoe UI");
HFONT hFont = ::CreateFontIndirect(&lf);
//HICON hIcon = original icon to use as a source
//I'm using a large 256x256 pixel icon
hIcon = (HICON)::LoadImage(theApp.m_hInstance, MAKEINTRESOURCE(IDI_ICON_GREEN_DIAMOND), IMAGE_ICON, 256, 256, LR_DEFAULTCOLOR);
ICONINFO ii = {0};
::GetIconInfo(hIcon, &ii);
BITMAP bm = {0};
::GetObject(ii.hbmColor, sizeof(bm), &bm);
SIZE szBmp = {bm.bmWidth, bm.bmHeight};
HDC hDc = ::GetDC(hWnd);
HDC hMemDC = ::CreateCompatibleDC(hDc);
HGDIOBJ hOldBmp = ::SelectObject(hMemDC, ii.hbmColor);
HGDIOBJ hOldFont = ::SelectObject(hMemDC, hFont);
::SetBkMode(hMemDC, TRANSPARENT);
::SetTextColor(hMemDC, RGB(255, 0, 0)); //Red text
//Draw text
//NOTE that DrawText API behaves in a similar way
::TextOut(hMemDC, 0, 0, L"Hello", 5);
::SelectObject(hMemDC, hOldFont);
::SelectObject(hMemDC, hOldBmp);
//We need a simple mask bitmap for the icon
HBITMAP hBmpMsk = ::CreateBitmap(szBmp.cx, szBmp.cy, 1, 1, NULL);
ICONINFO ii2 = {0};
ii2.fIcon = TRUE;
ii2.hbmColor = ii.hbmColor;
ii2.hbmMask = hBmpMsk;
//Create updated icon
HICON hIcon2 = ::CreateIconIndirect(&ii2);
//Cleanup
::DeleteObject(hBmpMsk);
::DeleteDC(hMemDC);
::ReleaseDC(hWnd, hDc);
::DeleteObject(ii.hbmColor);
::DeleteObject(ii.hbmMask);
::DeleteObject(hFont);
and then I can display the icon in my window from OnPaint() handler (so that I can see how it turns out) as such:
::DrawIconEx(dc.GetSafeHdc(), 0, 0,
hIcon2,
256, 256, NULL,
::GetSysColorBrush(COLOR_BTNFACE),
DI_NORMAL);
So here's what I get:
To see what's going on pixel-wise in my hIcon2 I called GetDIBits on its ii.hbmColor from the code above. The resulting pixel array where my word "Hello" was supposed to be shown looked like this:
The pixels are encoded as BGRA in that memory dump, so the 4th byte in each DWORD stands for transparency: 0=transparent, FF=opaque. But in this case TextOut doesn't fill out transparency, or leaves it as 0, which is interpreted as "fully transparent." Instead it seems to pre-multiply it into the RGB colors themselves.
Note that if I keep looking further down the same bitmap, where the green diamond begins, the image pixels seem to have transparency bytes set correctly:
Any idea how to draw text so that the API could set those transparency bytes?
EDIT: As was suggested below I tried the following GDI+ method:
HGDIOBJ hOldBmp = ::SelectObject(hMemDC, ii.hbmColor);
Graphics grpx(hMemDC);
RectF rcfTxt(0.0f, 0.0f, (REAL)szBmp.cx, (REAL)szBmp.cy);
Font gdiFont(L"Segoe UI", 58.0f, FontStyleRegular, UnitPixel);
SolidBrush gdiBrush(Color(255, 0, 0));
StringFormat gdiSF;
gdiSF.SetAlignment(StringAlignmentNear);
gdiSF.SetFormatFlags(StringFormatFlagsNoWrap);
gdiSF.SetHotkeyPrefix(HotkeyPrefixNone);
//The reason I was using GDI was because I was setting
//spacing between letters using SetTextCharacterExtra()
//Unfortunately with GDI+ this does not work!
HDC hTmpDC = grpx.GetHDC();
::SetTextCharacterExtra(hTmpDC, -4); //This doesn't do anything!
grpx.ReleaseHDC(hTmpDC);
grpx.DrawString(L"Hello", 5, &gdiFont, rcfTxt, &gdiSF, &gdiBrush);
::SelectObject(hMemDC, hOldBmp);
and besides not being able to set character spacing (which I could with GDI using SetTextCharacterExtra) here's what I got (slightly enlarged for visibility):
So clearly still an issue with transparency.
Taken from an old post by Microsoft MVP Mike D Sutton here.
When you create a DC it initially has default 'stock' objects selected
into it, including the stock 1*1*1 Bitmap. Since there is a Bitmap
already selected into the DC when you call DrawText() it will still
try and render to it even though pretty much everything (apart from
one pixel) will be clipped.
What you need to do is to create a Bitmap,
either DDB or DIBSection, and select that into your DC before drawing
to it.
First though you need to find the size of your Bitmap since you
want it large enough to display your text in, so for that you use the
DrawText() call again on the initial DC but include the DT_CALCRECT
flag. What this does is rather than drawing anything it simply
measures how large the text is and dumps that into the RECT you pass
the call. From here you can go ahead and create your DIBSection using
those dimensions and select it into your DC. Finally perform your
existing DrawText ()call (you may also want to use SetBkMode/Color())
which will render the text to the DIBSection from which you can get at
the data.
This seems to work pretty well here:
HBITMAP CreateAlphaTextBitmap(LPCSTR inText, HFONT inFont, COLORREF inColour) {
int TextLength = (int)strlen(inText);
if (TextLength <= 0) return NULL;
// Create DC and select font into it
HDC hTextDC = CreateCompatibleDC(NULL);
HFONT hOldFont = (HFONT)SelectObject(hTextDC, inFont);
HBITMAP hMyDIB = NULL;
// Get text area
RECT TextArea = {0, 0, 0, 0};
DrawText(hTextDC, inText, TextLength, &TextArea, DT_CALCRECT);
if ((TextArea.right > TextArea.left) && (TextArea.bottom > TextArea.top)) {
BITMAPINFOHEADER BMIH;
memset(&BMIH, 0x0, sizeof(BITMAPINFOHEADER));
void *pvBits = NULL;
// Specify DIB setup
BMIH.biSize = sizeof(BMIH);
BMIH.biWidth = TextArea.right - TextArea.left;
BMIH.biHeight = TextArea.bottom - TextArea.top;
BMIH.biPlanes = 1;
BMIH.biBitCount = 32;
BMIH.biCompression = BI_RGB;
// Create and select DIB into DC
hMyDIB = CreateDIBSection(hTextDC, (LPBITMAPINFO)&BMIH, 0, (LPVOID*)&pvBits, NULL, 0);
HBITMAP hOldBMP = (HBITMAP)SelectObject(hTextDC, hMyDIB);
if (hOldBMP != NULL) {
// Set up DC properties
SetTextColor(hTextDC, 0x00FFFFFF);
SetBkColor(hTextDC, 0x00000000);
SetBkMode(hTextDC, OPAQUE);
// Draw text to buffer
DrawText(hTextDC, inText, TextLength, &TextArea, DT_NOCLIP);
BYTE* DataPtr = (BYTE*)pvBits;
BYTE FillR = GetRValue(inColour);
BYTE FillG = GetGValue(inColour);
BYTE FillB = GetBValue(inColour);
BYTE ThisA;
for (int LoopY = 0; LoopY < BMIH.biHeight; LoopY++) {
for (int LoopX = 0; LoopX < BMIH.biWidth; LoopX++) {
ThisA = *DataPtr; // Move alpha and pre-multiply with RGB
*DataPtr++ = (FillB * ThisA) >> 8;
*DataPtr++ = (FillG * ThisA) >> 8;
*DataPtr++ = (FillR * ThisA) >> 8;
*DataPtr++ = ThisA; // Set Alpha
}
}
// De-select bitmap
SelectObject(hTextDC, hOldBMP);
}
}
// De-select font and destroy temp DC
SelectObject(hTextDC, hOldFont);
DeleteDC(hTextDC);
// Return DIBSection
return hMyDIB;
}
If you need an example of how to call it then try something like this
(inDC is the DC to render to):
void TestAlphaText(HDC inDC, int inX, int inY) {
const char *DemoText = "Hello World!\0";
RECT TextArea = {0, 0, 0, 0};
HFONT TempFont = CreateFont(50, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, "Arial\0");
HBITMAP MyBMP = CreateAlphaTextBitmap(DemoText, TempFont, 0xFF);
DeleteObject(TempFont);
if (MyBMP) { // Create temporary DC and select new Bitmap into it
HDC hTempDC = CreateCompatibleDC(inDC);
HBITMAP hOldBMP = (HBITMAP)SelectObject(hTempDC, MyBMP);
if (hOldBMP) {
BITMAP BMInf; // Get Bitmap image size
GetObject(MyBMP, sizeof(BITMAP), &BMInf);
// Fill blend function and blend new text to window
BLENDFUNCTION bf;
bf.BlendOp = AC_SRC_OVER;
bf.BlendFlags = 0;
bf.SourceConstantAlpha = 0x80;
bf.AlphaFormat = AC_SRC_ALPHA;
AlphaBlend(inDC, inX, inY, BMInf.bmWidth, BMInf.bmHeight,
hTempDC, 0, 0, BMInf.bmWidth, BMInf.bmHeight, bf);
// Clean up
SelectObject(hTempDC, hOldBMP);
DeleteObject(MyBMP);
DeleteDC(hTempDC);
}
}
}
All credit to answer and code go to original posters on that forum, I've simply reposted it so that this answer will be valid if the links die.
This reply is coming almost 3 years after the question was posted, but people still consult these things long into the future. So I'll explain what's happening.
DrawText (and other GDI text functions) will work on a transparent bitmap. The text is not coming out black even though it displays that way. The alpha channel is set to 0 on all pixels the text draws to, overriding whatever alpha you had set previously. If you set an alpha value in SetTextColor the text will render all black. If you're feeling ambitious you can run through pixel by pixel and target anything not your fill color (which requires a single fill color) but the problem then becomes one of the nature of ClearType being overridden and all alphas are set to whatever you set them to. The text ends up looking very funky. If you use a constant alpha for your background fill you can simply do a blanket run across the entire bitmap's bits after the text is drawn and reset all the alpha values. Since you have to read a byte to determine if it's background or not, you might as well just set every pixel's alpha to whatever the standard alpha is for that image and bypass the slow compares. This works reasonably well and I've found it to be very acceptable. In this day and age, MS should have taken care of this long ago but it's not to be.
https://learn.microsoft.com/en-us/windows/win32/gdiplus/-gdiplus-antialiasing-with-text-use
Gdiplus::Bitmap bmp( your_Width, your_Height, PixelFormat64bppARGB);
//PixelFormat64bppARGB ARGB needed
FontFamily fontFamily(L"Arial");
Font font(&fontFamily, 29, FontStyleRegular, UnitPoint);
Gdiplus::RectF rectF(00.0f, 10.0f, your_Width, your_Height);
StringFormat stringFormat;
SolidBrush solidBrush(Color(63, 0, 0, 255));
stringFormat.SetAlignment(StringAlignmentCenter);
//solidBrush Color(63, 0, 0, 255) ARGB neede
graphics.SetTextRenderingHint(TextRenderingHintAntiAlias);
graphics.DrawString("your_text", -1, &font, rectF, &stringFormat, &solidBrush);
//TextRenderingHintAntiAlias this needed

SelectObject returns NULL with hbitmap created in constructor

I have a bitmap class that has a load function for loading the bitmap from either file path or resource ID. This part works fine.
void GtBitmap::Load()
{
LPTSTR szFileName;
szFileName = (LPTSTR)m_strPath.c_str();
// Check for valid .BMP file path
if (m_strPath.size() > 0)
{
// Open .BMP file
m_pFile = fopen(m_strPath.c_str(), ("rb"));
if (m_pFile != NULL)
{
m_hBitmap = (HBITMAP)LoadImage (GetModuleHandle(NULL), szFileName, IMAGE_BITMAP, 0, 0, LR_SHARED | LR_LOADFROMFILE);
GetObject(m_hBitmap, sizeof(m_bmap), &m_bmap);
int i = 1;
}
}
else if (m_intResourceID != 0)
{
m_hBitmap = (HBITMAP)LoadImage(GetModuleHandle(NULL), MAKEINTRESOURCE(m_intResourceID), IMAGE_BITMAP, 0, 0, LR_SHARED);
GetObject(m_hBitmap, sizeof(m_bmap), &m_bmap);
int i = 1;
}
}
However, when I try to render it in my code block, the SelectObject returns null. Here is the code for that section of the painter class.
void GtPainterGDI::GtDrawBitmap(GtRectI & target, GtBitmap & bitmap, bool blnOffset)
{
GtCanvas topCv = m_arrCanvas.back();
HDC hdcMem = CreateCompatibleDC(topCv.m_hdcParent);
HBITMAP hbmOld = (HBITMAP)SelectObject(hdcMem, bitmap.m_hBitmap);
DWORD lastError = GetLastError();
bool success = BitBlt(hdcMem, target.GetLeft(), target.GetTop(),
target.Width(), target.Height(), hdcMem, 0, 0, SRCCOPY);
SelectObject(hdcMem, bitmap.m_hBitmap);
DeleteDC(hdcMem);
};
The SelectObject() returns null and the image is not drawn. I can only get the image to show up if I use a LoadImage() in that paint function. However I don't want to load the image every time I want to paint. I should be able to load the image once in the Load function or constructor of the bitmap, then use the handle in the paint function.
If anyone could please provide an example of loading an image in a constructor and then painting it elsewhere in the codes WM_PAINT or equivalent painting function I would appreciate it. The code is a new version of the GT graphical user interface library. I plan on posting a new version on codeproject in the next few days or so. I have to do some cleanup first...
Thanks in advance.
HINSTANCE parameter in LoadImage should be NULL when loading the image from file. Use GetModuleHandle(NULL) only when loading from resource.
m_hBitmap = (HBITMAP)LoadImage(NULL, m_strPath.c_str(),
IMAGE_BITMAP, 0, 0, LR_LOADFROMFILE);
if (!m_hBitmap)
{
//report error
}
Also LR_SHARED is not necessary here.
When testing for file's exist, you can use std::ifstream. Example:
#include <fstream>
...
bool test = std::ifstream(m_strPath).good();
This will test for file and close the file handle right away.
Make sure to select hbmOld before deleting hdcMem:
HBITMAP hbmOld = (HBITMAP)SelectObject(hdcMem, bitmap.m_hBitmap);
BitBlt(...)
//SelectObject(hdcMem, bitmap.m_hBitmap); <<= remove this
SelectObject(hdcMem, hbmOld);
DeleteDC(hdcMem);

DrawIconEx leaving mask artifacts

I'm extracting jumbo icons for any given path using IImageList and SHGetFileInfo. Once I have that, I then render the HICON into a HBITMAP using DrawIconEx for eventual rendering with GDI+ Bitmap and Graphics objects.
Now, this all works great, except that when I do the final rendering of the bitmap, the very left edge always has a black artifact on it. This is true for pretty much any icon I get, and is always the left edge.
Any ideas where the dark line could be coming from?
The code I'm using is roughly:
1. Extract Icon:
// Get the image list index of the icon
SHFILEINFO sfi;
if (!SHGetFileInfo(pszPath, 0, &sfi, sizeof(sfi), SHGFI_SYSICONINDEX)) return NULL;
// Get the jumbo image list
IImageList *piml;
if (FAILED(SHGetImageList(SHIL_JUMBO, IID_PPV_ARGS(&piml)))) return NULL;
// Extract an icon
HICON hicon;
piml->GetIcon(sfi.iIcon, ILD_SCALE|ILD_TRANSPARENT, &hicon);
return hicon;
2. Generate Bitmap
HDC hDC = GetDC(NULL);
HDC hMemDC = CreateCompatibleDC(hDC);
HBITMAP hMemBmp = CreateCompatibleBitmap(hDC, x, y);
HBITMAP hResultBmp = NULL;
HGDIOBJ hOrgBMP = SelectObject(hMemDC, hMemBmp);
HBRUSH hbr = CreateSolidBrush(bg);
RECT rr = { 0, 0, 256, 256 }; // jumbo icons
FillRect(hMemDC, &rr, hbr);
DeleteBrush(hbr);
DrawIconEx(hMemDC, 0, 0, hicon, size, size, 0, NULL, DI_NORMAL);
hResultBmp = hMemBmp;
hMemBmp = NULL;
SelectObject(hMemDC, hOrgBMP);
return hResultBitmap;
3. Render GDI+ Bitmap to "window bitmap":
Bitmap *b = ::New Bitmap(hResultBitmap, NULL);
Graphics graphics(hdc);
graphics.SetTextRenderingHint(TextRenderingHintClearTypeGridFit);
SolidBrush bgbrush(Color(255, 255, 255, 255));
Rect r(0, 0, hwnd_w, hwnd_h);
graphics.FillRectangle(&bgbrush, r);
graphics.SetInterpolationMode(InterpolationModeHighQualityBicubic);
Rect r(5, 5, 128, 128);
graphics.DrawImage(dpd->image_to_draw, r);
Wow, I spent another while last night playing with it. It's the ILD_SCALE in IImageList::GetIcon.
Get rid of that and it all works perfectly fine again. Go figure …
1. Extract Icon:
// Get the image list index of the icon
SHFILEINFO sfi;
if (!SHGetFileInfo(pszPath, 0, &sfi, sizeof(sfi), SHGFI_SYSICONINDEX)) return NULL;
// Get the jumbo image list
IImageList *piml;
if (FAILED(SHGetImageList(SHIL_JUMBO, IID_PPV_ARGS(&piml)))) return NULL;
// Extract an icon
HICON hicon;
piml->GetIcon(sfi.iIcon, ILD_TRANSPARENT, &hicon);
return hicon;

Reset existing HBITMAP as desktop background (Win32)

I wish to create a transparent window over the desktop.
For that purpose I've created an HDC with a background of the desktop (created HBITMAP of the desktop and applied it to my HDC), and invoked UpdateLayeredWindow.
So far, so good.
for performance issues i need to keep a persistent GDI+ object, meaning my HDC and HBITMAP need to stay the same handles between paintings (assuming the desktop DC didn't change), same as in this question.
In the first painting iteration all goes well. in the second painting iteration, since the HDC and HBITMAP haven't changed, I repaint on the existing HDC, meaning i get double images (the background is not erased).
Here's a code example of what I'm doing:
bool SomeUI::Draw()
{
BLENDFUNCTION blend = {0};
POINT ptPos = {0};
SIZE sizeWnd = {0};
POINT ptSrc = {0};
BOOL bUpdate = FALSE;
// Get the client rect
RECT rctWindow;
bool bGot = GetWindowRect(rctWindow);
if (!bGot)
return false;
// Get the desktop's device context
HDC hDCDesktop = GetDC(NULL);
if (!hDCDesktop)
return false;
int nWidth = abs(rctWindow.right - rctWindow.left);
int nHeight = abs(rctWindow.bottom - rctWindow.top);
// Create 32Bit bitmap to apply PNG transparency
VOID *ppvBits = NULL;
BITMAPINFO BitmapInfo = {0};
BitmapInfo.bmiHeader.biSize = sizeof(BITMAPINFOHEADER);
BitmapInfo.bmiHeader.biWidth = nWidth;
BitmapInfo.bmiHeader.biHeight = nHeight;
BitmapInfo.bmiHeader.biPlanes = 1;
BitmapInfo.bmiHeader.biBitCount = 32;
BitmapInfo.bmiHeader.biCompression = BI_RGB;
HBITMAP hBmp = CreateDIBSection(hDCDesktop, &BitmapInfo, DIB_RGB_COLORS, &ppvBits, NULL, 0);
if (!hBmp || hBmp==(HBITMAP)ERROR_INVALID_PARAMETER)
goto releaseHandles;
// Create a compatible DC and select the newly created bitmap
if (!m_hDC)
{
m_hDC = CreateCompatibleDC(hDCDesktop);
if (!m_hDC)
goto releaseHandles;
SelectObject(m_hDC, hBmp);
}
else
{
///////////////////////////////////////////////////////////////////////
//
// The problem lies here, this is where I need to reset the HBITMAP
// according to the desktop here (to have a transparent DC to work on)
//
///////////////////////////////////////////////////////////////////////
}
// The drawing logic
bool bInnerDraw = Draw(m_hDC);
if (!bInnerDraw)
goto releaseHandles;
// Call UpdateLayeredWindow
blend.BlendOp = AC_SRC_OVER;
blend.SourceConstantAlpha = 255;
blend.AlphaFormat = AC_SRC_ALPHA;
sizeWnd.cx = nWidth;
sizeWnd.cy = nHeight;
ptPos.x = rctWindow.left;
ptPos.y = rctWindow.top;
bUpdate = UpdateLayeredWindow(m_hWnd, hDCDesktop, &ptPos, &sizeWnd, m_hDC, &ptSrc, 0, &blend, ULW_ALPHA);
if (!bUpdate)
goto releaseHandles;
releaseHandles:
// releasing handles
}
Any ideas?
Found the Answer:
In order to reset the persistent HBITMAP, (reminder: it needs to stay the same handle), we'll set the desktop background of that area to a temporary HBITMAP and copy it to the persistent HBITMAP.
To achieve that (copying from one HBITMAP to the other), We'll create a temporary HDC and select the temporary HBITMAP to it, and copy the temporary HDC to the persistent HDC, usint BitBlt
Here's the code:
hBmpTemp = CreateDIBSection(hDCDesktop, &BitmapInfo, DIB_RGB_COLORS, &ppvBits, NULL, 0);
if (!hBmpTemp || hBmpTemp==(HBITMAP)ERROR_INVALID_PARAMETER)
goto releaseHandles;
HDC hTempDC = CreateCompatibleDC(NULL);
if (!hTempDC)
goto releaseHandles;
SelectObject(hTempDC, hBmpTemp);
::BitBlt(m_hPersistentDC, 0, 0, nWidth, nHeight, hTempDC, rctWindow.left, rctWindow.top, SRCCOPY);
::DeleteDC(hTempDC);

Windows 7 and ScreenShot.cpp GDI+ PNG problemo

was using XP without issue for a long time. switched to 7 and trying to capture screenshots with my previously functioning code no longer works. simple concept and relatively generic code...just find the window that i call and save it as a .png. any ideas what might make this bad boy run again? can't debug with my current setup, but it makes it all the way and spits out the error message after bmp->save(...) ...couldn't save image file. edit: also a file does get created/saved, but it is blank and not written to. perhaps the bitmap encoding or GDI is screwed up?
bool CScreenShot::Snap(CString wintitle, CString file, CString& ermsg)
{
ermsg = ""; // no error message
// create screen shot bitmap
EnumWinProcStruct prm = {0, (LPSTR)(LPCTSTR)wintitle, 0};
// Find the descriptor of the window with the caption wintitle
EnumDesktopWindows(0, EnumWindowsProc, (LPARAM)&prm);
if(!prm.hwnd)
{
ermsg.Format("couldn't find window \"%s\"", wintitle);
return false;
}
// Make the window the topmost window
SetWindowPos(prm.hwnd, HWND_TOPMOST, 0, 0, 0, 0, SWP_NOSIZE | SWP_NOMOVE);
Sleep(300);
// Get device context for the top-level window and client rect
HDC hDC = GetDC(prm.hwnd);
RECT rc;
GetClientRect(prm.hwnd, &rc);
HDC memDC = CreateCompatibleDC(hDC);
// Set the size and color depth for the screen shot image
BITMAPINFO bmpInfo;
memset(&bmpInfo, 0, sizeof(bmpInfo));
bmpInfo.bmiHeader.biSize = sizeof(bmpInfo.bmiHeader);
bmpInfo.bmiHeader.biWidth = rc.right - rc.left;
bmpInfo.bmiHeader.biHeight = rc.bottom - rc.top;
bmpInfo.bmiHeader.biPlanes = 1;
bmpInfo.bmiHeader.biBitCount = 24;
bmpInfo.bmiHeader.biCompression = BI_RGB;
bmpInfo.bmiHeader.biSizeImage = bmpInfo.bmiHeader.biWidth * bmpInfo.bmiHeader.biHeight * 3;
// Create memory buffer and perform a bit-block transfer of the color data from the window to the memory
LPVOID addr;
HBITMAP memBM = CreateDIBSection(memDC, &bmpInfo, DIB_RGB_COLORS, &addr, 0, 0);
HGDIOBJ stdBM = SelectObject(memDC, memBM);
BOOL OK = BitBlt(memDC, 0, 0, bmpInfo.bmiHeader.biWidth, bmpInfo.bmiHeader.biHeight, hDC, 0, 0, SRCCOPY);
ReleaseDC(prm.hwnd, hDC);
SetWindowPos(prm.hwnd, HWND_NOTOPMOST, 0, 0, 0, 0, SWP_NOSIZE | SWP_NOMOVE);
// Initialize GDI+.
GdiplusStartupInput gdiplusStartupInput;
ULONG_PTR gdiplusToken;
if(GdiplusStartup(&gdiplusToken, &gdiplusStartupInput, NULL) != Ok)
{
ermsg.Format("couldn't start GDI+");
return false;
}
// Create a Bitmap object for work with images defined by pixel data from the GDI HBitmap and the GDI HPalette.
Bitmap* bmp = ::new Bitmap(memBM, DIB_RGB_COLORS);
SelectObject(memDC, stdBM);
DeleteObject(memBM);
DeleteDC(memDC);
// Find the encoder for "image/png" mime type
CLSID encoderClsid;
EncoderParameters encoderParameters;
GetEncoderClsid(L"image/png", &encoderClsid);
encoderParameters.Count = 0;
// Convert file name to Unicode (wide-char) string.
WCHAR fn[_MAX_PATH];
MultiByteToWideChar(CP_THREAD_ACP, MB_PRECOMPOSED, file, file.GetLength() + 1, fn, _MAX_PATH);
// Save the screen shot into the specified file using image encoder with the mime style "image/png"
if(bmp->Save(fn, &encoderClsid, &encoderParameters) != Ok)
{
ermsg.Format("couldn't save image file \"%s\"", file);
return false;
}
::delete bmp;
GdiplusShutdown(gdiplusToken);
return true;
}
The error message implies that you're trying to save the file to a folder that you don't have permission to write to. Many folders such as Program Files are now protected. Since you didn't include the path in your sample code I'm unable to determine if this is the actual problem.
Edit: Another possibility is that the Bitmap is improperly constructed which causes the Save to fail. The second parameter to the constructor is supposed to be a handle to a palette, I think DIB_RGB_COLORS would be invalid here and you should use NULL. Also there are a couple of caveats noted in the Microsoft documentation and perhaps the different OS versions react differently when you break the rules:
You are responsible for deleting the GDI bitmap and the GDI palette. However, you should not delete the GDI bitmap or the GDI palette until after the GDI+ Bitmap::Bitmap object is deleted or goes out of scope.
Do not pass to the GDI+ Bitmap::Bitmap constructor a GDI bitmap or a GDI palette that is currently (or was previously) selected into a device context.
win7 won't accept encoderParameters.Count == 0 for some reason. Set that == 1 and you should be all set.
you probably could also just remove that parameter from Save() (overloaded)