[sane-devel] Erratic problems with saned/xsane

Henning Meier-Geinitz henning@meier-geinitz.de
Sat, 7 Dec 2002 13:25:37 +0100


On Sat, Dec 07, 2002 at 01:47:17AM -0300, John Coppens wrote:
> Xsane starts up well, even though I'd rather have have it _not_ check
> the parallel port, if I mean to have it run through the daemon. I didn't
> find how to do that.

I have never tried that but you could play with permissions of the
epson.conf file and/or libsane-epson.so.version. Setting them so only
saned can load the backend may work.

Another way would be to link xsane to the net backend directly,
without using libsane-dll. 

> But then, communication between saned and xsane seems to be very
> unreliable. Erratically, I get the famous 'invalid argument' messages,
> when I try for example to change the mode from gray to color, or change
> the resolution. It's just unpredictable. Sometimes it works, sometimes
> not. I restarted xsane many times, and each time it behaves differently.
> Sometimes it works completely, sometimes not al all.

Can you go on after such an error without restarting xsane? Or does
every other action fail after the first error? Does scanning work,
even if errors occured before? Do the errors onlöy occur when
setting/reading options?

> All versions are the last ones - fresly downloaded. Kernel is 2.4.18,
> and the scanner is an Epson ActionScanner II (GT-5000M). All points
> to unreliable comm between xsane and saned...

The communication flow is like this:

xsane <- linking -> libsane-dll <- dynamic loading -> libsane-net <- network
network -> saned <- linking -> libsane-dll <- dynamic loading -> libsane-epson

I guess it doesn't happen when you don't use saned, but access the
epson backend directly?

Do you get the same problems with other frontends (xscanimage,
quiteinsane, scanimage)? No? -> xsane problem (unlikely)

Does it happen with other backends? If you don't have other scan
hardware, try the "test" backend. It does also happen? --> bug in
saned or net. I think that's unlikely, too, but I'm biased. We had
bugs in saned/net before but they resulted in crashes.

You can try if it happens in debug mode, too. Comment out saned in
inetd.conf, restart inetd, and run saned in daemon mode from the
command line: saned -d128
This will also enable debug messages. If the problem occurs in this
mode, too, you can also set epson debugging:

SANE_DEBUG_EPSON=128 saned -d128

Please send us a debug log of a session where such an invalid argument
error occured.

Please also enable debugging on the client side:

SANE_DEBUG_NET=255 xsane

Please also send this log (of the same xsane session).