Probable flush update problem in source

[luke]

On Tue, 17 Jun 2008, Samuel Thibault wrote:

> > As for the flush change: is that ready for testing, or is it still in the 
> > state it was when I accidentally tried it before?
> 
> Nothing happened yet because we weren't able to reproduce the issue.
> We'll try to understand what happens now that we know it's the flush
> move.

Okay.  I am using a DEC Express (old) running on a serial port of course.  
I can rebuild this, and give you a full kernel config if you want it; 
but I'll wait to hear whether there are any specific tests you want me to 
run (E.G. modular/non-modular, different premption models, etc.).

The config from when it happened last time is here:

http://suntzu.tacticus.com/config-2.6.18-speakup4

Having tested it again, I note that on all of these git speakup versions 
from the last week, after it says "DECTalk Express found" (first spoken 
message, now that I have changed speakup_synth to speakup.synth), there is 
a long delay, and we are into network card detection before it actually 
starts talking again.

That is important here, because for the flush modified version, I am 
unable to shut the speech up right away from my USB keyboard, and by the 
time I can, the thing has choked, and even after it does shut up, it 
takes me nearly 45 seconds of hitting review keys (between numpad 2 and 
numpad enter) before the speech will come back.  When it does, of course, 
it is sluggish as I have described.

Just for fun, I entered speakup help, and every key pressed was exhibiting 
that half second or so delay.

Not so outside of speakup help, when at least the non-speakup keys would 
speak at their normal responsiveness level.

HTH

Luke