speechd-discuss
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RFC: Disable the ALSA driver by default.


From: Luke Yelavich
Subject: RFC: Disable the ALSA driver by default.
Date: Fri, 19 Sep 2014 08:09:49 +1000

On Sat, Aug 30, 2014 at 01:23:41AM AEST, Trevor Saunders wrote:
> On Fri, Aug 29, 2014 at 04:14:00PM +1000, Luke Yelavich wrote:
> > Hey folks.
> > One thing I have seen mentioned in many a discussion about 
> > speech-dispatcher in the recent past is about the state of the ALSA output 
> > driver, and its tendency to crash. I am wondering whether it is worth 
> > considering disabling the ALSA driver by default for now, and only building 
> > it if the user explicitly enables it at build time. For now, we have the 
> > libao driver as an alternative, and libao can be configured to use ALSA 
> > directly. Longer term, it would be better to have a direct ALSA driver, but 
> > until someone can take the time to fix its bugs, I'd rather users have a 
> > stable ALSA experience for now.
> > 
> > I haven't tested the ALSA driver myself recently, but I used to have 
> > regular crashers when I did use it, and that was on a single CPU/single 
> > core system. Given that the ALSA driver issues are likely threading 
> > related, I dare say it is much worse on multi-CPU/multi-core systems.
> 
> istr it was actually some sort of buffer under / over run.  Anyway I
> don't think its gotten much worse over the years as I've gone from
> single / double core machines to 4 or 8 core ones.
> 
> > This is something I am considering for the 0.8.1 release. Any thoughts, or 
> > alternative suggestions are welcome.
> 
> I think the goal of getting people to not use the alsa driver when they
> have alternatives makes sense.  However I wonder if it wouldn't make
> more sense to just change our defaults to prefer libao to alsa.  I think
> we could probably just switch the order of the alsa and libao checks in
> configure.ac to accomplish that.  I do wonder though why it is that we
> don't add all the audio backends we find to default_audio_method given
> it supports more than one at least in theory.

This makes sense, so for 0.8.1 I'll switch the order for now, and we can adjust 
things further in the future.

Luke



reply via email to

[Prev in Thread] Current Thread [Next in Thread]