[Openicc] SANE device color management

[Richard Hughes]

On 17 January 2011 18:43, Greg Troxel <gdt at ir.bbn.com> wrote:
> I think it's best to keep dependencies to an absolute minimum.

Right, agreed. I've added the SANE support with a build time
--enable-sane flag (just like --enable-polkit and --enable-gui) -- but
we could do better -- see below:

> It might make sense to have a small program that links with libsane and
> to have that packaged separately, and have the core colord try to call
> it having dropped privileges.   That would then enable other ways to
> enumerate scanners on platforms where there's some better-than-sane way
> (not sure if there are, but it seems unsound to presume there aren't).

I think a better way would be to have the sane client code dlopened at
startup if present. This would allow distro-packagers to ship a
colord-sane subpackage with just the compiled .so object in and the
daemon to load anything that's present in a specific directory.

I don't think it's worth having a new process just for the sane
devices, but I might re-open that line of thinking if libsane makes
colord crash in real life [1].

Richard.

[1] in old versions of gnome-color-manager we used sane in a thread,
which was notoriously crashy with some SANE drivers. colord doesn't
try threading sane, and seems to make libsane a little happier.