Speechd sd_generic async Limitations

[Hynek Hanke]

Thomas Lloyd wrote:
> I am at the end of the first development stage of the SAPI server. It
> runs well but i have been experimenting with a purge feature on the
> server side. This allows me to cancel any existing speech and then
> accept a new phrase to be synthesised. 
>   
Great, this is very important.
> My question is around this is there the ability from sd_generic to
> control the start and stop of speech messages or does the server have to
> remain dumb and just read everything it is passed.
>   
No, the server makes very good use of start/stop functions
either explicitly (the client calling for a stop) or implicitly
when handling messages of different priorities.

In sd_generic, for this to work, the very important things
are a) the binary must not exit before speech is stopped
b) it must stop speaking (producing audio output)
on receiving SIGKILL. This is limiting for sure, but it is
the simple way how sd_generic is designed and has been
useful with many simple command line speech tools.

If you can't make use of this, the next step would be
to design a proper SD module or to extend sd generic
with something, e.g. sending another signal than SIGKILL
to request a stop and/or not require the binary termination
and have some other method to inform the generic module
that speech has stopped.

Have a nice day & thanks for your work,
Hynek Hanke