baz-import does not record Arch merges

John Arbash Meinel john at arbash-meinel.com
Fri Nov 11 16:46:20 GMT 2005


Aaron Bentley wrote:
> 
> John A Meinel wrote:
> | I was surprised, though, to look at the revision-store files for these
> | revisions, and only see a single parent for all of them. Because I knew
> | I was merging from the other branch.
> 
> There are some branches of baz-import that do support merge metadata.
> It's just that Robert and I haven't quite got everything synced up. You
> can check out
> 
> http://panoramicfeedback.com/opensource/bzr/bzrtools-baz2bzr/
> and
> http://people.ubuntu.com/~robertc/baz2.0/bzrtools-baz2bzr/
> 
> Robert's branch supports some of his unmerged bzrlib changes.

I thought you had that working, which was why I was surprised to see
that it didn't.
Do you have an estimate when these will be brought into your mainline
bzrtools?

> 
> | Also, I understand the desire to use the revision identifier
> | "Arch:archive at somewhere/cat--branch--0.1--patch-10", however, on
> | Windows, the colon (:) is an illegal character in filenames.
> 
> Doh!
> 
> | Which to me seems that we need a different translation for arch revision
> | ids. I realize this is coming a little bit late, since most of launchpad
> | would be converted to using bzr instead of baz, meaning there are a
> | whole lot of revision ids out there with a colon in them.
> 
> Well, we were planning on bumping the namespace again when Robert's
> changes got merged in, so maybe we'll format them differently.

Yeah, I'm not sure what would be good. Colon makes for an awfully nice
namespace separator.
We could go the Mac route, and encode ":" as a "/" (mac does the
opposite, encoding / as a :), effectively pushing the revisions into
their own directories. :)

We could do "Arch-"
And just live with that.
Or do all caps, with a space
'ARCH '

> 
> |
> | The other possibility is to leave the revision identifier the way it is
> | (colon is nice from a logical perspective), and just to have the stores
> | know to look for the file under a different name. We could smash it to
> | underscore, or replace it with something else.
> 
> Martin's been reluctant to encode the IDs, but the only forbidden
> characters so far were '/' and maybe '\'.

I thought newline was forbidden. If it wasn't, we probably should make
it be. I know for file-ids we are being more generous, and allowing
spaces, etc, because Arch would use them. (Though I thought the real
arch id would read anything you wrote, but had all sorts of things
squashed back to _)

John
=:->

> 
> Aaron
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 256 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20051111/f00c7ef1/attachment.pgp 


More information about the bazaar mailing list