Re: gstreamermm option group does not init

[Marcus Brinkmann <marcus brinkmann ruhr-uni-bochum de>]

At Thu, 19 Jun 2008 19:38:49 -0400,
José Alburquerque <jaalburquerque cox net> wrote:
> > If there was a function g_option_group_get_parse_hooks, we could use
> > that to get the init_pre and init_post hooks of gstreamer, and replace
> > them with our own function that calls the gstreamer functions and does
> > some extra work.
> >
> > However, to really make this work, we also need to know the user data
> > for the hooks, which is set at g_option_group_new time.  For this user
> > data there is no getter and not even a setter function.  Both would be
> > needed to complete the wrapping of the parse hooks.
> >
> > So, in summary, what's needed from glib to complete this are:
> >
> > void g_option_group_get_parse_hooks (GOptionGroup *group,
> >                                      GOptionParseFunc *pre_parse_func,
> >                                      GOptionParseFunc *post_parse_func);
> > void g_option_group_set_user_data (GOptionGroup *group, gpointer user_data);
> > gpointer g_option_group_get_user_data (GOptionGroup *group);
>
> You can try.  I'm not sure what the result would be.  If submitted and 
> they like the idea, I think they might accept a patch.  For now, there 
> are a few other things that need attention, but trying to enhance 
> gstreamermm by making it deal appropriately with GStreamer's 
> GOptionGroup would certainly be nice.

I hate my idea already.  The same user data is also used for the error
hook (not a big problem, but wrapping it requires yet another getter
function), and also for each G_OPTION_ARG_CALLBACK option, and we
definitely don't want to wrap each of those.  Bummer.  The proper
solution is to have different user data for different purposes, but
that is a slightly bigger change on which the glib people might have
some opinions.  This will probably slow things down.

So, here is the work around:  After parsing options, I call:

  int argc_hack = 0;
  char *argv_hack = NULL;
  Gst::init (argc_hack, argv_hack);

Which completes initialization.  By the way, I didn't get a response
with regards to the discrepancy of initialization calls in Gst::init
vs Gst::init_check.  Not that I know much about what these functions
do, but init and init_check look kinda out of sync.

I totally agree that getting GOption integration to work properly is
not a high priority item.  I'll wait a couple of days, and if no other
ideas pop up I will add the end result of the discussion plus the work
around to the bug report for reference at a later time.

Thanks,
Marcus