<div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Oct 27, 2016 at 11:07 PM, Leo Liu <span dir="ltr"><<a href="mailto:leo.liu@amd.com" target="_blank">leo.liu@amd.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div bgcolor="#FFFFFF" text="#000000"><div><div class="h5">
    <br>
    <br>
    <div class="m_2199568300170825623moz-cite-prefix">On 10/27/2016 01:20 PM, Nayan Deshmukh
      wrote:<br>
    </div>
    <blockquote type="cite">
      
      <div dir="ltr"><br>
        <div class="gmail_extra"><br>
          <div class="gmail_quote">On Thu, Oct 27, 2016 at 10:17 PM, Leo
            Liu <span dir="ltr"><<a href="mailto:leo.liu@amd.com" target="_blank">leo.liu@amd.com</a>></span>
            wrote:<br>
            <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
              <div class="m_2199568300170825623gmail-HOEnZb">
                <div class="m_2199568300170825623gmail-h5"><br>
                  <br>
                  On 10/27/2016 12:20 PM, Nayan Deshmukh wrote:<br>
                  <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                    On Thu, Oct 27, 2016 at 10:38:30AM -0400, Leo Liu
                    wrote:<br>
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <br>
                      On 10/24/2016 09:55 AM, Nayan Deshmukh wrote:<br>
                      <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                        Suggested-by: Leo Liu <<a href="mailto:leo.liu@amd.com" target="_blank"></a><a class="m_2199568300170825623moz-txt-link-abbreviated" href="mailto:leo.liu@amd.com" target="_blank">leo.liu@amd.com</a>><br>
                        Signed-off-by: Nayan Deshmukh <<a href="mailto:nayan26deshmukh@gmail.com" target="_blank"></a><a class="m_2199568300170825623moz-txt-link-abbreviated" href="mailto:nayan26deshmukh@gmail.com" target="_blank">nayan26deshmukh@gmail.com</a>><br>
                        ---<br>
                           src/gallium/auxiliary/vl/vl_w<wbr>insys.h   
                          |  4 ++<br>
                           src/gallium/auxiliary/vl/vl_w<wbr>insys_dri3.c
                        | 89 +++++++++++++++++++++++++++---<wbr>-<br>
                           2 files changed, 83 insertions(+), 10
                        deletions(-)<br>
                        <br>
                        diff --git a/src/gallium/auxiliary/vl/vl_<wbr>winsys.h
                        b/src/gallium/auxiliary/vl/vl_<wbr>winsys.h<br>
                        index 26db9f2..7c56b48 100644<br>
                        --- a/src/gallium/auxiliary/vl/vl_<wbr>winsys.h<br>
                        +++ b/src/gallium/auxiliary/vl/vl_<wbr>winsys.h<br>
                        @@ -59,6 +59,10 @@ struct vl_screen<br>
                              void *<br>
                              (*get_private)(struct vl_screen *vscreen);<br>
                        +   void<br>
                        +   (*set_output_texture)(struct vl_screen
                        *vscreen, struct pipe_resource *buffer,<br>
                        +                         uint32_t width,
                        uint32_t height);<br>
                        +<br>
                      </blockquote>
                      I think it's more appropriate if it's called
                      "set_back_texture_from_output"<br>
                    </blockquote>
                    Yeah that makes more sense.<br>
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                              struct pipe_screen *pscreen;<br>
                              struct pipe_loader_device *dev;<br>
                           };<br>
                        diff --git a/src/gallium/auxiliary/vl/vl_<wbr>winsys_dri3.c
                        b/src/gallium/auxiliary/vl/vl_<wbr>winsys_dri3.c<br>
                        index 2929928..44d6f4c 100644<br>
                        --- a/src/gallium/auxiliary/vl/vl_<wbr>winsys_dri3.c<br>
                        +++ b/src/gallium/auxiliary/vl/vl_<wbr>winsys_dri3.c<br>
                        @@ -56,6 +56,7 @@ struct vl_dri3_buffer<br>
                              struct xshmfence *shm_fence;<br>
                              bool busy;<br>
                        +   bool is_external_texture;<br>
                              uint32_t width, height, pitch;<br>
                           };<br>
                        @@ -71,6 +72,9 @@ struct vl_dri3_screen<br>
                              xcb_special_event_t *special_event;<br>
                              struct pipe_context *pipe;<br>
                        +   struct pipe_resource *output_texture;<br>
                        +   uint32_t output_texture_width;<br>
                        +   uint32_t output_texture_height;<br>
                              struct vl_dri3_buffer
                        *back_buffers[BACK_BUFFER_NUM]<wbr>;<br>
                              int cur_back;<br>
                        @@ -105,7 +109,8 @@ dri3_free_back_buffer(struct
                        vl_dri3_screen *scrn,<br>
                              xcb_free_pixmap(scrn->conn,
                        buffer->pixmap);<br>
                              xcb_sync_destroy_fence(scrn->c<wbr>onn,
                        buffer->sync_fence);<br>
                              xshmfence_unmap_shm(buffer->sh<wbr>m_fence);<br>
                        -   pipe_resource_reference(&buff<wbr>er->texture,
                        NULL);<br>
                        +   if (!buffer->is_external_texture)<br>
                        +      pipe_resource_reference(&buffe<wbr>r->texture,
                        NULL);<br>
                              if (buffer->linear_texture)<br>
                                  pipe_resource_reference(&buffe<wbr>r->linear_texture,
                        NULL);<br>
                              FREE(buffer);<br>
                        @@ -236,13 +241,24 @@
                        dri3_alloc_back_buffer(struct vl_dri3_screen
                        *scrn)<br>
                              templ.format = PIPE_FORMAT_B8G8R8X8_UNORM;<br>
                              templ.target = PIPE_TEXTURE_2D;<br>
                              templ.last_level = 0;<br>
                        -   templ.width0 = scrn->width;<br>
                        -   templ.height0 = scrn->height;<br>
                        +   if (scrn->output_texture) {<br>
                        +      templ.width0 =
                        (scrn->output_texture_width) ?<br>
                        +                   
                         scrn->output_texture_width :<br>
                        +                   
                         scrn->output_texture->width0;<br>
                        +      templ.height0 =
                        (scrn->output_texture_height) ?<br>
                        +                     
                        scrn->output_texture_height :<br>
                        +                     
                        scrn->output_texture->height0;<br>
                        +   } else {<br>
                        +       templ.width0 = scrn->width;<br>
                        +       templ.height0 = scrn->height;<br>
                        +   }<br>
                              templ.depth0 = 1;<br>
                              templ.array_size = 1;<br>
                              if (scrn->is_different_gpu) {<br>
                        -      buffer->texture =
                        scrn->base.pscreen->resource_c<wbr>reate(scrn->base.pscreen,<br>
                        +      buffer->texture =
                        (scrn->output_texture) ?<br>
                        +                        scrn->output_texture
                        :<br>
                        +                       
                        scrn->base.pscreen->resource_c<wbr>reate(scrn->base.pscreen,<br>
                                                                       
                                       &templ);<br>
                      </blockquote>
                      Why not just make it in the same line?<br>
                      <br>
                    </blockquote>
                    I was just trying to limit characters to 78 as given
                    in mesa coding style.<br>
                  </blockquote>
                  <br>
                </div>
              </div>
              It's really matters if just over a few chars. which way do
              you think it will look better?
              <div>
                <div class="m_2199568300170825623gmail-h5"><br>
                  <br>
                </div>
              </div>
            </blockquote>
            <div>I would prefer it in a single line, anyways I just
              checked that it used in single line in entire </div>
            <div>vl code.  </div>
            <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
              <div>
                <div class="m_2199568300170825623gmail-h5">
                  <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                    <br>
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                                 if (!buffer->texture)<br>
                                    goto unmap_shm;<br>
                        @@ -257,7 +273,9 @@
                        dri3_alloc_back_buffer(struct vl_dri3_screen
                        *scrn)<br>
                                    goto no_linear_texture;<br>
                              } else {<br>
                                 templ.bind |= PIPE_BIND_SCANOUT |
                        PIPE_BIND_SHARED;<br>
                        -      buffer->texture =
                        scrn->base.pscreen->resource_c<wbr>reate(scrn->base.pscreen,<br>
                        +      buffer->texture =
                        (scrn->output_texture) ?<br>
                        +                        scrn->output_texture
                        :<br>
                        +                       
                        scrn->base.pscreen->resource_c<wbr>reate(scrn->base.pscreen,<br>
                                                                       
                                       &templ);<br>
                      </blockquote>
                      Same as above .<br>
                      <br>
                      <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                                 if (!buffer->texture)<br>
                                    goto unmap_shm;<br>
                        @@ -271,11 +289,20 @@
                        dri3_alloc_back_buffer(struct vl_dri3_screen
                        *scrn)<br>
                                                                     
                        usage);<br>
                              buffer_fd = whandle.handle;<br>
                              buffer->pitch = whandle.stride;<br>
                        +   buffer->width = templ.width0;<br>
                        +   buffer->height = templ.height0;<br>
                        +   buffer->is_external_texture =
                        (scrn->output_texture) ?<br>
                        +                                 true :<br>
                        +                                 false;<br>
                      </blockquote>
                      Same as above.<br>
                      <br>
                      <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                        +   scrn->output_texture = NULL;<br>
                        +   scrn->output_texture_width = 0;<br>
                        +   scrn->output_texture_height = 0;<br>
                        +<br>
                              xcb_dri3_pixmap_from_buffer(sc<wbr>rn->conn,<br>
                                                          (pixmap =
                        xcb_generate_id(scrn->conn)),<br>
                                                         
                        scrn->drawable,<br>
                                                          0,<br>
                        -                               scrn->width,
                        scrn->height, buffer->pitch,<br>
                        +                             
                         buffer->width, buffer->height,
                        buffer->pitch,<br>
                      </blockquote>
                      The width/height here should be same as buffer
                      texture width/height.<br>
                      That's case before, now that is changed to<br>
                      <br>
                             templ.width0 =
                      (scrn->output_texture_width) ?<br>
                                           
                      scrn->output_texture_width :<br>
                                           
                      scrn->output_texture->width0;<br>
                             templ.height0 =
                      (scrn->output_texture_height) ?<br>
                                           
                       scrn->output_texture_height :<br>
                                           
                       scrn->output_texture->height0<wbr>;<br>
                      <br>
                    </blockquote>
                    I did this to support clipping for vdpau target
                    display as we<br>
                    may have to display a portion of the buffer instead
                    of the entire buffer<br>
                    <br>
                    In the patch you suggested you were using scrn
                    width/height for the buffer<br>
                    width/height but if there is a screen resize event
                    then we are sending a small<br>
                    buffer to X with extra width/height and would lead
                    to error and hence I used<br>
                    buffer width/height<br>
                  </blockquote>
                  <br>
                </div>
              </div>
              scrn size always updated when resize event happens, and
              back buffer size is updated accordingly.<br>
              <br>
            </blockquote>
            <div>Exactly but when this happens the back buffer size is
              updated accordingly but the output texture will still</div>
            <div>be of smaller size and won't get get updated until the
              next frame so it will also cause errors when we have a</div>
            <div>pixmap of screen size. I tried setting it to screen
              size instead of buffer size and got the following errors</div>
            <div>
              <div><br>
              </div>
              <div>X11 error: BadAlloc (insufficient resources for
                operation)</div>
              <div>X11 error: BadDrawable (invalid Pixmap or Window
                parameter)</div>
              <div>X11 error: BadPixmap (invalid Pixmap parameter)</div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    <br></div></div>
    You got misunderstood it.<br>
    <br>
    You problem is here:<span class=""><br>
    <br>
       templ.width0 = (scrn->output_texture_width) ?<br>
                          scrn->output_texture_width :<br>
                          scrn->output_texture->width0;<br>
           templ.height0 = (scrn->output_texture_height) ?<br>
                           scrn->output_texture_height :<br>
                           scrn->output_texture-><wbr>height0;<br>
    <br></span>
    should be like:<br>
    <br>
       templ.width0 = scrn->output_texture->width0;<br>
       templ.height0 = scrn->output_texture->height0;<br>
    <br>
    <br>
    requested pixmap needs size according to that back buffer, not clips
    size, not scrn size.<br>
    <br></div></blockquote><div> Oh now I get it.</div><div>This will work fine but then we have to somehow support cliping in vdpau, any thoughts Christian?</div><div><br></div><div>Regards,</div><div>Nayan.</div><div>   </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div bgcolor="#FFFFFF" text="#000000">
    It was scrn size before, because the scrn size is same as back
    buffer size.<br>
    <br>
    Regards,<br>
    Leo<div><div class="h5"><br>
    <br>
    <br>
    <br>
    <blockquote type="cite">
      <div dir="ltr">
        <div class="gmail_extra">
          <div class="gmail_quote">
            <div><br>
            </div>
            <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
              The major problem causing resize corruption is now you
              request pixmaps to X dri3 ext. with wrong size.<span class="m_2199568300170825623gmail-"><br>
                <br>
                xcb_dri3_pixmap_from_buffer(sc<wbr>rn->conn,<br>
                                                 (pixmap =
                xcb_generate_id(scrn->conn)),<br>
                                                 scrn->drawable,<br>
                                                 0,<br>
                -                               scrn->width,
                scrn->height, buffer->pitch,<br>
                +                               buffer->width,
                buffer->height, buffer->pitch,<br>
                <br>
              </span></blockquote>
            <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><span class="m_2199568300170825623gmail-"><br>
              </span>
              <div>
                <div class="m_2199568300170825623gmail-h5">
                  <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      thus causing the corruption appears when resizing.<br>
                      <br>
                      <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                                                         
                        scrn->depth, 32,<br>
                                                          buffer_fd);<br>
                              xcb_dri3_fence_from_fd(scrn->c<wbr>onn,<br>
                        @@ -287,8 +314,6 @@
                        dri3_alloc_back_buffer(struct vl_dri3_screen
                        *scrn)<br>
                              buffer->pixmap = pixmap;<br>
                              buffer->sync_fence = sync_fence;<br>
                              buffer->shm_fence = shm_fence;<br>
                        -   buffer->width = scrn->width;<br>
                        -   buffer->height = scrn->height;<br>
                              xshmfence_trigger(buffer->shm_<wbr>fence);<br>
                        @@ -310,6 +335,7 @@ dri3_get_back_buffer(struct
                        vl_dri3_screen *scrn)<br>
                           {<br>
                              struct vl_dri3_buffer *buffer;<br>
                              struct pipe_resource *texture = NULL;<br>
                        +   bool allocate_new_buffer = false;<br>
                              assert(scrn);<br>
                        @@ -318,8 +344,30 @@ dri3_get_back_buffer(struct
                        vl_dri3_screen *scrn)<br>
                                 return NULL;<br>
                              buffer =
                        scrn->back_buffers[scrn->cur_b<wbr>ack];<br>
                        -   if (!buffer || buffer->width !=
                        scrn->width ||<br>
                        -       buffer->height != scrn->height) {<br>
                        +   /* This is normal case when our buffer is
                        smaller<br>
                        +    * than the screen this will be same for
                        external<br>
                        +    * texture<br>
                        +    */<br>
                      </blockquote>
                      Why do you change to size comparison to < from
                      != ?, it will be a waste when<br>
                      from larger window to small one.<br>
                    </blockquote>
                    Since I am setting  buffer width/height to texture
                    width/height which will be<br>
                    always greater than the screen size as vdpau
                    allocates slightly larger buffers<br>
                  </blockquote>
                  <br>
                </div>
              </div>
              Do you consider the VA-API case?<span class="m_2199568300170825623gmail-"><br>
                <br>
              </span></blockquote>
            <div>I did not give much thought to it, need to see how
              VA-API allocates buffer. </div>
            <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><span class="m_2199568300170825623gmail-">
                <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                    <br>
                  <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      +   if (!buffer || buffer->width <
                      scrn->width ||<br>
                      +       buffer->height < scrn->height)<br>
                      +      allocate_new_buffer = true;<br>
                      +<br>
                      +   /* If we were using a external texture buffer
                      and<br>
                      +    * the texture is not provided then we need a
                      new<br>
                      +    * buffer<br>
                      +    */<br>
                      +   if (buffer &&
                      buffer->is_external_texture &&<br>
                      +       !scrn->output_texture)<br>
                      +      allocate_new_buffer = true;<br>
                    </blockquote>
                    Can you elaborate this case? I cannot think of any.<br>
                    the texture is output surface, it should be always
                    provided, otherwise the<br>
                    where vdpau mixer render to.<br>
                  </blockquote>
                  vl_winsys_dri3.c is also used by va, and since we did
                  not change their<br>
                  code it will not set the output texture and would
                  expect us to send a<br>
                  buffer on which it will copy the output.<br>
                </blockquote>
                <br>
              </span>
              Yeah, we should definitely keep previous code path for
              VA-API.<br>
              <br>
              But this case is not fit to VA-API,  because there is no
              "is_external_texture" for it.<span class="m_2199568300170825623gmail-"><br>
                <br>
                +   if (buffer && buffer->is_external_texture
                &&<br>
                +       !scrn->output_texture)<br>
                +      allocate_new_buffer = true;<br>
                <br>
                <br>
              </span>
              This always is false for VA-API, right?</blockquote>
            <div>Yes I will remove it then. </div>
            <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><span class="m_2199568300170825623gmail-"><br>
                <br>
                <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                  <br>
                  <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                    You allocate a back buffer, but that cannot be used
                    as output surface by<br>
                    vdpau. right?<br>
                  </blockquote>
                  Yes. But this case won't occur anyway<br>
                </blockquote>
                <br>
              </span>
              Please remove that to avoid the confusion if that's not
              necessary.<span class="m_2199568300170825623gmail-"><br>
                <br>
                <br>
                <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                    but I just added it for safety as any<br>
                  state tracker would either set output texture or not
                  for all frames.<br>
                </blockquote>
                <br>
              </span>
              If that's the case, the handling here is still not
              correct, like being said, the back buffer won't be the
              target for output.<br>
              <br>
              Regards,<br>
              Leo
              <div class="m_2199568300170825623gmail-HOEnZb">
                <div class="m_2199568300170825623gmail-h5"><br>
                  <br>
                  <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <br>
                    Regards,<br>
                    Nayan<br>
                    <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                      <br>
                      Regards,<br>
                      Leo<br>
                      <br>
                      <br>
                      <blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
                        +<br>
                        +   /* In case of a single gpu we need to get
                        the<br>
                        +    * handle and pixmap for the texture that is
                        set<br>
                        +    */<br>
                        +   if (buffer &&
                        buffer->is_external_texture &&<br>
                        +       !scrn->is_different_gpu)<br>
                        +      allocate_new_buffer = true;<br>
                        +<br>
                        +   if (allocate_new_buffer) {<br>
                                 struct vl_dri3_buffer *new_buffer;<br>
                                 new_buffer =
                        dri3_alloc_back_buffer(scrn);<br>
                        @@ -332,6 +380,13 @@ dri3_get_back_buffer(struct
                        vl_dri3_screen *scrn)<br>
                                 vl_compositor_reset_dirty_are<wbr>a(&scrn->dirty_areas[scrn->cur<wbr>_back]);<br>
                                 buffer = new_buffer;<br>
                                 scrn->back_buffers[scrn->cur_<wbr>back]
                        = buffer;<br>
                        +   } else if (buffer->is_external_texture) {<br>
                        +      /* In case of different gpu we can reuse
                        the linear<br>
                        +       * texture so we only need to set the
                        external<br>
                        +       * texture for copying<br>
                        +       */<br>
                        +      buffer->texture =
                        scrn->output_texture;<br>
                        +      scrn->output_texture = NULL;<br>
                              }<br>
                              pipe_resource_reference(&textu<wbr>re,
                        buffer->texture);<br>
                        @@ -627,6 +682,19 @@
                        vl_dri3_screen_get_private(str<wbr>uct vl_screen
                        *vscreen)<br>
                           }<br>
                           static void<br>
                        +vl_dri3_screen_set_output_tex<wbr>ture(struct
                        vl_screen *vscreen, struct pipe_resource
                        *buffer,<br>
                        +                           uint32_t width,
                        uint32_t height)<br>
                        +{<br>
                        +   struct vl_dri3_screen *scrn = (struct
                        vl_dri3_screen *)vscreen;<br>
                        +<br>
                        +   assert(scrn);<br>
                        +<br>
                        +   scrn->output_texture = buffer;<br>
                        +   scrn->output_texture_width = width;<br>
                        +   scrn->output_texture_height = height;<br>
                        +}<br>
                        +<br>
                        +static void<br>
                           vl_dri3_screen_destroy(struct vl_screen
                        *vscreen)<br>
                           {<br>
                              struct vl_dri3_screen *scrn = (struct
                        vl_dri3_screen *)vscreen;<br>
                        @@ -744,6 +812,7 @@
                        vl_dri3_screen_create(Display *display, int
                        screen)<br>
                              scrn->base.set_next_timestamp =
                        vl_dri3_screen_set_next_timest<wbr>amp;<br>
                              scrn->base.get_private =
                        vl_dri3_screen_get_private;<br>
                              scrn->base.pscreen->flush_fron<wbr>tbuffer
                        = vl_dri3_flush_frontbuffer;<br>
                        +   scrn->base.set_output_texture =
                        vl_dri3_screen_set_output_text<wbr>ure;<br>
                              return &scrn->base;<br>
                      </blockquote>
                    </blockquote>
                  </blockquote>
                  <br>
                </div>
              </div>
            </blockquote>
          </div>
          <br>
        </div>
      </div>
    </blockquote>
    <br>
  </div></div></div>

</blockquote></div><br></div></div>