[sane-devel] MacOS X SCSI support and other changes in SANE CVS

[Henning Meier-Geinitz]

Hi,

On Fri, May 02, 2003 at 09:35:56AM +0900, Peter O'Gorman wrote:
> Okay, that and changing the link order (darwin ld is *very* wierd) 
> allows scanimage to build.

Sorry, that doesn't work (at least) on Linux.

-       @$(LIBTOOL) $(MLINK) $(LINK) saned.o $(LIBSANEI) $(LIBSANE) $(LIBLIB) \
+       @$(LIBTOOL) $(MLINK) $(LINK) saned.o $(LIBLIB) $(LIBSANEI) $(LIBSANE) \
                $(LIBS)

LIBLIB must be placed after LIBSANEI because LIBLIB defines symbols
that are used in LIBSANEI (e.g. md5_*).

About which symbol is the linker complaining if you use the old link
order?

I'm surprised about this problem, too. I think some people built
sane-backends without such changes but I may be wrong. That
Makefile.in hasn't been changed for quite some time. Are the Darwin
versions that different?

If there is no other solution, we could use different link orders if
we are compiling on Darwin. But I fear that this breaks other Darwin
systems.

> saned does not build however because darwin does not have poll(2).

Oh well, I thought Darwin some sort of Unix? :-)

> Even though configure checks this, saned goes ahead and uses it
> anyway :(

configure checks for poll.h because OS/2 doesn't have poll.h.

I've added a check to configure now. So IPV6 support is disabled if
poll isn't found. It would be nicer to have a poll substitude (using
select?). Anyone?

> Attached is a patch which will get you as far as scanimage building.

Thanks again. I've applied the lalloca part and the configure changes.

A new unofficial snapshot is here:

http://www.meier-geinitz.de/sane/snapshots/

Bye,
  Henning