On 4/20/22 08:54, Matteo Bruni wrote:
On Mon, Apr 18, 2022 at 8:34 AM Giovanni Mascellani gmascellani@codeweavers.com wrote:
From: Zebediah Figura zfigura@codeweavers.com
tests/conditional.shader_test | 4 +- tests/d3d12_test_utils.h | 8 ---- tests/hlsl-for.shader_test | 6 +-- tests/shader_runner.c | 69 +++++++++++----------------------- tests/texture-load.shader_test | 8 ++-- tests/trigonometry.shader_test | 32 ++++++++-------- tests/utils.h | 8 ++++ 7 files changed, 54 insertions(+), 81 deletions(-)
diff --git a/tests/conditional.shader_test b/tests/conditional.shader_test index e665ac1d..42307777 100644 --- a/tests/conditional.shader_test +++ b/tests/conditional.shader_test @@ -15,5 +15,5 @@ float4 main(float tex : texcoord) : SV_TARGET
[test] draw quad -probe rect rgba ( 0, 0, 319, 480) (0.9, 0.8, 0.7, 0.6) -probe rect rgba (321, 0, 640, 480) (0.1, 0.2, 0.3, 0.4) +probe ( 0, 0, 319, 480) rgba (0.9, 0.8, 0.7, 0.6) +probe (321, 0, 640, 480) rgba (0.1, 0.2, 0.3, 0.4)
Nothing about the patch itself, just something that I noticed: these "explicit" rect coordinates are a bit awkward since we switched to RENDER_TARGET_WIDTH / RENDER_TARGET_HEIGHT in the shader runner. OTOH specifying e.g. one-pixel offsets with normalized coords isn't particularly nice either.
I don't see it as particularly necessary, although I don't object to it either. Part of the point of RENDER_TARGET_WIDTH was to ensure that if we in the future wanted to make the RT size customizable, we wouldn't miss anything while changing the code to account for that.
If we do make the coörds normalized, we could probably just subtract a pixel or two from the edges after un-normalizing them.