I'm trying out pre37 on my Linux 2.2.3 system and I can't get it to
compile. In include/atalk/adouble.h, it mentions the syscall
__NR_sendfile, which isn't in my asm/unistd.h. Here's gcc's complaint:
gcc -p -DSENDFILE_FLAVOR_LINUX -DHAVE_IFNAMEINDEX -O2
-fomit-frame-pointer -fsigned-char -Wunused -Wuninitialized
-I../../include -c ad_open.c
../../include/atalk/adouble.h: In function `sendfile':
In file included from ad_open.c:38:
../../include/atalk/adouble.h:50: `__NR_sendfile' undeclared (first use
this function)
../../include/atalk/adouble.h:50: (Each undeclared identifier is reported
only once
../../include/atalk/adouble.h:50: for each function it appears in.)
On Thu, 30 Dec 1999, a sun wrote:
>
> hi all,
>
> here's something on the eve of the year before some arbitrarily
> specified millenium. it has a few more things ticked off from the
> list:
>
> 1) password changing via dhx. the client-side is a little
> buggy, so i don't check a few things that i should.
>
> 2) a workaround for a problem with large quantum sizes in
> os 9. you can now report whatever quantum size you wish
> to the client. netatalk doesn't care, but the appleshare
> client doesn't seem to like large values. read
> config/afpd.conf.
>
> 3) sundry os x bits. i still haven't fixed
> everything. however, netatalk is now using the same
> values as apple's appletalk router.
>
> 4) more interface detection code. for those systems that
> support if_nameindex, you can now specify
> -DHAVE_IFNAMEINDEX. that should include glibc 2.1 +
> linux 2.2, solaris 8 (nee 2.8), and i think some of the
> newer bsd's.
>
> afpd will now use the ip address of the first valid
> interface that it can find if your dns is screwed up and
> you don't have the hostname in /etc/hosts.
>
> QUESTION: does SIOCGIFCONF just not work with ddp on
> solaris, or does it just not work in general? if it
> doesn't work with ddp, does it report an error when
> used, or does it report something random?
>
> 5) afpd will no longer report an error if it can't set the
> desktop directory's ownership/permissions.
>
> the site:
> ftp.cobaltnet.com:/pub/users/asun/testing/pre-asun2.1.4-37.tar.gz
>
> this is almost a release candidate. i'll take another look at what's
> needed for os x, but that's about all i'll do. unless a kerberos uam
> spec with details of what gets passed back and forth with the
> appropriate bit sizes and what not miraculously appears in my inbox,
> the kerberos uam won't work. you'll just have to use pam. for those
> interested in writing their own uams, let me know if the interface
> makes sense to you. read include/atalk/uam.h for the exported
> interface and the modules in etc/uams for actual
> implementations.
>
> also, i would like reports for the various bsd's on the necessary
> flags to get the runtime library linking to work correctly. here's
> what i need: os rev, architecture, and flags. i think the elf-based
> systems are all the same, but i'm not sure.
>
>
>
>
> -- and now for more on the future --
>
> please let me know what's desired for the asun2.3.x series. here's the
> list of things that i would like to see done:
> 1) db support. actually, quite a bit of this is already done. i
> just have to go through and finish it and integrate it with
> afpd.
>
> 2) incorporation of the papd/pap patches. note: for printer
> authentication, i would like to see use of the same uams as
> afpd uses. if you look in etc/uams, you'll see that i already
> have stubs for printer authentication support.
>
> 3) permission/ownership overhaul. i already have some patches for
> client-based administration, so i'll probably base some of
> this off that. basically, this will add the ability to specify
> an administrative user/group.
>
> 4) afp client support so that you can have an ftp-style
> interface to an afp server.
>
> 5) SLP support so that we can do nbp-style registration via
> tcp/ip. if someone's willing to work on an opensource SLP
> library (there's already an rfc with c language bindings), i'd
> appreciate it.
>
> 6) autoconf support. i have some patches for this already. this
> should make building afp/tcp much easier on random unix-style
> oses as well as avoiding headaches with runtime dynamic
> linkage that the bsd's seem to foster.
>
> -a
> [EMAIL PROTECTED]
>
----------------------------------------------
Andrew McNabb
Argus Systems Group
[EMAIL PROTECTED]
----------------------------------------------
