Re: [PATCH v6 0/13] MR626: vkd3d-utils: Constant buffer reflection without vkd3d-shader.

[Henri Verbeet (＠hverbeet)]

> +/* Return a pointer to data in a code blob, with bounds checking. */
> +static const void *get_data_ptr(const struct vkd3d_shader_code *code,
> +        uint32_t offset, uint32_t count, uint32_t size)
> +{
> +    if (!require_space(offset, count, size, code->size))
> +    {
> +        ERR("Offset %#x and size %#x exceeds section size %#zx.\n", offset, size, code->size);
> +        return NULL;
> +    }
> +
> +    return (const char *)code->code + offset;
> +}

In principle that's an issue with the input data, not the implementation. So WARN instead of ERR, I'd argue.

I have a slight preference for uint8_t over char when referring to byte arrays instead of strings.
> +static HRESULT get_string(const struct vkd3d_shader_code *code, uint32_t offset, char **ret)
> +{
> +    const char *str;
> +    char *end;
> +
> +    if (offset >= code->size)
> +    {
> +        ERR("Offset %#x exceeds size %#zx.\n", offset, code->size);
> +        return E_INVALIDARG;
> +    }
> +
> +    str = (const char *)code->code + offset;
> +    if (!(end = memchr(str, 0, code->size - offset)))
> +    {
> +        ERR("String at %#x is not properly zero-terminated.\n", offset);
> +        return E_INVALIDARG;
> +    }
> +
> +    if (!(*ret = vkd3d_memdup(str, end + 1 - str)))
> +        return E_OUTOFMEMORY;
> +    return S_OK;
> +}

Likewise for the ERRs here.

Just a thought, but could we get shader runner support for reflection tests?