> On 5/6/08, ?tienne Bersac <bersace03 at gmail.com> wrote: > > Hi, > > > > Thanks alan fo this post. I agree with everything on you list. > > Just add > > something : would it be possible to just "ensure" that all usb > > scanner > > backends support libusb:xxx:yyy as "fallback" device name ? > > i think we can do that, but write it in a more general way- backends > must allow the 'name' argument to sane_open to be either the > SANE_Device.name the backend returned in sane_get_devices, or the > device name used by the sanei_ functions, if those names differ. > > still not sure how to document that requirement for the front-end > author, because those device names will vary quite a bit between > platforms...
Now that you talk about platforms, if I may jump into the discussion... I am a developper for Haiku <http://haiku-os.org/>, a Free Software rewrite of BeOS, and I updated the SANE BeOS port for Zeta (BeOS R6) some time ago (I sent diffs which went into cvs). Before porting to Haiku I tried the other day to build SANE cvs under Zeta, but had to fix several issues. I suppose it'd be nice if those could be fixed for a release so at least it builds again on BeOS. IIRC some non-standard (linux specific) types were used (u_int8_t ?) in place of standard ones, without even including the correct headers, so I had to fix many backends. That might not only affect BeOS btw. A better fix would be to use the standardized type instead, with the correct header of course. Plus some other fixes. I don't have a patch around, I'll make one ASAP. Fran?ois.