[merge][0.91] set default format to dirstate-tags

Robert Collins robertc at robertcollins.net
Tue Aug 14 13:50:05 BST 2007


On Tue, 2007-08-14 at 21:53 +1000, Martin Pool wrote:
> This builds on my previous patch that adjusts append_revision and so on.
> 
>  * set the default branch format and bzrdir format -- the concrete
> change caused by this is that we now use just a last-revision marker,
> and have tags

>  * remote branches now support tags -- of course whether it actually
> works depends on the remote branch format

do you mean RemoteBranch ? or branches over SFTP?

> There is some divergence between formats about whether you are allowed
> to point a branch to a revision that's not in its repository.  A few
> methods assume that it is present and look at the graph.  The tests
> that rely on this being possible seem to do so only accidentally, so I
> have just adjusted them to use a revision that is present.

I'm not entirely clear what you mean here. I guess you are saying that
'tests which are using branches with a revision id *not* present in the
repository seem to be doing so accidentally so they no longer do'. (My
confusion is in the binding of 'this').

> I'd like to merge this as soon as we branch off 0.90 so that we can
> get the most testing with this format in place.

Amen :).

I think the news should be more clear about incompatabilities. we've had
a SURPRISES before. Something like 'Newly inited branchess will not work
with bzr below 0.15. Existing branches will interoperate fine with bzr
0.90, but if upgraded will cease being compatible with bzr below 0.15.'

I think I'd prefer to see append_revision deleted, or the prior contract
upheld, because this is less likely to result in 'it did the wrong
thing' surprises.

You duplicate the IN DEVELOPMENT heading in NEWS in your patch, as well
as changing the version number.

When is branch.supports_tags() called? Would it be better for
RemoteBranch to proxy the method to the smart server, than to
unconditionally report True? (Put another way, what happens if it
reports true and a client tries to use tags, but the underlying branch
doesn't ? - and if this doesn't cause failures (other than tag commands
not tagging ;)), why do we have 'supports_tags' as a method?

bb:tweak
-Rob

-- 
GPG key available at: <http://www.robertcollins.net/keys.txt>.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20070814/873ecd9f/attachment.pgp 


More information about the bazaar mailing list