[sane-devel] SANE_Device issue

[Perry Hutchison]

Ralph Little <skelband at gmail.com> wrote:

> Hi,
>
> On Wed, May 28, 2025 at 9:32???PM Ben Olden-Cooligan <ben.cyanfish at gmail.com>
> wrote:
>
> > It's very hacky but I get around this in NAPS2 by creating a temporary
> > config dir with only the one backend for the specified device.
>
> It certainly sounds a bit janky, but I didn't think of this. It might
> be the only option to get the information without an API change.

I haven't looked into the code, but just thinking about how something
like this "has to" work I would expect sane_get_devices() to do, roughly:

    for b in backends
    {
        ...
        b->get_devices();
        ...
    }

IOW each backend has to already implement a specified interface that
probes for devices supported by that backend and (somehow) loads the
results into SANE_Device.

Is there some reason why a program like xsane, when told which
backend to use, cannot just call that one backend's get_devices()
function directly -- at the same point where it would otherwise
call sane_get_devices()?  This is effectively what Ben's solution
accomplishes by limiting "backends" to the one of interest, but
without having to create and populate a scratch directory somewhere.

> The other option actually is just to replace that descriptive
> information with something nondescript like "None" or "Unknown".
> Doesn't look all that great but it would avoid the delay on startup.
>
> Cheers,
> Ralph
>
> > https://github.com/cyanfish/naps2/blob/master/NAPS2.Sdk/Scan/Internal/Sane/SaneScanDriver.cs#L324
> >
> > Certainly it would be nice if there was an API for this but there doesn't
> > seem to be.
> >
> > On Wed, May 28, 2025, 21:24 Ralph Little <skelband at gmail.com> wrote:
> >
> >> Hi Devs,
> >> I'm looking at an issue that affects xsane that I have had a number
> >> of complaints about.
> >>
> >> When xsane starts, it calls sane_get_devices() regardless of whether
> >> or not a specific device was specified by the user. The main issue
> >> is the long delay before xsane starts up while device polling is
> >> happening, which *can* be annoying long.
> >>
> >> On the face of it, this seems pointless if we already know which
> >> device the user wants to use. However, in various places, xsane
> >> uses the descriptive information for that device that we get in the
> >> SANE_Device structure which can only be got from sane_get_devices()
> >> This seems like a gaping hole in the protocol that the information
> >> about a specific device cannot be determined without searching for
> >> all devices.
> >>
> >> Does anyone have any ideas how we might get around this without
> >> breaking backwards compatibility in the SANE API and have
> >> I missed some official method whereby this information might be
> >> obtained?
> >>
> >> Cheers,
> >> Ralph