Cannot merge bundles created without referencing another branch.

Alexander Belchenko bialix at ukr.net
Mon Oct 8 21:17:57 BST 2007


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Aaron Bentley пишет:
> Alexander Belchenko wrote:
>> Yes, IIRC I'm also stepped into this behavior when I wrote the patch
>> for thunderbird at win32 support. And this behavior ('.' branch already had all
>> necesary revisions) breaks the way old bundles working.
> 
> There is a very simple solution to this: don't use "." as a submit
> branch.  It's totally inappropriate for the purpose.  It is not the
> branch you want to submit your changes to.
> 
> Merge directives are for contributing to an ongoing project, and the
> submit branch should be the branch you want to merge your changes into.
>  If there's no such branch, then merge directives aren't appropriate for
> your purpose, and you should zip or tar up the branch directory instead.
> 
> If you absolutely, positively, *must* use a merge directive, then create
> an empty branch and use that as your submit branch.

For me send, merge-directives and bundles are three
different things that solves different purposes.

> The old way -r worked was a constant problem, because people who tried
> to do cherry-picking were instead making it impossible to merge their
> changes.

It's not true in one edge case: when I exactly know that remote branch
at revision r1, but my local branch on revision r2, and I want to
"push" my changes to remote side via bundle. It's not merge process,
it's pseudo-push (bundle + pull), because bzr allows me to pull bundles w/o problems.

>> Yes, I remember
>> your explanation about using this to reduce size of bundle.
> 
> I'm not sure what you mean.

Once I sent you the bundle that was generated against '.' and you was unable to merge it.

>> But in my
>> opinion if we produce bundle against '.' then it's more understandable
>> and straightforward for bzr users if bundle will contain all revisions
>> inside.
> 
> Producing a bundle against "." is ludicrous, and anyone who does it
> deserves what they get.

I'm give up here. You're God of bundles, I'm only little human,
and sometimes I'm willing to use bundles in the way that breaks
commandments.

>> Maybe it's more suitable to use bundle command instead of send,
>> but bundle in the future will die IIUC, and there is way to nowhere.
> 
> The bundle command is essentially the same thing as the send command.
> 
>> BTW, I just realized, that new send behavior will breaks my workflow
>> to obtain new changes via my bzrdev.bialix.com site.
>> I produce bundles from bzr.dev mirror on my site and then download
>> bundles to my computer.
> 
> Why does this break your workflow?

Because I do "pseudo-push" and produce bundle for this purpose against '.'.
As you noted above I need to use reference branch, and reference branch
is the only *right* way to do.
But it's more complex for me, than just producing bundle against '.'.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHCpB1zYr338mxwCURAg8rAKCTymXCHoblR/NebvyBWgVW+QYq4gCeKNdo
lMByNJSZt99wzQokNwm/1fY=
=7oSS
-----END PGP SIGNATURE-----



More information about the bazaar mailing list