CVS rtag corrupting files

[Ross Patterson]

To start off, I wouldn't believe this was happening if I didn't see it with my 
own two eyes.  It has happened before, but I wasn't expecting it then so I 
wasn't prepared to capture the results and report it.  Would you believe ...

During a "cvs rtag" of a large repository (~131K files, ~1.7GB) that touches 
almost all of the files, a small number of files are having four bytes of 
binary zeroes inserted at the start of the file.  Another small number (often 
the same number, but I'm not certain) of files are being truncated by a four 
bytes.  On occasion both problems happen to the same file.

The environment is client-server, with Red Hat Linux 7.3 + cvs 1.11.1p1 as the 
client and RHL 7.2 + cvs 1.11.1p1 as the server.  We're using the pserver 
protocol, and there's no NFS access to the filessystem where the server keeps 
the repository.  The repository was intact before starting the "cvs rtag" 
command and there was no access to it during the command (verified by 
assorted logfiles and "cvs history -a -e").  There are no disk problems on 
the server, and no files outside the repository are showing any damage.

Does anyone have any idea what's going on?  I've been able to fix the damaged 
files, but I'm *really* concerned about how to keep this from happening in 
the future.
-- 
Ross A. Patterson
Chief Technology Officer
CatchFIRE Systems, Inc.
5885 Trinity Parkway, Suite 220
Centreville, VA  20120
(703) 563-4164