[Spice-devel] [PATCH] server: red_current_add_equal - don't push a drawable to the middle of the pipe if it depends on surfaces.

[Yonit Halperin]

On 08/31/2010 11:24 AM, Gerd Hoffmann wrote:
> On 08/31/10 10:11, Yonit Halperin wrote:
>> On 08/31/2010 11:05 AM, Gerd Hoffmann wrote:
>>>> +static inline int is_drawable_independent_from_surfaces(Drawable
>>>> *drawable)
>>>> +{
>>>> + int x;
>>>> +
>>>> + for (x = 0; x< 3; ++x) {
>>>> + if (drawable->surfaces_dest[x] != -1) {
>>>> + return FALSE;
>>>> + }
>>>> + }
>>>> + return TRUE;
>>>> +}
>>>
>>> What happens if this meets qemu without the surfaces init fix?
>>> Will it break or just work less efficient?
>>>
>>> cheers,
>>> Gerd
>>>
>>
>> I think that if the server will decide to video stream an area, it can
>> cause a delay in between frames if the pipe is busy.
>
> Ok, no breakage, good.

>
> While we are at it: How should surfaces_dest (and surfaces_rect) be
> filled for spice 0.4 commands? Right now they are just zero-initialized
> (see red_get_compat_drawable() in red_parse_qxl.c).
>
Best to fill surfaces_dest with -1. For consistency (though not 
necessary right now), in QXL_COPY_BITS scenario, fill surfaces_dest[0] 
with the drawable surface id, and surfaces_rect[0] with the rect of the 
drawabele transformed to src_pos (i.e.,  starts at src_pos and its width 
and hight are equal to the drawable destination).

> Oh, and I've just seen self bitmaps are not handled yet.
>
> spice 0.4 has:
>
> uint16_t bitmap_offset;
> QXLRect bitmap_area;
>
> spice 0.6 has:
>
> uint8_t self_bitmap;
> QXLRect self_bitmap_area;
>
> I suspect the area can just be used as-is. What about bitmap_offset /
> self_bitmap?
>
if bitmap_offset != 0, it means self_bitmap = TRUE

> cheers,
> Gerd
>