[sane-devel] net/saned problems with EPSON backend

[Major A]

> > Is it possible that there is a problem in net/saned that only comes to
> > light when there is a physical network involved, as opposed to a
> > loopback interface?
> 
> Yes, that's possible. We had such bugs in saned some years ago. I
> think it was the other way round (only localhost buggy), but that

I think I jumped the gun here. I can't reproduce the exact behaviour
that I had before, but there still are differences (they might be
random though, not dependant on what network interface is used).

I'm still using 1.0.11 (more about that later).

I have a suspicion, though: as it is at the moment, the options that
cause trouble are the ones which allow from a selection of strings AND
which use strings with whitespace in them. Currently (that might
change tomorrow if there really is something random to it) the only
option that can be set without an "invalid argument" is the "Focus
2.5mm above glass", but only if xsane runs on the REMOTE
machine. Anything else (including the other choice in this option,
"Focus on glass"), results in an "invalid argument".

It doesn't make a difference as to whether I run saned from inetd or
in debug mode. In debug mode, it doesn't print anything out of the
ordinary.

> Try to reproduce with different frontends (e.g. scanimage). Does it
> really only happen over the net and not with net:localhost:epson? Does
> it always happen? After you get the error, can you go on or does saned
> crash?

I also tried scanimage, with very interesting results. Run as root
(always making sure the right device was set with -d), scanimage works
just fine. As a regular user, however, it doesn't, it gives "invalid
arguments". Here, behaviour seems to be the same across both machines.

saned never crashed after the error.

> You could also try sane-backends 1.0.12-pre1. There have been some
> changes to saned, but I don't think they are related to your problem.

I just tried 1.0.12-pre1 on my own scanner at home, no problems at
all. I modified the coolscan backend to include string names with
whitespace, still no error.

Now I've also tested 1.0.12-pre1 on the machines in question: xsane
still does the same weird things (but only through the net/saned, not
directly), but now both machines are identical in behaviour. Also,
scanimage seems to work just fine now. I haven't rebuilt xsane, I
don't think it would make a difference.

What do you think? What can I try next? Wherever the problem is, it's
somewhere between saned and the net and epson backends, with some
involvement of the frontend...

  Andras

===========================================================================
Major Andras
    e-mail: andras at users.sourceforge.net
    www:    http://andras.webhop.org/
===========================================================================