[PATCH 1/5] wined3d: Fix sRGB clears with no support for ARB_framebuffer_sRGB.
matteo.mystral at gmail.com
Fri Jan 27 15:22:25 CST 2017
2017-01-27 22:00 GMT+01:00 Henri Verbeet <hverbeet at gmail.com>:
> On 27 January 2017 at 19:52, Matteo Bruni <mbruni at codeweavers.com> wrote:
>> - if (!gl_info->supported[ARB_FRAMEBUFFER_SRGB] && needs_srgb_write(context, state, fb))
>> + if (!gl_info->supported[ARB_FRAMEBUFFER_SRGB]
>> + && ((!(context->d3d_info->wined3d_creation_flags & WINED3D_SRGB_READ_WRITE_CONTROL)
>> + && fb->render_targets && fb->render_targets->format_flags & WINED3DFMT_FLAG_SRGB_WRITE)
>> + || state->render_states[WINED3D_RS_SRGBWRITEENABLE]))
> I think this looks suspicious, why is this more correct?
> In particular, wouldn't this enable sRGB clears on formats without
> WINED3DFMT_FLAG_SRGB_WRITE? Patch 2/5 hints that that may be
> intentional, but in that case, shouldn't this be independent of
> ARB_framebuffer_sRGB support?
Yes, the idea is to do the "manual" sRGB correction of the clear color
when ARB_framebuffer_sRGB is not supported even if the format doesn't
The story behind this goes like "oh, the test always expects the sRGB
corrected color, which seems weird for formats / GPUs not supporting
sRGB. Let me fix the test. Oh, wait, the test now fails on the GF4 Go
(which doesn't advertise WINED3DFMT_FLAG_SRGB_WRITE for the relevant
texture format) and actually it was passing before. I guess wined3d is
So, this is the fix spurred from that test result, in that we do the
manual clear color sRGB correction if ARB_framebuffer_sRGB is not
supported AND we're on d3d9, regardless of the format flags. If
ARB_framebuffer_sRGB is supported we enable GL_FRAMEBUFFER_SRGB in
context_apply_clear_state() instead and the condition there should
still be fine (I think?).
More information about the wine-devel