On Tue, 10 Jun 2003, Henning Meier-Geinitz wrote: > The backend is called "v4l".
OK, I was not at my Linux box and did not remember its name. > > That is, only the first "scan" > > produces a (more or less) correct image, while the others have the > > succesive scanlines badly shifted. Is it normal? > > If I remeber correctly, it didn't happen with my card the last time I > checked. > > > Can it be avoided? > > I know I can use other software (like XawTV) to take snapshots, but if > > there is such a SANE backend, it should work correctly. > > The code of v4l suffers severe bitrot. The TODO mentions: > > * v4l: > - Geometry support is missing > - Frequency setting is missing Maybe it is the problem? I have a PAL system. But anyway, the first scan of the session is OK. Only the later ones have this problem. > - check if v4lctl distributed with xawtv can be used > - Use #define _LINUX_TIME_H 1 just before including videodev.h or > videodev2.h to avoid > trouble if /usr/include/linux is a link. > - Add v4l updates from Martin Palm, see > http://www.meier-geinitz.de/sane/misc/sane-v4l-update.tar.gz > - compilation seems to break when video4linux version 2 is installed? > Don't think so. I have and use v4l2 on my system (kernel 2.4.21-pre5, but with v4l2 patch), all the applications are linked against it (those which use v4l, of course). v4l2 has a compatibility mode, which makes it appear as v4l1 to the non-v4l2-compliant applications. But as v4l2 is not as mature as v4l1 and its API has slightly changed IIRC, this might perturb some programs. > > Martin Palm worked on the v4l backend but I haven't heard from him for > quite some time. YOu can try the code in the above mentioned archive > but I don't remember if it really works. > > You could bug Martin Palm or Julien Blache (who also wanted/wants to > work on v4l) a bit :-) > > > Also, a v4l2 backend would be welcome too. > > Yes, please go ahead and write it :-) > Would do it, but unfortunately don't speak C enough for this. Levente
