[sane-devel] buttons-daemon for SANE genesys scanner

[m. allan noah]

On Sat, Jan 10, 2009 at 3:56 PM, Pierre Willenbrock
<pierre at pirsoft.dnsalias.org> wrote:
> m. allan noah schrieb:
>> On Sat, Jan 10, 2009 at 2:55 PM, Pierre Willenbrock
>> <pierre at pirsoft.dnsalias.org> wrote:
>>> m. allan noah schrieb:
>>>> On Sat, Jan 10, 2009 at 1:11 PM, Pierre Willenbrock
>>>> <pierre at pirsoft.dnsalias.org> wrote:
>>>>> m. allan noah schrieb:
>>>>>> On Fri, Jan 9, 2009 at 5:02 PM, Pierre Willenbrock
>>>>>> <pierre at pirsoft.dnsalias.org> wrote:
>>>>> [...]
>>>>>>> Another question: Is it okay to only look at the hardware state if the
>>>>>>> frontend asks for the state of the option? That way shorter presses can
>>>>>>> be lost, if the frontend does not poll often enough.
>>>>>> In my case, the scanners are smart enough to buffer the button presses
>>>>>> for 3 seconds, and I effectively read the status of all the buttons
>>>>>> from the scanner every time you ask for the value of the first option.
>>>>>> So, as long as the front-end reads within that 3 second window, no
>>>>>> presses are lost.
>>>>>>
>>>>>> If your machines dont buffer, then you might need a thread just to
>>>>>> read the status really quickly? Do you know how frequently the windows
>>>>>> driver reads the buttons?
>>>>> The genesys chips don't help at detecting hardware buttons. The only
>>>>> sensor logic in the chip is for the home-sensor. The rest is GPIO.
>>>> Is it possible for a user to press and release the button and have the
>>>> driver 'lose' it if you dont read the GPIO pins during the 'pressed'
>>>> inteval?
>>> Exactly like that. And the windows driver has the same problem.
>>
>> Ahh, I love cheap scanners :) How fast would you have to poll the
>> scanner to get short presses? 10Hz?
>
> In my exerience, at 10Hz it is hard to press short enough for the press
> to not get noticed. 2Hz is too slow, 8Hz is okay, and above 12Hz it is
> impossible to lose presses, even when hammering the button like mad. At
> least, you cannot be sure if the software lost the press or you didn't
> press the button properly.
>
> The shortest single presses i managed were 20ms, averaging at 40ms.
> Normal presses are 70-110ms. Now one could calculate the odds for a
> missed button press, but i will not go there.

I would expect those numbers to be similar to other "unbuffered"
scanners, so lets just say that the front-end should loop over the
values of the buttons at 10Hz. It is alot easier for the frontend to
read more frequently than for backends to add threads/IPC. And, if
reading the button status is a heavy weight operation, the backend
could return cached data occasionally.

allan
-- 
"The truth is an offense, but not a sin"