WIP audio in server

[Luke Yelavich]

On Fri, Feb 26, 2016 at 09:34:51AM AEDT, Jeremy Whiting wrote:
> If anyone has other feedback on the patch itself I would appreciate it.

Ok, some comments, this is a bit of a brain dump, as its almost EOD here...

I see you create a thread in src/server/audio.c in speechd_audio_init, which 
sets everything up, and then essentially waits for audio_close_requested to be 
set to TRUE, which is not done anywhere in code. Can I assume this is a 
WIP/TODO part of this work?

In the audio thread method, you also add the fd to the GLib main loop to listen 
for activity from modules, and then subsequently add more descriptors for 
modules. I guess you are trying to have all audio related stuff be done in that 
audio thread? If so, I don't think its correct, given my understanding of GLib 
main loops. GLib main loops poll and respond to fd activity in the thread they 
are in, so at the moment you have a thread sitting there doing nothing, since 
nothing changes the state of the semaphore or audio_close_requested. Meanwhile 
all audio activity is in the main thread running of the main thread's main loop.

If you want to use the same sort of mechanism for audio in another thread, you 
have to create a new main context, and create a new main loop with that 
context, and add all descriptors to poll based on that main loop and context. 
YOu would then use g_main_loop_quit in the same thread to close down and clean 
up.

I can have a crack at updating the code to use a separate main loop as per 
above if you like, although I am not sure what you are planning to do with that 
semaphore, as in where you think it will be used.

Luke