Chown problem with arla

[Jim Nance]

---Magnus Ahltorp <map at stacken.kth.se> wrote:
>
> Sorry for not answering this until now, but we've been trying to track
> the problem.

Thanks!

> The problem is most probably caused by a file system function
> returning a positive errno (negative errno:s are used in the Linux
> kernel). I don't see how that can happen, though. I have seen this
> problem, but only once on one machine since the last snapshot. Do you
> have a test case that will cause this problem?

Sort of.  I see this problem when I try to build a
large set of programs and the source code and libraries
are on AFS and the object files are on NFS.   Part of
the build process does this:

cd $OBJDIR; $(CC) $(OBJS); mv a.out $(PROGNAME)

At first everthing works fine.  Eventually, at some
random place in the build process, the programs
start showing up with out the executable bit set, due
to the the chown problem.  Once that starts happening
it happens every time I make that chown call.  I have
not checked to see if other chown calls work but
judging from the number of executables w/o the x bit
set I suspect that they all fail.  I also have problems
with arlad dying during the builds so I suspect that
there is some uninitialized pointer or something like
that going on.

Has anyone else attempted to build large projects like
mozilla or X11R6 in an AFS file system?  I usually end
up with arla related problems when I try this but I
never see anyone else complaining about it.  It makes
me wonder if something is "special" about my setup.

Thanks,

Jim
_________________________________________________________
DO YOU YAHOO!?
Get your free @yahoo.com address at http://mail.yahoo.com