Re: [PATCH v2 0/1] MR808: d2d1/tests: Mark some more failing tests as todo.

[Nikolay Sivov (＠nsivov)]

On Thu Sep  8 19:12:58 2022 +0000, **** wrote:
> Zebediah Figura replied on the mailing list:
> On 9/8/22 13:58, Alexandre Julliard wrote:
> > @@ -7268,6 +7289,7 @@ static void test_draw_geometry(BOOL d3d11)
> >               "PBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8"
> >               "FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwU"
> >               "PBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8ZDxkPGQ8ZDxk3i8A");
> > +    todo_wine_if (!match && is_nvidia)
> >       ok(match, "Figure does not match.\n");
> >       match = compare_figure(&ctx, 160,   0, 160, 160, 0xff652e89, 0,
> >               "vi5kPGQ8ZDxkPGQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwU"
> > @@ -7276,6 +7298,7 @@ static void test_draw_geometry(BOOL d3d11)
> >               "PBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8"
> >               "FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwU"
> >               "PBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8ZDxkPGQ8ZDxk3i8A");
> > +    todo_wine_if (!match && is_nvidia)
> >       ok(match, "Figure does not match.\n");
> >       match = compare_figure(&ctx, 320,   0, 160, 160, 0xff652e89, 0,
> >               "vi5kPGQ8ZDxkPGQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwU"
> > @@ -7284,6 +7307,7 @@ static void test_draw_geometry(BOOL d3d11)
> >               "PBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8"
> >               "FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwU"
> >               "PBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8ZDxkPGQ8ZDxk3i8A");
> > +    todo_wine_if (!match && is_nvidia)
> >       ok(match, "Figure does not match.\n");
> >       match = compare_figure(&ctx, 480,   0, 160, 160, 0xff652e89, 0,
> >               "yC5aRlpGWjxkPGQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwU"
> > @@ -7292,6 +7316,7 @@ static void test_draw_geometry(BOOL d3d11)
> >               "PBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8"
> >               "FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwU"
> >               "PBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8FDwUPBQ8ZDxkPGQ8ZDxk3i8A");
> > +    todo_wine_if (!match && is_nvidia)
> >       ok(match, "Figure does not match.\n");
> >   
> >       match = compare_figure(&ctx,   0, 160, 160, 160, 0xff652e89, 64,
> I see the 2nd and 3rd failure on radeonsi (and radv), and all four 
> failures on llvmpipe (and lavapipe). This is with Mesa 22.2.0-rc3. 
> Nikolay, Henri, do neither of you see test failures, and with which 
> drivers if so?

I only run it on Intel 8086:0416, and that shows no failures. It's also an old one, I don't know if newer mesa updates change anything there as much as they do for current gpus. We had failures like that before I think, the solution was to tweak geometries or transforms to make it more reliable.