Try this with darcs-2:

rm afile
darcs record
darcs mv anotherfile afile

It will say that it can't proceed since afile already exists (at  
least on Windows), even though afile actually doesn't exist.

Zooko,

I just added a new test to "mv.pl" and sent you the patch. See if it passes for
you with the darcs in question. If not, it's a Windows-specific issue. If so,
see if you can figure out the difference between your case and the test case. 

   Mark

For some reason the subsequent discussion didn't get added to this ticket.  The
bug is Windows-specific.  I haven't tried to reproduce it recently, because I
don't like doing a job better suited for a robot (i.e. a buildbot).

So, here's what I've decided:  I'm not going to do the job of a buildbot:
running the normal darcs unit tests on different platforms.  However, I will set
up and maintain a buildslave which will do that for me.  I can get my company to
let us use a Windows, a Cygwin, and a few different Linux systems, and I'll use
my own Solaris system.  Then darcs can have a set of automated buildslaves
similar to this set: http://allmydata.org/buildbot/waterfall?show_events=false

I will also help set up and configure the buildmaster.  The buildmaster issues
instructions to buildslaves, saying, basically "Do a darcs checkout of the
latest version, do a make, then do a make test, and report back the results.".

The buildmaster can be triggered to issue these commands any time a new patch is
pushed into the darcs repository, by editing the _darcs/prefs/defaults and
entering some sort of trigger script in there.

Here's how to install buildbot:

http://buildbot.net/repos/release/README

Zooko,

Buildbot is already installed on darcs.net (by me), but I was waiting to get a
little further before announcing it. However, I totally stalled on it, so if you
could pick it up, that would be great. (Assuming server access issues are worked
with David). I'll forward you my status e-mails directly.

Zooko: looking at http://buildbot.darcs.net, I see that Mark's build test is
passing (although, we can't tell for Windows, due to issue840)

Could you see what is different about your case and the one in Mark's test?

I cannot reproduce it on last released darcs on Windows (2.2.1).

echo > afile
echo > anotherfile
darcs init
darcs record -lam "test"
del afile
darcs record -lam "test"
darcs mv anotherfile afile
darcs record -lam "test"


I'll mark it dead; please reopen it with more information if you still think it 
still exists.