[Sugar-devel] gst-plugins-espeak port to gstreamer 1.0

[Aleksey Lim]

On Sun, Aug 19, 2012 at 12:48:40PM +0200, Daniel Narvaez wrote:
> On 19 August 2012 09:02, Aleksey Lim <alsroot at sugarlabs.org> wrote:
> >> I still think returning an arbitrary
> >> size buffer is wrong and should be fixed at some point.
> >
> > The original buffer size should not matter, if I got gst docs right,
> > only `offset` and `size` does matter
> > http://gstreamer.freedesktop.org/data/doc/gstreamer/head/gstreamer/html/gstreamer-GstBuffer.html#gst-buffer-new-wrapped-full
> >
> 
> That's my understanding too. But we are not passing gst_espeak_create
> size as size to gst_buffer_new_wrapped_full, because of this
> 
>     switch (g_atomic_int_get (&self->track)) {
>     case ESPEAK_TRACK_WORD:
>     case ESPEAK_TRACK_MARK:
>         size_to_play = events (self, spin, size_to_play);
>         break;
>     default:
>         size_to_play = whole (spin, size_to_play);
>         break;
>     }

Yeah, that might be a problem.
In any case, that seems to work in gst-0.10/1.0.

I don't want fixing this issue in current code, it is overbloated with
design that was not fully implemented and seems to be an overkill
at the end (it was a try to process continues synch, but, it seems to be
more reasonable to process text synch separated with gst stop/start
sessions). I started "simplify" branch for that reason.

> >> >> gst-launch-1.0 espeak text="Hello world" ! autoaudiosink
> >
> > I compiled gst from git and got broken `autoaudiosink` behaviour. For
> > example
> >
> >     gst-launch -v -m audiotestsrc ! audioconvert ! audioresample ! autoaudiosink
> >
> > fails with the same error as
> >
> >     gst-launch-1.0 espeak text="Hello world" ! autoaudiosink
> >
> > So, it seems to gst development code fails.
> 
> I have a few days old gstreamer tree where autoaudiosink seems to be
> fine. I need to apply the following to make espeak work though (it was
> in my patch, http://lists.sugarlabs.org/archive/sugar-devel/2012-August/038891.html).
> Can we get this in?
> 
> diff --git a/src/gstespeak.c b/src/gstespeak.c
> index 4560512..1dd097e 100644
> --- a/src/gstespeak.c
> +++ b/src/gstespeak.c
> @@ -152,7 +152,7 @@ static void gst_espeak_init (GstEspeak * self) {
>              "rate", G_TYPE_INT, espeak_get_sample_rate (),
>              "channels", G_TYPE_INT, 1, NULL);
> 
> -    gst_base_src_set_format (GST_BASE_SRC (self), GST_FORMAT_DEFAULT);
> +    gst_base_src_set_format (GST_BASE_SRC (self), GST_FORMAT_TIME);
>  }
> 
>  static void gst_espeak_finalize (GObject * self_) {

Original intention was a try to have bytes buffers, but it seems
something that something was implemented wrong. Time buffers seems to
work better, so I' pushed this code.

-- 
Aleksey