cvs update is uploading entire sandbox to server

[Nelson Bolyard]

I use CVS in client-server mode, running ssh between a client on WinXP 
and a server on Linux.  I have a large workarea (sandbox) containing well 
over 20MB of source code.  It's all been working for me very well for YEARS.
Client is 1.11.22.  Don't know server version.

But suddenly this past week, when I attempt to do a "cvs udpate" command, to
merge any changes from the responsitory into my local sandbox, I find that 
the CVS client wants to upload my entire sandbox, with all file contents, 
up to the server.  I start the update, and I see my system sending out 
SSH traffic as fast as it can until it has sent ~12 Megabytes of source.  
then the actual update begins, which is very quick.  

When I use the -t option with cvs update, I can see it naming the files as 
it sends them to the server.  

When it's all over, it seems to have worked as expected, and done the usual
things, except for the pause for several minutes while it uploads the entire
contents of the sandbox (including subordinate directories).  The CVS udpate
ultimately succeeds.  There's just this LONG delay while copies of the files 
get unnecessarily uploaded.  

It appears to me that this extra upload is a big waste of time and bandwidth
and doesn't actually accomplish anything useful.  I'm retty sure it was not 
behaviing this was a few weeks ago.  Is there some environment variable, or 
contents of a .cvsrc file or a flie in the CVS sub-directories that can cause 
this to happen?  I'm pretty sure that there are no files that have 
inconsistent file name capitalization in the sanbox.  

It only seems to affect cvs update commands.  cvs stat (for example) 
completes with the expected timeliness.  

Any tips on how to get it to stop uploading these sources for an update 
command would be appreciated.  This is very reproducible if you have ideas 
for me to try.

Thanks.

/Nelson B

-- 
Want an e-mail address like mine?
Get a free e-mail account today at www.mail.com!