[Pkg-bazaar-maint] builddeb questions

Reinhard Tartler siretart at tauware.de
Sat Jun 30 19:00:48 UTC 2007


James Westby <jw+debian at jameswestby.net> writes:

> Hi all,
>
> I am working on import code for builddeb at the moment, and I wanted to
> ask for help on a couple of questions.
>
> The first is tagging, some of the code needs to be identify certain
> revisions, so far I have only needed to identify pure upstream versions,
> and so I used a tag of
>
>   upstream-<version>
>
> for those. Does that look right, or should it just be '<version>'?
>
> I can see a need soon to identify released versions of the package, so
> again I would like to design a tag scheme for this. It could just be the
> version number, but there are things like uploads to different
> distributions to consider. It would be possible to prefix the tag with
> the name of the distribution the upload is on its way to (from the
> changelog), but that makes it harder to actually find a version again.
>
> Perhaps I should just assume that uploads to different distributions
> (including things like etch-backports, stable-proposed-updates) go in a
> different branch, so just the version number is used.
>
> Has anyone got any bright ideas?

TBH, I intended to just use a new section in the already existing config
file .bzrbuilddeb.conf for that. I need to mark the loompoints in form
of a ordered list anyway, which cannot be accomplished with only using
tags. Since I have to do that, I think it would feel natural to mark
upstream versions and uploads there as well. This doesn't mean that we
should use bzr tags, it's rather that this way we can let the user
choose the naming schema of the tags.

> The other question is what to do on imports of packages. For a normal
> package you end up with
>
>   packaging  ----*----*----*----*
>                 /         /    /
>   upstream ----*---------*----*
>
> and for native packages you just let linear history.

Basically yes. In practice, I notice that things are a bit more
complicated when you take into account that you might have branches in
your packaging, like we have with the bzr uploads to backports.org, or
uploads to ubuntu. Not sure if we can expect any importer to figure this
out automatically, though. In fact, I don't think so.

> However there may be some packages that go through a period of being
> native and back (one way changes may be quite common). I am unsure of
> what to do in those cases. Does it make sense to do the following

>
>   packaging  ----*----*             *----*
>                 /    / \           /    /
>   upstream ----*----*---*----*----*----*
>                         ^         ^
>                         |         |
>   first native version_/           \_upstream of first non-native.
>
>
> It makes the history a bit funny as the upstream branch contains
> packaging work, but it seems like that is right, as that is a native
> package.

For that use case, it seems natural to me. Right.

-- 
Gruesse/greetings,
Reinhard Tartler, KeyID 945348A4
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 213 bytes
Desc: not available
Url : http://lists.alioth.debian.org/pipermail/pkg-bazaar-maint/attachments/20070630/1da6b57b/attachment.pgp 


More information about the Pkg-bazaar-maint mailing list