Re: ethlcd driver improvements

On 2011-03-28 21:50, Markus Dolze wrote:
> There is no way for a driver to signal the core to exit once the driver
> is running, so using exit() is the only option if you want to quit LCDd.
> However no cleanup of memory is performed then.
I also was searching the way of clean exit from driver, but I can see
that returning -1 works only in init function. So I used the same
method like ie. in hd44780-ftdi.c. Do you think it would be better if
I free some (PrivateData?) structures before? I know it would be a lot
better if I do some kind of clean exit - but if I am correct, it
also needs some changes not only in my driver.

> What may be reasons why a send / receive fails? If it is the TCP
> connection that broke maybe reconnecting may be an idea?
 From experience with my device I can tell that it sometimes hung (hardware).
Then even ping packet are not replied. The only way to fix such a state
is to cycle device's power. But this situation is very very rare - two-three
times per year, when the device is working fine the rest of time (working
non-stop).
So I also thought about reconecting, but it doesn't help in this case,
especialy because if I even successfuly reconnect after device start, then
it won't work without executing commands in init function.

> Not related to you patch: If your ethlcd device can handle more than one
> byte of input you may take advantage of the small flush buffer available
> now. Have a look at hd44780-lcd2usb.c on how to make use of it.
I asume you're talking about tx_buffer struct. I'll think about it if
I change the device protocol, because now it unfortunately cannot handle
more data in one packet. Thank you for your tip.

regards,
-- 
Mariusz Białończyk
jabber/e-mail: [email protected]
http://manio.skyboo.net
_______________________________________________
LCDproc mailing list
[email protected]
http://lists.omnipotent.net/mailman/listinfo/lcdproc