[Spice-devel] [PATCH spice-gtk v3 1/6] display-gst: check codec type before creating decoder

Frediano Ziglio fziglio at redhat.com
Fri Jun 9 15:52:29 UTC 2017


> 
> 
> > On 8 Jun 2017, at 16:47, Christophe Fergeau <cfergeau at redhat.com> wrote:
> > 
> > On Thu, Jun 08, 2017 at 10:39:53AM -0400, Frediano Ziglio wrote:
> >>> 
> >>> On Thu, Jun 08, 2017 at 03:55:11PM +0200, Victor Toso wrote:
> >>>>> 
> >>>>> Not really worth all that discussion, your patch is ok :)
> >>>> 
> >>>> Ah, I liked the discusison mostly because this would be a spice <->
> >>>> spice-gtk communication bug and I'm not sure either if we should go for
> >>>> CRITICAL or WARNING messages in such cases.
> >>>> 
> >>>> I mean, in the past, for spice server, there was discussion about when
> >>>> to assert(). The result was that only when it would be a bug inside the
> >>>> component (never on client's input/configuration for instance)
> >>>> 
> >>>> Anyway, thanks for the discussion :)
> >>> 
> >>> To tie this in with the logging discussion, you can read the latest
> >>> comment from ebassi in this blog post about g_log_structured ;)
> >>> 
> >>> https://blog.gtk.org/2017/05/04/logging-and-more/
> >>> 
> >>> Christophe
> >>> 
> >> 
> >> Wrong thread? Looks like the structured logging would be more
> >> appropriate in Chistophe D thread.
> > 
> > Nope, the comment I refer to
> > 
> > «  Warnings (and criticals) are meant to be used for programmer error
> > and internal state consistency checks — and the expectation is that
> > anything that follows a warning is undefined behaviour. Personally, I
> > prefer asserting much more than just warning, because that catches real
> > programmer errors, but warnings usually have more understandable
> > messages. »
> 
> I see a fair number of warnings when starting spicy, notably a number of
> assertions.
> 
> (spicy:3327): GSpice-CRITICAL **: display_handle_stream_data: assertion 'st
> != NULL' failed
> (spicy:3327): GSpice-CRITICAL **: display_handle_stream_data: assertion 'st
> != NULL' failed
> 
> Many such messages, making log output hard to read (even on the traces
> branch), but not annoying enough that we fix them. Better than a crash,
> though.
> 
> Christophe
> 

This message should be printed if data for a stream not present are
received or bad handling of events.

Do you have some capture of these? Maybe is a server issue.
What were you doing?

Frediano


More information about the Spice-devel mailing list