hg status issue on windows with tip?

[Stephen Darnell]

> On Wed, 2005-08-10 at 12:14 +0100, Stephen Darnell wrote:
> > I've just been catching up with the latest changes, and I'm 
> > getting a problem:
> 
> Sounds like a path normalisation problem.  I'll see if I can 
> dust off a copy of Windows, and take a look.
> 
> <b

Thanks Bryan.

Obviously if you can suggest where to dig I'll do some digging.

Chopping through the changes, the problem seems to start at change 836:

C:\mercurial\latest>hg log -r 836
changeset:   836:1fe3b14c7044
parent:      827:a61728b58dc0
parent:      835:9de3535caae8
user:        mpm at selenic.com
date:        Thu Aug 04 22:27:41 2005
summary:     Merge with TAH

The changes shown by the web interface for this change don't seem
to include the diffs for all files.
http://www.selenic.com/hg/?cmd=changeset;node=1fe3b14c704408d57e0b33efda
af5adba6d87617

Clicking on the filediff links, doesn't show any difference for
util.py, bdiff.c, and hg.py  but presumably these did actually
change at that revision (presumably joining of two heads).
But a diff should be shown - surely?
But I can't immediately see how to do it.

What is worse, is the diff of commands.py shown by hgweb seems to be
just the merging patchups, and does not include the 'real' changes.

Doing the diff the hard way (two tree's sync'd to 835 and 836)
shows what I would consider to be the 'real' diffs.

In there, the pathto() function uses os.sep, but presumably some files
are in canonical / form already?

There's also some normalisation being done in hg.py (walk&traverse)
and some os.path.join's that will presumably be creating names with
backslashes in them.

I'm not sure yet whether either of these are causing the problem though.

Regards,
  Stephen