Re: possible bug with _WriteSnapshot

[Urs Hoelzle <urs@otis.stanford.edu>]

> 	I grabbed the Self binaries and scripts a couple of days ago.  I'm
> running it on a Sun SPARC IPC under SunOS 4.1.1.  The very first
> problem I had is with getting _WriteSnapshot to write a complete
> snapshot (the system seems to work fairly well other than this).  When
> I try to write the snapshot, it goes away for about 5 seconds, and
> then comes back with the fatal error message below.  This sounds more

> > 'snap' _WriteSnapshot
> cannot write to file: Interrupted system call
> fatal error: write error

This is the infamous "SunOS 4.1.1 write bug"; basically, the write()
system call violates its contract by returning with EINTR before it
has written everything.  This bug occurs in SunOS 4.1.1 only (not in
4.1).  It doesn't always appear; writes over NFS normally seem to
work, and writes to /tmp also work if you're running tmpfs.  We got
the bug only when writing to a local disk if I remember correctly.

A kernel patch to fix the write problem is available from Sun (bug ID
1052649).  I forgot their 800 number, but your sysadmin should know.

Hope that helps,

-Urs