Problems with committing a single file - grabs everything

[Hazlett, Susan G.]

> -----Original Message-----
> From: Matt Mackall [mailto:mpm at selenic.com]
> Sent: Thursday, January 06, 2011 3:18 PM
> To: Hazlett, Susan G.
> Cc: mercurial at selenic.com
> Subject: RE: Problems with committing a single file - grabs everything
> 
> On Thu, 2011-01-06 at 14:36 -0800, Hazlett, Susan G. wrote:
> > > -----Original Message-----
> > > From: Matt Mackall [mailto:mpm at selenic.com]
> > > Sent: Thursday, January 06, 2011 2:28 PM
> > > To: Hazlett, Susan G.
> > > Cc: mercurial at selenic.com
> > > Subject: RE: Problems with committing a single file - grabs everything
> > >
> > > On Thu, 2011-01-06 at 13:35 -0800, Hazlett, Susan G. wrote:
> > > > > -----Original Message-----
> > > > > From: Matt Mackall [mailto:mpm at selenic.com]
> > > > > Sent: Thursday, January 06, 2011 1:08 PM
> > > > > To: Hazlett, Susan G.
> > > > > Cc: mercurial at selenic.com
> > > > > Subject: Re: Problems with committing a single file - grabs
> everything
> > > > >
> > > > > On Thu, 2011-01-06 at 11:34 -0800, Hazlett, Susan G. wrote:
> > > > > > When he looked at the commit, it took *all* the files rather
> than
> > > just
> > > > > > config.suggest.  We rolledback, tried it again and this time,
> the
> > > > > > commit took only the config.suggest as he had originally wanted.
> > > > > > Thinking it an anomaly, we moved on, but 5 minutes later he came
> > > back
> > > > > > and was able to reproduce the problem.
> > > > >
> > > > > That's odd. Perhaps there's something interesting hiding in:
> > > > >
> > > > > hg showconfig --debug
> > > >
> > > > [Hazlett, Susan G.]
> > > > I'm including output from both environments.  The first is from
> where
> > > the problem occurred, the second is mine where I could not reproduce.
> > > (I've removed only the username line from both to avoid publishing the
> > > email).
> > >
> > > Hmm, nothing obvious there. What versions are you using? Are you using
> > > subrepos? Does the working directory have two parents (hg parents or
> hg
> > > summary)?
> >  [Hazlett, Susan G.]
> >
> > No subrepos, single parent.
> >
> > % hg summary
> > parent: 1417:6857128369aa tip
> >  Fix error in comment.
> > branch: default
> > commit: 3 modified, 5 unknown
> > update: (current)
> > % hg parents
> > changeset:   1417:6857128369aa
> > tag:         tip
> > user:        <removed email SGH>
> > date:        Thu Jan 06 10:49:36 2011 -0800
> > summary:     Fix error in comment.
> >
> > % hg --version
> > Mercurial Distributed SCM (version 1.5.2)
> 
> Hmm, that's a bit old, but I don't know of any bugs in this area.
> Is this easily reproducible on the machine in question? Do all files
> show up in the "HG:" lines in the commit editor?
> 
> There's not much we can do for here at this point if we can't reproduce
> it; you might be able to dig deeper with --debug, or even --debugger, or
> by instrumenting the code with some print statements. The interesting
> bits are in localrepo.py:commit.

Yes, all of the files show up in the commit editor, and after the post-rollback, we were able to reproduce on that machine.  I'll take a look at those interesting bits and see what comes out of that and let you know.

Susan