[Tts-project] Speech-dispatcher/pulseaudio integration

[Dennis Filder]

Hi,

recently I installed speechd-up (under Buster) which broke my
pulseaudio setup.  Some searching revealed that this is a known issue.
As this bugged me I cooked up something that works at least for me and
figured this might be worth sharing as it would allow speechd-up and
espeakup to work correctly out of the box with pulseaudio.

I'm posting here mainly to ask what the appropriate channel for
discussing this would be.  Most of the bugs for this (e.g. #859926)
assign this to speech-dispatcher, thus I'm writing here first, but I
disagree with that assessment.  In my opinion this is ultimately a
pulseaudio (pkg-pulseaudio-devel) issue, but it also affects
accessibility (pkg-a11y-devel) as the misconfiguration can leave blind
users stranded.  So where should I take this?

In short, the backbone of my approach was to make pulseaudio
publically expose the environment variables PULSE_SERVER and
PULSE_COOKIE and to implement a crude start/stop hook with systemd
ExecPost{Start,Stop} directives that makes pulseaudio trigger the
change-over as it starts/stops.  Other programs like
speech-dispatcher, speechd-up and espeakup can then, if properly
configured, make use of that infrastructure to release the audio
device and instead connect to pulseaudio once it is ready, but use
ALSA directly if pulseaudio does not start for some reason.

Regards,
D. Filder.