On Tue, 2007-03-20 at 12:18 +0100, Julien BLACHE wrote: > David Zeuthen <da...@fubar.dk> wrote: > > Hi, > > > So I'm curious if a) you think this is a good idea; and b) whether such > > a patch would be able to go into mainline SANE? Thanks for considering! > > Looks good to me, as long as the added code in SANE is propery > #ifdef'd out etc. Let's try not to turn SANE into a Linux-specific > piece of code :)
Yeah, it would all be #ifdef'ed out (btw, HAL runs on Solaris and FreeBSD too these days) > Also, how do you manage different HAL versions ? Is the HAL plugin API > stable ? > > What would be the impact in terms of: > - added code I think there would be 200-300 LOC in a new file that and maybe 5-10 lines of code in dll.c. The add-on itself would probably be around 500-100 LOC of code but that would be independent on SANE code apart form linking in libsane. The add-on could use glib but it's not really necessary and I don't see it using it. > - added library dependencies (and here I'm especially worried about > bringing in the infamous GLib for the DBus stuff in dll.c, so if we > could avoid it, that'd be nice) Yeah, it would only pull in libdbus and libdbus have no library requirements except the C library. Specifically glib wouldn't be used (D-Bus does not use glib anywhere mostly because glib doesn't handle OOM). I'll try writing the patch today but I just need a scanner where the libsane driver supports buttons. Any recommendations for a cheap scanner I should get? Btw, how does the sane-backends build system work? I couldn't find the Makefile.am files anywhere... David