The existing NFS daemon seems to work minimally, with one small
change. I can't make the change myself, since (1) I don't have a
gnu account, and (2) I don't have the appropriate papers filed with
the FSF. I will fill a bug with the debian bts, but it will be
later this evening.
I say that it works minimally, since it crashes when I try this
command sequence (on my Linux machine):
% mount -t nfs hurd:/ /mnt/hurd
% find /mnt/hurd -name '*' | xargs grep zoobin
I don't expect to find 'zoobin' in the output, but I do expect to
schlep all the disk over from hurd to the Linux machine and search
through it. This seemed like something which would exercise the
nfs daemon, at least to find out if there are resource links. I
did notice that the data seemed to be right for the two or three
files I looked at. I will think about other tests later on.
The short form is that the file hurd/libdiskfs/fhandle.h should define
char filler2[28 - (sizeof (int) + sizeof (int) + sizeof (long))];
not this one:
int filler2[28 - sizeof (int) + sizeof (int) + sizeof (long)];
and then everything has to be recompiled - particularly libdiskfs* and
ext2fs.static. The second misses a pair of parentheses, and declares
filler2 to be an array of ints, and not of chars.