2012/1/14 Saulius Krasuckas saulius2@ar.fi.lt:
- On Tue, 27 Dec 2011, Matteo Bruni wrote:
dlls/d3d9/tests/visual.c | 388 ++++++++++++++++++++++++++++++++-------------- 1 files changed, 269 insertions(+), 119 deletions(-)
Hello Matteo,
this is http://source.winehq.org/git/wine.git/commitdiff/8dee7989f242b8ea624abc3b1fe...
The patch removed 6 failures on XP on real hardware (nVidia FX5200): http://test.winehq.org/data/18c20964e1e22f9aa974cc774bf77da6fc013716/xp_s2-s...
visual.c:4057: Test failed: proj: Pixel 162/122 has color 0x00000000, expected 0x00FFFFFF visual.c:4062: Test failed: proj: Pixel 158/178 has color 0x00000000, expected 0x00FFFFFF visual.c:4069: Test failed: proj: Pixel 318/118 has color 0x00ffffff, expected 0x00000000 visual.c:4071: Test failed: proj: Pixel 322/118 has color 0x00ffffff, expected 0x00000000 visual.c:4075: Test failed: proj: Pixel 322/122 has color 0x00ffffff, expected 0x00000000 visual.c:4078: Test failed: proj: Pixel 318/178 has color 0x00000000, expected 0x00FFFFFF
and introduced 14 new ones: http://test.winehq.org/data/277361d7be206e1ff7634998f0d750f7c6d66498/xp_s2-s...
visual.c:3729: Test failed: D3DTTFF_COUNT3 | D3DTTFF_PROJECTED - bottom right: Pixel (401, 361) has color 00000000, expected 00ffffff visual.c:3729: Test failed: D3DTTFF_COUNT3 | D3DTTFF_PROJECTED - bottom right: Pixel (401, 419) has color 00000000, expected 00ffffff visual.c:3729: Test failed: D3DTTFF_COUNT3 | D3DTTFF_PROJECTED - bottom right: Pixel (481, 359) has color 00ffffff, expected 00000000 visual.c:3729: Test failed: D3DTTFF_COUNT3 | D3DTTFF_PROJECTED - bottom right: Pixel (481, 361) has color 00ffffff, expected 00000000 visual.c:3729: Test failed: D3DTTFF_COUNT3 | D3DTTFF_PROJECTED - bottom right: Pixel (479, 359) has color 00ffffff, expected 00000000 visual.c:3729: Test failed: D3DTTFF_COUNT3 | D3DTTFF_PROJECTED - bottom right: Pixel (479, 419) has color 00000000, expected 00ffffff visual.c:3729: Test failed: D3DTTFF_PROJECTED (like COUNT3 | PROJECTED, texcoord has only 3 components) - bottom right: Pixel (401, 361) has color 00000000, expected 00ffffff visual.c:3729: Test failed: D3DTTFF_PROJECTED (like COUNT3 | PROJECTED, texcoord has only 3 components) - bottom right: Pixel (401, 419) has color 00000000, expected 00ffffff visual.c:3729: Test failed: D3DTTFF_PROJECTED (like COUNT3 | PROJECTED, texcoord has only 3 components) - bottom right: Pixel (479, 361) has color 00000000, expected 00ffffff visual.c:3729: Test failed: D3DTTFF_PROJECTED (like COUNT3 | PROJECTED, texcoord has only 3 components) - bottom right: Pixel (479, 419) has color 00000000, expected 00ffffff visual.c:3729: Test failed: 0xffffffff (like COUNT3 | PROJECTED, texcoord has only 3 components) - top left: Pixel (81, 121) has color 00000000, expected 00ffffff visual.c:3729: Test failed: 0xffffffff (like COUNT3 | PROJECTED, texcoord has only 3 components) - top left: Pixel (81, 179) has color 00000000, expected 00ffffff visual.c:3729: Test failed: 0xffffffff (like COUNT3 | PROJECTED, texcoord has only 3 components) - top left: Pixel (159, 121) has color 00000000, expected 00ffffff visual.c:3729: Test failed: 0xffffffff (like COUNT3 | PROJECTED, texcoord has only 3 components) - top left: Pixel (159, 179) has color 00000000, expected 00ffffff
The same machine running Win7 didn't exhibit any changes: http://test.winehq.org/data/18c20964e1e22f9aa974cc774bf77da6fc013716/win7_s2... http://test.winehq.org/data/277361d7be206e1ff7634998f0d750f7c6d66498/win7_s2...
This probably has something to do with older nVidia driver version being used on Win7 - v96.85 - than the version used on XP - v163.75 .
But on Win7 somewhat similar change is seen on virtual machine from Francois, only the failure count increases by 28 here: http://test.winehq.org/data/18c20964e1e22f9aa974cc774bf77da6fc013716/win7_fg... http://test.winehq.org/data/277361d7be206e1ff7634998f0d750f7c6d66498/win7_fg...
Would you mind fixing this please:)?
S.
I somehow expected some test failures, since this is really testing "invalid" d3d9 states combinations. Those results show that there are some minor discrepancies with the actual output on the screen, but the bottom line is that invalid settings still work, which was the main objective of the tests. So adding some broken() around is probably okay. I'll send a patch.
So, I was more or less expecting this and I did already notice the test failures, but thanks for the nice summary and heads up ;)
I think I'll just ignore the virtual machine test failures: too many tests are failing already and I don't think those results are interesting anyway.