Found the culprit.
Basically it was a resource exhaustion problem. My basic font
rendering was:
SelectObject(hdc, font);
hbm =3D CreateCompatibleBitmap( hdcDisplay, w, h );
hbmOld=3DSelectObject(hdc,hbm);
TextOut(hdc, 0, 0, buf, strlen(buf));
SelectObject(hdc,hbmOld);
GetDIBBits( ... ); //get BMI info
GetDIBBits( ... ); //get actual data
free( bits );
free( bmi );
SelectObject( hdc, hfontOld );
The font in use is retained across calls, as is the DC I created
(using CreateCompatibleDC()). So this function basically just selects
a font into the DC, creates a bitmap, selects the bitmap, renders
text, deselects the bitmap, grabs the bits, frees them, unselects the
font.
The problem was that I was forgetting to call DeleteObject( hbm ) -- I
was creating a new bitmap every time the function was called, which is
why it was happening that "1% of times" -- it took probably a few
hundred calls (at like 20 calls/second) for it to manifest.
My guess is that HBITMAPs were exhausted or something (these were
pretty small bitmaps, probably 8K or so each), but instead of dying on
me and reporting an error where you'd expect, it actually died on me
inside GetDIBBits() after I received a valid HBITMAP. The NEXT time
through it would then die on CreateCompatibleBitmap(), so I'm assuming
that some kind of resource acquisition/lock was failing inside of
GetDIBBits() that was failing even though I had a valid HBITMAP.
Adding DeleteObject() made the problem go away, so I'm a happy
programmer (for now). Still not sure why GetLastError() wasn't
reporting anything, but oh well.
Brian
|
