[Python-Dev] repository locks

[Fred L. Drake, Jr.]

Fredrik Lundh writes:
 > I'm a totally clueless CVS user, and I've managed to check
 > at least 114 things into the repository without leaving any
 > locks behind me.

  The checkins aren't causing the problem -- these are read locks on
the directories.

 > Is it just sheer luck, or is there some flaw in CVS waiting
 > for me to stumble upon?

  I think there's probably a bug in the CVS server, but not sure, and
I haven't dug into the CVS code to figure it out.  It looks like some
operation is locking the directory (update, or a diff), and then the
connection to the client disappears, and the server exits without
cleaning up the lock.  As long as we're reading from the repository,
that doesn't matter -- the problem is when we go to write to the
repository, the CVS server talking to us still thinks there's a
reader, so we're denied access.
  Another possibility is that there's a stuck CVS server (with a dead
connection) holding the lock, and it hasn't exited at all.  But that's
less likely.
  The problem is likely because we're a very active project with lots
of simulataneous repository access on an overloaded service.  I've
noticed that the other services at SourceForge (message forums,
support managers, etc.) have been really slow the past 36 hours as
well.
  Clearly, we van blame it all on Skip.  ;)  <Just kidding!>


  -Fred

-- 
Fred L. Drake, Jr.  <fdrake at beopen.com>
BeOpen PythonLabs Team Member