On 9/23/19 16:43, Henri Verbeet wrote:
it's worth, note also that ARB_buffer_storage explicitly adds the ability to draw from persistently mapped buffers.
Yes, sure. Both my Nvidia and Intel cards on which I tested this have ARB_buffer_storage extension, but none of them allow drawing from buffers mapped without appropriate flag.
Do you think it is ok then to just drop the unstable test (that second attempt draws)? I am afraid it might be too fragile regardless, the failure that I initially expect here is already relying on hitting the direct map case in wined3d buffer.c, which is likely to change sooner or later.
Sure. There might be something to be said for keeping the test but ignoring the result only on Wine, but I don't have a strong opinion on the matter.
I would prefer to keep it then and ignore Wine result for test 0.