Can someone help describe the workflow when using bzr-svn...

[John Szakmeister]

On Fri, May 1, 2009 at 8:04 AM, Jelmer Vernooij <jelmer at samba.org> wrote:
> John Szakmeister wrote:
[snip]
>> I realize that... nut it's awkward for svn users to see.  It looks
>> like I changed much more than I actually did.  And makes merging that
>> much more difficult, because you don't know which revision of the
>> commit to apply to a branch.  We ran into this in one case, and
>> decided it was we definitely didn't want to directly push our local
>> branches back into trunk after that.
>>
> Perhaps it needs to be mentioned more prominently in the user guide.

I definitely would.  It also has another side-effect.  We cherry pick
revisions to a stable branch.  Because of the re-ordering, we had to
later cherry pick the reversion and re-application of that commit,
otherwise we were getting conflicts all the time.  Part of this is
probably a Subversion issue (it could use a better merge algorithm).
However, others who use just the SVN client, only have what SVN
offers.  It definitely caught me off guard, but it does match the
merge flow seen in 'bzr qlog'.

[snip]
>> Unfortunately, I can't do that.  But I did take the time to reproduce
>> it in a script.  The short form: you make have an svn:author value
>> set.  When I was playing with this stuff locally, I just gave anon
>> write access to make things go.  Turns out that's what causes the
>> backtrace.  I've attached a script demonstrating this.  If you
>> uncomment the commented lines (and remove the one svn ci line), you'll
>> see everything work as expected.
>>
> Thanks, that's useful. This is now fixed in the 0.6 branch.

Awesome!  Thanks!

-John