Peter, et al,

My joy is short-lived.  Two attempts at building a linux kernel on a coda
filesystem have failed with spurious truncated *.o files.  Different files
both times.

For another data point, a "make boot" insists upon rebuilding absolutely
everything from the beginning on a subsequent restart.  Granted that the
kernel build process is a bit kludgy (though getting better), but it is
substantially smarter than this on an NFS or local filesystem.

To my eyes, the datestamps look valid and I cannot understand why make
thinks it needs to rebuild everything.  Is there a difference in how
stat() reports file information on a coda filesystem that could account
for this?

What would you like me to try in pursuit of the file corruption issues?

Steve