On 27.02.2017 14:47, Józef Kucia wrote:
On Mon, Feb 27, 2017 at 1:49 PM, Sebastian Lackner sebastian@fds-team.de wrote:
As you might be aware, the problem on Nvidia binary can also be worked around by using slightly different offsets (basically reverting 1bed10abee7d2439a5c7358596c860c454f06b53). Do you think its better to disable it, or should we just use different pixel center offsets?
The offsets used by 1bed10abee7d2439a5c7358596c860c454f06b53 are more correct, e.g. because of filling conventions. I don't think using different offsets is a correct thing to do.
I'm not suggesting to change it in general, but the question is if it is reliable enough to be used as a quirk on Nvidia drivers. It is sufficient to pass all the Wine tests at least. If you think manual fixup is better, thats also fine ofc.
We also have a code path which handles the pixel offsets correctly when ARB_clip_control is not available. On the other hand, we may want to reconsider the ARB_viewport_array detection logic when we'll add support for multiple viewports in wined3d.