Re: FIX: USB Timeout problems (Clie, etc)

[Matt Davey <mcdavey mrao cam ac uk>]

On Mon, 2004-09-06 at 04:52, JP Rosevear wrote:
> On Sun, 2004-09-05 at 13:31 +0100, Matt Davey wrote:
> > After poking around the pilot-link / gnome-pilot source, it
> > appears that the timeout sent to pilot-link should be in
> > milliseconds, whereas gnome-pilot thinks it is in seconds.
> 
> I don't think so, if you look at serial.c:pi_serial_accept you will see
> it multiplies ps->accept_to (set in pi_accept_to) by 1000, converting in
> to milliseconds.

I see that multiply, but right before that multiplication there is a
call to _poll that uses the original (milliseconds) timeout.  On the
device I was testing (a Clie TJ35) the timeout failure was in the poll,
not in the socket_init or handshake code.  It's a close run thing: 2ms
fails most of the time, 5ms succeeds most of the time.

This does push the problem from gnome-pilot into pilot-link, as I would
imagine the best thing is for pilot-link to multiply the timeout by 1000
before the poll.

Matt
-- 
Matt Davey  		I stopped believing in Santa Claus when I was six. 
mcdavey mrao cam ac uk  Mother took me to see him in a department store and
                        he asked for my autograph.  -- Shirley Temple Black