On 8 March 2017 at 20:29, Stefan Dösinger stefandoesinger@gmx.at wrote:
*) Accept the Nvidia XRGB color key results with X != 0. I am open to removing this test entirely as we're accepting a lot of different outcomes now.
Yeah, probably. It does seem consistent with the idea that the "X" channel is undefined, and I don't think that particular test was motivated by application behaviour.
Am 2017-03-09 um 11:00 schrieb Henri Verbeet:
On 8 March 2017 at 20:29, Stefan Dösinger stefandoesinger@gmx.at wrote:
*) Accept the Nvidia XRGB color key results with X != 0. I am open to removing this test entirely as we're accepting a lot of different outcomes now.
Yeah, probably. It does seem consistent with the idea that the "X" channel is undefined, and I don't think that particular test was motivated by application behaviour.
https://www.winehq.org/pipermail/wine-patches/2005-August/019938.html mentions the game Black Mirror, and it seems that the game requires that the X channel is ignored for color keying - this behavior makes sense, but one of our broken() results is different. It appears that the behavior I've seen on system surfaces on Windows would break the game. I originally thought it's possible that the game uses 16 bit surfaces and sets higher bits in the color key, but apparently this is not the case.
I'll spend a few minutes to try to break the game's transparency on Windows, simply running it with the basic display driver might do the job. But I think since the game requires a certain behavior re the X channel we should keep the test. The other 3 behaviors (X doesn't influence color keying, but may be copied or set to 0x00 or 0xff) are all fine for the game.
-
Henri Verbeet -
Stefan Dösinger