1.9rc1 countdown

[Jelmer Vernooij]

On Fri, 2008-10-31 at 13:04 -0500, John Arbash Meinel wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Martin Pool wrote:
> > On Thu, Oct 30, 2008 at 8:41 PM, Martin Pool <mbp at canonical.com> wrote:
> >> On Thu, Oct 30, 2008 at 7:44 PM, Martin Albisetti <argentina at gmail.com> wrote:
> >>> On Thu, Oct 30, 2008 at 8:37 AM, Martin Pool <mbp at canonical.com> wrote:
> >>>> * 1.9 format (uses btree indexes) -- and make this the default?
> >>> I know this has been discussed somewhere, but I can't find it.
> >>> Is this format rich-root?
> >> It still has rich-root as a separate format, but this issue was
> >> mentioned in that thread.  I'd like to just make it rich-root and ask
> >> people to upgrade, but I'd like to understand if there would be any
> >> deleterious effects in doing so.
> > 
> > One effect is that upgrading to rich-root is relatively slower than
> > just updating the indexes or other similar transitions, since we need
> > to update all inventories.
> > 
> > I spoke to Robert about this briefly on irc.
> > 
> >  * We should check that not only are inventories updated correctly
> > when moving in to a rich-root form, but also their sha is updated in
> > the revision object.
> >  * People may have repositories in the wild that (incorrectly, maybe
> > through using very old buggy versions?) mix rich-root and non content,
> > and we should ensure check/reconcile will detect and fix this.  Robert
> > has a patch for check but not for reconcile.
> > 
> > <lifeless> and yes, while the default isn't rich root, I think there
> > are clear drawbacks to adding a new format that is a trapdoor for
> > users on the default
> > 
> > I can't parse that.  If we made the default format rich-root, then it
> > would be the default.
> > 

> So one of the problems with upgrading to rich-root is that when someone
> commits in a rich-root repo, that patch cannot be applied to a
> non-rich-root repository.
> 
> We've run into this in the past when people were experimenting, and
> accidentally upgraded something to rich-root. They had to downgrade, and
> do another commit before we could merge there changes.
> 
> I believe what Robert is saying is that "I think there are drawbacks to
> making a trapdoor format the default."
> 
> For example, once rich-root is default doing:
> 
> bzr init-repo .
> bzr branch http://upstream
> 
> Will convert to the new format, and they will no longer be able to merge
> their changes back into "upstream" without upgrading it.
That's going to have to happen at some point though, if rich roots are
ever going to make it into a default format.

Rich root formats have been supported since bzr 1.0, so it should be
possible to upgrade upstream to some format that can be accessed by
older clients. Users that really need compatibility with older versions
of bzr can use --format=pack-0.92 when creating their shared repository.

Cheers,

Jelmer
-- 
Jelmer Vernooij <jelmer at samba.org> - http://samba.org/~jelmer/
Jabber: jelmer at jabber.fsfe.org
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 315 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20081031/b8ccf357/attachment.pgp