On 5/27/22 13:11, Zebediah Figura wrote:
On 5/26/22 01:51, Chip Davis wrote:
From: Chip Davis cdavis5x@gmail.com
Signed-off-by: Chip Davis cdavis5x@gmail.com
dlls/wined3d/device.c | 202 ++++++++++++++++++++++++++++++++++++++ dlls/wined3d/wined3d.spec | 1 + include/wine/wined3d.h | 49 +++++++++ 3 files changed, 252 insertions(+)
This doesn't seem like an improvement. I think it potentially makes sense to use wined3d format information (plus the feature level) to more accurately match the flags Windows reports—instead of reporting exactly what the backend is actually capable of—but I don't see a need to move the whole implementation to wined3d, not when it means adding a lot of new flags to the API, and not when d3d11 is the only direct user.
Separately, I think changes to the flags we report should be made in a much more granular fashion, one per commit. Applications can be rather sensitive to these flags, and I'd rather make any regressions easily bisectable.
FWIW, I can see an argument for returning the format attrs and caps directly, instead of querying for one at a time—i.e. make the wined3d API more similar to the d3d11 API without introducing a 1:1 translation. I'm not sure it's *necessary*, but if it makes things simpler I'm not opposed.