hg: jdk7/tl/jdk: 6843995: Added RowsetFactory and Deprecate COMMIT_ON_ACCEPT_CHANGES, make constants final that needed to be.

Mark Wielaard mark at klomp.org
Sat Aug 22 13:02:18 PDT 2009

Hi Andrew,

On Fri, 2009-08-21 at 20:35 +0100, Andrew John Hughes wrote:
> 2009/8/21 Mark Wielaard <mark at klomp.org>:
> > On Thu, 2009-08-20 at 15:40 -0700, Mark Reinhold wrote:
> >> This change was integrated prematurely.  I've rolled it back in the
> >> jdk7/tl/jdk repository.
> >
> > If at all possible, please don't do this. It plays havoc with already
> > checked out repos and/or automatic clones/backups. You force people to
> > recreate their local repos if you purge a commit that was already public
> > like this. Could you use a normal hg backout instead in the future? That
> > will make sure the integrity of the repo isn't compromised.
> >
> It would be much cleaner to just do a commit reversing the patch (mjw,
> is that what you mean by a backout or something else?)

Yes. "hg backout" creates a reverse changeset that "cancels out" an
earlier commit (but does so in a way that keeps the integrity of
repository, plus the whole history). See also: hg backout --help

> I really don't like the idea of sanitising the repos to remove all
> mistakes and there's also no guarantee everyone will do this and push
> it in every forest.
> > Thanks,
> >
> > Mark

More information about the compiler-dev mailing list