[gst-devel] Implementation Notes

[Ronald S. Bultje]

Hi Ryan,

On Thu, 2005-02-10 at 15:44, Ryan Norton wrote:
> 1.  The most annoying one - the GstStreamInfo associated with the
> video/playbin didn't have an "object" property - i.e. it would crash when I
> called GST_PAD_REALIZE on it afterwards... after googling forever, I found
> out from an old totem diff that there is a "pad" property that it has that
> does the same thing (?), so
> 
> my workaround was
> ----
>                 g_object_get (info, "object", &pad, NULL);
> ----
> turned into
> ----
>             //Maybe newer gstreamer 0.8+ is supposed to have "object"...?
>             //but a lot of old plugins still use "pad"?
>             pspec = g_object_class_find_property (
>                         G_OBJECT_GET_CLASS (info), "object");
> 
>             if (!pspec)
>                 g_object_get (info, "pad", &pad, NULL);
>             else
>                 g_object_get (info, "object", &pad, NULL);
> ----
> 
> 2.   After setting the playbin to GST_STATE_READY any calls on it afterwards
> (changing the state, changing the uri, etc.) would trigger a ton of
> assertions until the program crashed.  My workaround was just to create a
> new playbin and correspoding video and audio sinks whenever a new file was
> loaded...  not the best solution, but it worked :).

Both sound like you use a fairly old gst-plugins release. Not bad, but
playbin is still extremely young so we tend to have some obvious bugs in
our first version that are long-time fixed in the newest releases...
Please retry with latest releases.

> 3.  When pausing the playbin and seeking immediately in it,querying the
> current position always returned the old postion (most of the time the last
> position before pause).  Basically what I did was cache the current position
> right before pause and everytime it was set afterwards until the playbin was
> set back to a play state and return the cached postion from my get position
> call until it was played again.

Yeah, known problem, depends on the media type actually, we never
defined how that should work so most elements just do an actual seek in
their next iteration. This will be done differently in 0.9. I'll try to
fix some elements in 0.8, although I can't fix the problem in general.

Thanks,

Ronald

-- 
Ronald S. Bultje <rbultje at ronald.bitfreak.net>