[sane-devel] Umax Astra 2200 under OpenBSD - working

Henning Meier-Geinitz henning at meier-geinitz.de
Mon Aug 22 17:12:04 UTC 2005


On Mon, Aug 22, 2005 at 09:22:42AM -0700, alancorey at yahoo.com wrote:
> I persisted, on a rainy day, and got my Astra 2200
> working with sane-backends-1.0.16 and libusb-0.1.10a.

Ah, very fine.

> The first thing I found, on the libusb site in the
> mailing list area, was a patch by Robert Heller for
> FreeBSD.


I propose to send this to the libusb mailing list (or if they have a
bug tracker, to taht one).
> I had to force sane-backends-1.0.16 to use libusb by
> doing 
> configure --enable-libusb 
> because it still claimed to not be able to find usb.h
> My usb.h is where it's supposed to be:
> cpia# ls -la /usr/local/include/usb.h
> -rw-r--r--  1 root  wheel  8321 Aug 20 11:34
> /usr/local/include/usb.h
> OpenBSD also has a /usr/include/dev/usb/usb.h, I don't
> know if that might be causing the problem by maybe
> being found first.

I don't think so. Maybe on OpenBSD the compiler/preprocessor just
doesn't look for headers in /usr/local/include/ by default? The
autoconfig scripts look in the default paths only. Something like this
may help:

CPPFLAGS="-I/usr/local/include/" ./configure

> When I looked I found that the versions were 1.16
> instead of 1, so I copied the umax one to
> libsane-umax.so.1 and it worked.  I think I've seen
> some installs that symlink the current version to an
> earlier standard one like .1 but nothing like that
> happened here.

This seems to be a real problem on OpenBSD. The links are created by
libtool automatically on all other platforms i can test (including
FreeBSD and NetBSD). So maybe there is a problem with libtool on

SANE used to create these links itsself (still commented in
backend/Makefile.in). But this created other trouble so it was removed
some years ago.

Maybe the dll backend should be more flexible in which extensions it
looks for.

> Lamp control under xscanimage doesn't seem to work,
> although I've got it enabled in umax.conf and the
> description from scanimage -L says the capability
> exists.  The Umax Windows driver can do it, but only
> if you keep the silly thing running in your tray.
> A really useful feature in xsane would be an option to
> turn off the scanner lamp when the program closes, if
> the current device supports it.

This is backend-dependent. Some of "my" backends just turn off the lamp
at "sane_close()", others provide an option for the user to decide. 

> And another that may be a typo: when looking at a
> newly-scanned grayscale image, in the size statement
> at the lower left corner, it tells me the image is 1
> *bit* per color instead of 1 byte.  Scared me.  When I
> do a color scan it still says 1 bit per color.

It also says 1 buit per color in 1 and in 16 bit mode so this one
seems to be fixed (?). Also for Oliver.

> Enabling TIFF and JPEG in sane-backends doesn't seem
> to work.  I did
> ./configure --enable-tiff --enable-jpeg
> but looking at the config.log it seems like it
> couldn't find the libraries.  They are there, and show
> up if I do ldcofig -r.  I'm not sure what versions
> it's looking for.

tiff and jpeg are enabled by default.

Maybe the header file is not found if it's in /usr/local? CPPFLAGS may
help, also CFLAGS="-L..." if the lib is in an unusual directory. As
above, the configure script just looks in the default directories.

> XSane's default pnm format doesn't
> seem to be very universally supported,

pnm is the standard interchange format if nothing else works. If a
viewer or image manipulation program can't handle this it's broken in
my opinion.

> and totally uncompressed. 

This is intentional.

> Gimp knows it, but qiv, Mozilla, Photoshop don't.

qiv can display pnm (just tested). Mozilla opens display (ImageMagick)
here to show it. I can't test photoshop.

> IrfanView does, so I suppose I can batch convert scans in that.

For batch conversion, imagemagick is the right tool.

But xsane can save as png, jpeg or tiff :-)


More information about the sane-devel mailing list