[sane-devel] Epson 1640SU, no devices found, RH 8.0
Mon, 15 Dec 2003 22:00:10 +0000
Content-Type: text/plain; charset=us-ascii
See my previous email, in which I explained the situation. There are
some errors in your text, which I'll correct:
On Mon, Dec 15, 2003 at 08:28:56PM +0100, Oliver Rauch wrote:
> The problem is that in redhat 9.0 there is listed /usr/lib/sane in=20
> the second problem is that there is a link /usr/lib/sane/libsane.so.1 to
> a libsane library in the same directory. (this makes ldconfig changing th=
ldconfig will create links for sonames when they are missing.
> There also is a copy of a libsane.so.1 library in /usr/lib but this=20
> seems not to be used ( I think /usr/lib/sane is searched before /usr/lib =
It is used, of course, as ldd will tell you. However, the search
order that glibc uses recently changed.
> It is a buggy configuration in redhat 9.0 and this also causes
> problems when the rpm is uninstalled and sane-backends is
> installed from source with -prefix=3D/usr because the entry /usr/lib/sane
> is not removed from /etc/ld.so.conf when the sane-backends rpm
> is uninstalled.
Did you test this? I believe it to be incorrect, and I think that the
entry is removed.
In any case, I already explained that future sane-backends packages
will omit the ld.so.conf tampering. In fact, the current development
Fedora package has this removed right now.
At present there is a sane-backends package in the works which undoes
the hacky (IMHO) libsane.so.1-force in ltmain.sh. I understand that
your preferred solution would be to remove the incorrect ld.so.conf
tampering; however starting with new fixed packages at this stage
would only delay things, while the difference between the two fixes is
quite minimal and will be largely unnoticed. At this stage I'd rather
get SANE working again on Red Hat Linux systems.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)
-----END PGP SIGNATURE-----