[Spice-devel] [PATCH spice-server 3/3] red-worker: Reuse code to process display command

[Frediano Ziglio]

> 
> On Tue, Jan 24, 2017 at 07:26:15AM -0500, Frediano Ziglio wrote:
> > > 
> > > On Tue, Jan 10, 2017 at 03:59:38PM +0000, Frediano Ziglio wrote:
> > > > Code to read and process display commands were the same
> > > > so use a common function for better reuse.
> > > > 
> > > > Signed-off-by: Frediano Ziglio <fziglio at redhat.com>
> > > > ---
> > > >  server/red-worker.c | 38 ++++++++++++++++++--------------------
> > > >  1 file changed, 18 insertions(+), 20 deletions(-)
> > > > 
> > > > diff --git a/server/red-worker.c b/server/red-worker.c
> > > > index 394a935..49ec8e8 100644
> > > > --- a/server/red-worker.c
> > > > +++ b/server/red-worker.c
> > > > @@ -169,6 +169,19 @@ static RedDrawable *red_drawable_new(QXLInstance
> > > > *qxl)
> > > >      return red;
> > > >  }
> > > >  
> > > > +static gboolean red_process_surface_cmd(RedWorker *worker,
> > > > QXLCommandExt
> > > > *ext, gboolean loadvm)
> > > > +{
> > > > +    RedSurfaceCmd surface_cmd;
> > > > +
> > > > +    if (red_get_surface_cmd(&worker->mem_slots, ext->group_id,
> > > > &surface_cmd, ext->cmd.data)) {
> > > > +        return FALSE;
> > > > +    }
> > > > +    display_channel_process_surface_cmd(worker->display_channel,
> > > > &surface_cmd, loadvm);
> > > > +    // do not release resource as is released inside
> > > > display_channel_process_surface_cmd
> > > 
> > > I'd use this as an opportunity to improve that comment to something
> > > like:
> > > 
> > > // do not release resource ('release_info_ext') as it will be released
> > > inside
> > > // display_channel_surface_unref() once the last reference is dropped
> > > 
> > 
> > I think this comment is half an improvement and half worsening.
> > The improvement is "('release_info_ext')", the worsening
> > display_channel_surface_unref.
> > Is up to display_channel_process_surface_cmd how to release the resource,
> > for instance can free resource directly on double create or surface too
> > big.
> > 
> > Perhaps
> > 
> >   // do not release resource ('release_info_ext') as
> >   // display_channel_process_surface_cmd() will take care of
> 
> Ah, different perspective on this. I think what you want to say is that
> display_channel_process_surface_cmd() will take ownership of the
> resource (ie we no longer need to care about it).
> 
> My point was more along the line that if you look at
> display_channel_process_surface_cmd() code, there is nothing releasing

This is wrong.
Just you should not look at how display_channel_process_surface_cmd is
implemented, this is up to DisplayChannel object, not red worker.
Unfortunately you cannot declare ownership in C so this should be
documented for display_channel_process_surface_cmd (as the comment does,
perhaps would be worth adding a comment in DisplayChannel header/code).

> resources, so the code aluded to in the comment does not exist there!
> They are only released once the last reference is dropped, in
> display_channel_surface_unref().
> 
> So I'm  fine with something like "display_channel_process_surface_cmd()
> takes ownership of 'release_info_ext', we don't need to
> release it ourselves"
> 

Yes, this is fine.

> Christophe
> 

Frediano