probably bug? merge A B and merge B A produce different set of conflicts

Alexander Belchenko bialix at ukr.net
Fri Nov 30 12:25:32 GMT 2007


Aaron Bentley пишет:
>>> My best guess would be poor choice of a least common ancestor, but I
>>> really need the branches in question to know.
>> C:\work\Bazaar\devel\case.insensitive.wt>bzr bundle --no-patch >
>> case.insensitive.wt.bundle
>>
>> Attached. If this bundle lacks some info then I'll publish it on my server.
> 
> Thanks.  I think that's enough.
> 
> You say you get different conflicts, but I can't reproduce this-- I get
> conflicts in the same files, either way:

> This is with bzr.dev revno 3042 and case.insensitive.wt revno 2992

Hmm, I can't reproduce it too with current bzr.dev. And I'm not sure 
about past report, I think I saw what I saw, but it was last night and 
my eyes could be a bit tired.

> I agree it doesn't seem to make sense that you would get conflicts in
> files you haven't changed.
> 
> What happened is you did a criss-cross merge at revno 2991.  You merged
> your mv.change.case branch into your case.insensitive.wt branch.  This
> meant that there was no least common ancestor (LCA), so we had to look
> much farther back in history to find a common ancestor.
> 
> If you use annotate merge, (bzr merge --weave) you should only get
> conflicts in NEWS.

Yes, it is. Thanks for explanation.
Most of the time weave merge does not works any bit better than usual 
3-way merge, but in the case of criss-cross merge it actually do
the right thing.



More information about the bazaar mailing list