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

Re: [SpeechIO-50] EOFs (was: Re: [SpeechIO-12] speechd v0.39)

> Have you ever tested what you just said ?  
> I just did.  >  had the same (non)effect as >>.  

when I do:
xterm1: mkfifo spk
xterm1: cat spk
xterm2: cat > spk
xterm2: sdf
xterm2: sdf
xterm2: ^D

the first cat exits when I do the ^D in the second xterm...it always behaves
this way for me.

> Yes, an implimentation as a kernel module could be more robust and solid.
> But from what I see, the *only* difference is when something opens to
> over-write instead of append, and starts causing dropped lines.  Yes, it's
> a flaw, but I think not a major one, and not significant enough to loose
> the power/speed of development/ease of maintenence of Perl, and add the
> extra complications of building, maintaining, and installing a kernel
> module.  

It wouldn't really stop our use of perl -- we'd jsut be reading from the
kernel module implemented device file instead of a FIFO.  That'd be the
only difference.

I've already done a first implementation of the kernel module -- it's at:


under projects multipipe.  It needs some polishing, but it gets around
the immediate problem with the EOF that FIFO's don't hanlde the way we want.
as far as blocking writers so only 1 can write at a time, that's still
something I need to work out.

> I figure we'll end up agreeing on something eventually.
> Could you do the input w/ a really small kernel module that just sent
> output to another program -- basically consisting of most of what speechd
> is ?  So we don't loose Perl....
> How were you thinking about doing this ?

see multipipe -- I've already started it (I though I announced it to you
guys, but I guess I might not have).  I think the only change that would
need to be made is to have speechd read from the multipip_out device file
instead of the FIFO.


"From a certain point onward there is no longer any turning back. That is the
point that must be reached." 
    -- Kafka
mortis@voicenet.com                            http://www.voicenet.com/~mortis

** (debug) Blackmail[57992] would have rejected this message for 1 reasons. **