ReceiveTempSMS/Sweden-Number
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

opengl32: Drop the ChoosePixelFormat test as it is very unreliable.

Browse Source
This commit is contained in:
Roderick Colenbrander 2008-04-30 10:54:19 +00:00 committed by Alexandre Julliard
parent 21171c403f
commit 3f89e1a053
1 changed files with 0 additions and 75 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

75
dlls/opengl32/tests/opengl.c
View File

@ -176,80 +176,6 @@ static void test_pbuffers(HDC hdc)
else skip("Pbuffer test for offscreen pixelformat skipped as no offscreen-only format with pbuffer capabilities has been found\n");
}
static void test_choosepixelformat(HDC hdc)
{
int iPixelFormat;
int nFormats;
BOOL found=FALSE;
PIXELFORMATDESCRIPTOR pfd = {
sizeof(PIXELFORMATDESCRIPTOR),
1, /* version */
PFD_DRAW_TO_WINDOW |
PFD_SUPPORT_OPENGL |
PFD_DOUBLEBUFFER,
PFD_TYPE_RGBA,
32, /* 32-bit color depth */
0, 0, 0, 0, 0, 0, /* color bits */
0, /* alpha buffer */
0, /* shift bit */
0, /* accumulation buffer */
0, 0, 0, 0, /* accum bits */
0, /* z-buffer */
0, /* stencil buffer */
0, /* auxiliary buffer */
PFD_MAIN_PLANE, /* main layer */
0, /* reserved */
0, 0, 0 /* layer masks */
};
/* Below we test the behavior of ChoosePixelFormat. As documented on MSDN this
* function doesn't give any guarantees about its outcome. Programs should not
* rely on weird behavior of the function but unfortunately a few programs like
* e.g. Serious Sam TSE rely on it.
*
* MSDN documents of a few flags like double buffering / stereo that they can be set to DONTCARE.
* It appears that a 0 value on other options like alpha, red, .. means DONTCARE. The hypothesis
* is that ChoosePixelFormat returns the first available format which matches the criteria.
*
* This test tries to proof the DONTCARE behavior by passing an almost 'empty' pfd to
* ChoosePixelFormat. The pfd only has some really needed flags (RGBA, window, double buffer) set.
* Further a 32 bit color buffer has been requested. The idea is that when a format with e.g. depth or stencil bits
* is returned, while there are also 'better' candidates in the list without them (but located AFTER the returned one)
* that an option set to zero means DONTCARE. We try to proof this by checking the aux/depth/stencil bits.
* Proofing this behavior for the color bits isn't possible as all formats have red/green/blue/(alpha), so we assume
* that if it holds for aux/depth/stencil it also holds for the others.
*
* The test below passes at least on various ATI cards (rv250, r300) and Nvidia cards.
*/
iPixelFormat = ChoosePixelFormat(hdc, &pfd);
if(iPixelFormat) {
PIXELFORMATDESCRIPTOR pfd_tmp;
BOOL res;
int i;
memset(&pfd, 0, sizeof(PIXELFORMATDESCRIPTOR));
res = DescribePixelFormat(hdc, iPixelFormat, sizeof(PIXELFORMATDESCRIPTOR), &pfd);
nFormats = DescribePixelFormat(hdc, 0, 0, NULL);
/* Start testing from iPixelFormat, second formats start counting from index=1, so use '<=' */
for(i=iPixelFormat; i<=nFormats; i++) {
memset(&pfd_tmp, 0, sizeof(PIXELFORMATDESCRIPTOR));
res = DescribePixelFormat(hdc, i, sizeof(PIXELFORMATDESCRIPTOR), &pfd_tmp);
if(!res)
continue;
/* Check if there is a format which better matches the requirements */
if((pfd_tmp.cAuxBuffers < pfd.cAuxBuffers) || (pfd_tmp.cDepthBits < pfd.cDepthBits) || (pfd_tmp.cStencilBits < pfd.cStencilBits))
found = TRUE;
}
/* When found=TRUE we were able to confirm our hypothesis */
ok(found == TRUE, "Unable to confirm DONTCARE behavior of unset pixelformatdescriptor flags\n");
}
}
static void test_setpixelformat(HDC winhdc)
{
int res = 0;
@ -472,7 +398,6 @@ START_TEST(opengl)
ok(res, "wglMakeCurrent failed!\n");
init_functions();
test_choosepixelformat(hdc);
test_setpixelformat(hdc);
test_colorbits(hdc);
test_gdi_dbuf(hdc);