Re: [PATCH vkd3d 4/6] vkd3d-shader: Implement hlsl_error() and hlsl_warning().

[Zebediah Figura (she/her)]

On 2/18/21 5:49 PM, Matteo Bruni wrote:
> On Wed, Feb 17, 2021 at 6:52 AM Zebediah Figura  wrote:
> > Signed-off-by: Zebediah Figura
> >
> > libs/vkd3d-shader/hlsl.c                 |  18 +-
> >  libs/vkd3d-shader/hlsl.h                 |   8 +-
> >  libs/vkd3d-shader/hlsl.l                 |   3 +-
> >  libs/vkd3d-shader/hlsl.y                 | 204 ++++++++++++-----------
> >  libs/vkd3d-shader/vkd3d_shader_private.h |  18 ++
> >  5 files changed, 147 insertions(+), 104 deletions(-)
>
> I generally like the changes to the error messages. Regardless, I have
> a couple of comments...
> > diff --git a/libs/vkd3d-shader/hlsl.y b/libs/vkd3d-shader/hlsl.y
> > index c33d97c6..9ffabbcb 100644
> > --- a/libs/vkd3d-shader/hlsl.y
> > +++ b/libs/vkd3d-shader/hlsl.y
> > @@ -768,12 +776,13 @@ static bool add_typedef(struct hlsl_ctx *ctx, DWORD modifiers, struct hlsl_type
> >      if ((type->modifiers & HLSL_MODIFIER_COLUMN_MAJOR)
> >              && (type->modifiers & HLSL_MODIFIER_ROW_MAJOR))
> > -
> >        hlsl_error(ctx, v->loc, "more than one matrix majority keyword");
> > -
> >        hlsl_error(ctx, v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER,
> > -
> >                "'row_major' and 'column_major' modifiers are mutually exclusive.");
> >
> >    ret = hlsl_scope_add_type(ctx->cur_scope, type);
> >    if (!ret)
> > -
> >        hlsl_error(ctx, v->loc,
> > -
> >                "redefinition of custom type '%s'", v->name);
> > -
> >        hlsl_error(ctx, v->loc, VKD3D_SHADER_ERROR_HLSL_REDEFINED,
> > -
> >                "Typedef '%s' is already defined.", v->name);
> >    vkd3d_free(v);
> > }
> >    vkd3d_free(list);
>
> This made me wonder whether defining a struct also implicitly defines
> a type (i.e. can you reference a struct type by its name without
> 'struct' even when there is no explicit typedef?). A quick test
> suggests that's the case, if that's confirmed we should probably fix
> it at some point.

It does, and it actually already works with our parser. I guess that
makes the message here slightly misleading (maybe "Type '%s'" would be
better?)
> > @@ -1772,15 +1783,11 @@ struct_declaration:
> >              if (!$3)
> >              {
> >                  if (!$2->name)
> > -
> >            {
> > -
> >                hlsl_error(ctx, @2,
> > -
> >                        "anonymous struct declaration with no variables");
> > -
> >            }
> > -
> >                hlsl_error(ctx, @2, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX,
> > -
> >                        "Struct type definitions cannot be anonymous.");
> >            if (modifiers)
> > -
> >            {
> > -
> >                hlsl_error(ctx, @1,
> > -
> >                        "modifier not allowed on struct type declaration");
> > -
> >            }
> > -
> >                hlsl_error(ctx, @1, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER,
> > -
> >                        "Modifiers are not allowed on struct type definitions.");
> >        }
> >
> >        if (!(type = apply_type_modifiers(ctx, $2, &modifiers, @1)))
>
> Splitting hairs, but I think "declaration" is a bit nicer for both of
> the above (the C spec refers to those as declarations, fwiw). The
> message could be improved to clarify that the declaration being
> anonymous is only a problem if there are no variables being defined at
> the same time (which I assume is what you mean with "Struct type
> definitions"), but I don't have any good practical suggestions.

Fair points, I'll change these.