Re: [PATCH]input: Change timer function to workqueue for gpio_keys driver

[Phil Carmody <ext-phil.2.carmody@nokia.com>]

On Thu, 2009-06-25 at 17:48 +0200, ext Alek Du wrote:
> On Thu, 25 Jun 2009 23:42:08 +0800
> Jani Nikula <ext-jani.1.nikula@nokia.com> wrote:
> 
> > On Thu, 2009-06-25 at 17:09 +0200, ext Alek Du wrote:
> > > On Thu, 25 Jun 2009 23:05:55 +0800
> > > Jani Nikula <ext-jani.1.nikula@nokia.com> wrote:
> > > > Actually scrap that, the input layer already ignores events with no
> > > > state changes, right?
> > > > 
> > > Yes, correct. I just want to reply your previous mail, but seems you find that. :-)
> > 
> > The point about your patch breaking debouncing is still valid, though.
> > 
> >
> How? If IRQ triggered then the delay work scheduled, after debouncing time, in work, it checks GPIO pin state again,
> if pin is active, send "1" to input layer -- key pressed, if de-active, send "0" -- no event. 
> 
> I really did test on my board, could you also try it out?

This is not a matter of testing, this error can be seen simply by
algorithm analysis - that's how Jani and I discovered the problem in the
first place.

If you stopped calling the delay after the first transition "debouncing
time" and simply called it a "delay" you might more easily see that it
does *no* debouncing at all. Imagine putting noise on the line
constantly - the original code's timer would never expire. Your timer
will expire after a delay, and while the line is still toggling
frantically - you've not debounced.

Please investigate the meaning and implications of "debouncing" before
claiming your code does it.

Phil