[Spice-devel] [PATCH 09/15] worker: remove cursor channel asserts

[Jonathon Jongsma]

On Mon, 2015-11-09 at 12:25 -0500, Marc-André Lureau wrote:
> 
> ----- Original Message -----
> > > 
> > > From: Marc-André Lureau <marcandre.lureau at gmail.com>
> > > 
> > > ---
> > >  server/cursor-channel.c | 6 +++---
> > >  server/red_worker.c     | 2 --
> > >  2 files changed, 3 insertions(+), 5 deletions(-)
> > > 
> > > diff --git a/server/cursor-channel.c b/server/cursor-channel.c
> > > index 6d648b1..fc3a057 100644
> > > --- a/server/cursor-channel.c
> > > +++ b/server/cursor-channel.c
> > > @@ -219,7 +219,7 @@ static void
> > > put_cursor_pipe_item(CursorChannelClient
> > > *ccc, CursorPipeItem *pipe_
> > >          return;
> > >      }
> > >  
> > > -    spice_assert(!pipe_item_is_linked(&pipe_item->base));
> > > +    spice_return_if_fail(!pipe_item_is_linked(&pipe_item
> > > ->base));
> > >  
> > >      cursor_item_unref(pipe_item->cursor_item);
> > >      free(pipe_item);
> > > @@ -277,7 +277,7 @@ static void
> > > red_marshall_cursor_init(RedChannelClient
> > > *rcc, SpiceMarshaller *bas
> > >      SpiceMsgCursorInit msg;
> > >      AddBufInfo info;
> > >  
> > > -    spice_assert(rcc);
> > > +    spice_return_if_fail(rcc);
> > >      cursor_channel = SPICE_CONTAINEROF(rcc->channel,
> > > CursorChannel,
> > >      common.base);
> > >  
> > >      red_channel_client_init_send_data(rcc,
> > > SPICE_MSG_CURSOR_INIT, NULL);
> > > @@ -410,7 +410,7 @@ static void
> > > cursor_channel_release_item(RedChannelClient
> > > *rcc, PipeItem *item, i
> > >  {
> > >      CursorChannelClient *ccc = RCC_TO_CCC(rcc);
> > >  
> > > -    spice_assert(item);
> > > +    spice_return_if_fail(item);
> > >  
> > >      if (item_pushed) {
> > >          cursor_channel_client_release_item_after_push(ccc,
> > > item);
> > 
> > Considering the default behavior of spice_return_if_fail is
> > aborting
> > like spice_assert I don't see the point of these changes.
> 
> The point is we should have less assert() in the code and be more
> resilient to memory allocation failures or wrong arguments received
> etc. Then the code should no longer assert() in return_if_fail().


I'm becoming convinced that we should get rid of all of these spice_*
equivalents to the glib logging / checking functions (spice_return_if,
spice_warning, etc). The fact that spice_return_if_fail() aborts but
g_return_if_fail() does not abort creates a potential for
inconsistencies and incorrect usage.


> 
> > 
> > > diff --git a/server/red_worker.c b/server/red_worker.c
> > > index f70ee9b..10dfd8b 100644
> > > --- a/server/red_worker.c
> > > +++ b/server/red_worker.c
> > > @@ -301,8 +301,6 @@ struct RedGlzDrawable {
> > >  pthread_mutex_t glz_dictionary_list_lock =
> > > PTHREAD_MUTEX_INITIALIZER;
> > >  Ring glz_dictionary_list = {&glz_dictionary_list,
> > > &glz_dictionary_list};
> > >  
> > > -#define NUM_CURSORS 100
> > > -
> > >  typedef struct RedWorker {
> > >      pthread_t thread;
> > >      clockid_t clockid;
> > > --
> > > 2.4.3
> > > 
> > 
> > I'll split this last hunk in
> > 
> > "worker: remove unused NUM_CURSORS define" (and I'll ack this)
> > 
> > Frediano
> > _______________________________________________
> > Spice-devel mailing list
> > Spice-devel at lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/spice-devel
> > 
> _______________________________________________
> Spice-devel mailing list
> Spice-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/spice-devel