ogg dev - May 2013 - libfishsound/liboggz: How to properly set e_o_s marker when encoding?

Perhaps I am really dense, but after writing an app to encode multi-stream
audio files using libfishsound and liboggz, I cannot figure out a way to
reliably determine the which packet is the FINAL packet output from the
encoder, so I can set the e_o_s flag on it.

What is the magic for determining WHEN to set the e_o_s marker on a packet
in the FishSoundEncoded callback function?

Even though I know when I write the last audio buffer with
fish_sound_encode_float(), I don't know how many packets my
FishSoundEncoded callback will receive as a result.

Therefore, how can I know which packet is indeed the last and therefore
set the e_o_s flag appropriately?

There seems to me no mechanism within FishSound to do this.

After thinking about it, I decided a good approach might be:

    a. fish_sound_flush()
    b. Let the callback queue all remaining packets via oggz_write_feed()
    c. Manually set the e_o_s packet on the final packet in each oggz
internal stream queue.
    d. Call oggz_run() to flush the queues to the output file.

Of course, this requires creating my own oggz function in order to gain
access to the internal per-stream queues.

But before I go that route:

   a. Is this the best way to accomplish e_o_s handling?

   b. Is the un-exported oggz_set_eos() function supposed to be used for
this purpose? It's there along with oggz_get_eos(), but is undocumented
and un-exported.

Any helpful insights will be much appreciated.

Regards,
Bob Ingraham