[cvsnt] Abort check-in twice on a file causes CVSNT to fill up hard drive

[Davies, Bron]

I just had a problem where a user attempted to commit a particular file twice and clicked on the abort button both times and then proceeded to commit again and it completed successfully.  Overnight, the RCS file for the object he checked in grew to 30+ GB and filled up the hard drive. A CVS process on the server was hung using @40% CPU probably still attempting to write to the file since any attempt I made to free up space on the drive still resulted in 0 free disk space until I killed it. The user's Tortoise version is 1.8.0. on WinXP SP2, CVSNT version is 2.0.34 using NTLM authentication on Win 2000 Server SP4.  I had to destroy the 30GB RCS file so that my user's could continue working so I'm afraid I don't have much else to go on.  I have a suspicion that it is due to the repeated canceling of the commit operation that is at the root of it, however I also suspect a bug in the server since that is where the errant process was running.  I am also posting to the Tortoise CVS list.

Bron Davies
BMC Software