Re: Glibc sunrpc failure?

[Jeff Bailey <jbailey at nisa dot net>]

On Sun, Jan 26, 2003 at 02:41:55PM -0800, Roland McGrath wrote:

> I think the plan at that time was not to change nothing.  I had a large
> matrix of configurations and type settings in front of me, and I must have
> interpreted some wrong.  We should get typesizes.h and the mig sizes in synch,
> but we'll have to figure out which should change.

hurd_types.defs defines the mig type as this:

type io_statbuf_t = struct[32] of int;

hurd_types.h has this:

typedef struct stat64 io_statbuf_t;

So the problem isn't so much that the mig type doesn't match: It's just
a stat64 struct packed into the stack of ints.  (We probably do have to
tweak the padding bytes at the end, though.  I can provide a patch for
that when I'm certain which way we're going with this).  The problem is
more that the filesystem servers are packing a different stat64 struct
than glibc is unpacking on a running Hurd system.  Recompiling the Hurd
servers with this new glibc would also solve the problem.  I'm fine with
that solution too.

Tks,
Jeff Bailey