[sane-devel] scanbuttond - project dead?

[Wilhelm Meier]

Am Mittwoch 19 November 2008 schrieb m. allan noah:
> On Wed, Nov 19, 2008 at 12:52 PM, Wilhelm Meier 
<wilhelm.meier at fh-kl.de> wrote:
> > Am Mittwoch 19 November 2008 schrieb m. allan noah:
> >> On Wed, Nov 19, 2008 at 10:15 AM, Wilhelm Meier
> >
> > <wilhelm.meier at fh-kl.de> wrote:
> >> > Hi,
> >> >
> >> > does anybody know if the scanbuttond project ist still active?
> >>
> >> I dont think so.
> >>
> >> > Why is this project using different backends?
> >>
> >> because that is how the author decided to do it, it is not part
> >> of SANE. Lately, SANE has begun to add support for buttons, but
> >> not all backends do it yet, and the 'frontend' daemons (in our
> >> 'experimental' tree) likely will need some work.
> >
> > Ok, I've begun some work on this.
> >
> > The daemons in the experimental tree miss some features:
> >
> > 1) they can't coexist with saned because they hold the devices
> > open while polling and don't "know" about saned
>
> correct. I suppose you could have them connect to saned instead of
> directly.

well, I don't think this will work (but I did not test it), because if 
the button-daemon uses saned to poll the devices, the device also 
remains opened via saned. So, saned forks a new child for the 
button-daemon and this child holds the device open. If another scan 
request (say scanimage via net-backend) arrives, saned forks a new 
child, which tries to open the same device and the gets device-busy. 
So the same situation as before.

The algorithm must give saned sort of priority over the polling. My 
proposal is to stop the polling of the button-daemon, if saned gets 
called via the network. To be not too intrusive to the saned-code, I 
made a wrapper for saned: this wrapper is called from inetd, then 
sends signal to the button-daemon to stop polling, forks the saned 
and waits for it, and then reactivates the polling.

>
> > 2) they lack a hal/dbus interface
>
> correct. there was some discussion of this (search the archives)
> but I don't recall if any code was produced for this or not.

Ok, found the discussion but no code ;-(

>
> > Actually I have 1) functional and I'm working on 2).
> >
> > The problem right now is, that there are only few backends, which
> > enumerate the buttons as sane-options. I have access to an epson
> > 1670, an avision av122 and a fujitsu fi-6130 for testing, but the
> > only one I can read the buttons as option is the fi-6130. Is
> > there any further work on this on the backend side?
>
> The fujitsu backend is probably the most complete regarding
> buttons. It is a fairly new part of the draft sane standard
> however, so it will take some time for other backends to implement
> support.
>
> allan



-- 
Wilhelm