bzr-svn / qbzr issue with qlog

Stephen J. Turnbull stephen at xemacs.org
Thu Jun 11 17:09:26 BST 2009


Russel Winder writes:

 > > How can you tell whether branch A came from B or B came from A? I
 > > realize you probably ran "bzr branch A B" at some point. At the moment,
 > > we don't record that event, so all we know is that you have 2 branches,
 > > that share some history, and have some new data in the tip.
 > 
 > Given that there was a branch of B from A shouldn't that metadata be
 > encoded?  Certainly it is needed if there is to be any hope of
 > traceability.  From an human knowledge and intention perspective A is
 > definitely the parent of B and will likely be used from that
 > perspective. 

That's not necessarily true.  In personal projects I regularly do
things like

git branch going-nowhere-fast master # copy branch at current head
git branch -f master making-progress # overwrite master with a branch that
                                     # may go somewhere in my lifetime

when I get stuck.  (Note that in git this is all done by reference
manipulation, so it's really cheap.)  I've also been in the situation
(using CVS) where the trunk got filled with a lot of release-oriented
bugfix cruft that turned out to be just about to collapse of its own
weight, and it would have been useful to simply "graft on" a new trunk
about 30 revisions back.  So for me there isn't a true trunk, like
some kind of 100m-tall redwood.  Rather it's more like a mangrove, and
all I can say is that some branches are more actively growing than
others are ... but that can and does change rather suddenly.

I don't claim any generality, and obviously this is unrelated to your
workflow.  The point is that I get the feeling that this is all very
workflow-specific.  I'd be uncomfortable with a patch that gave any
branch or workflow special treatment here.



More information about the bazaar mailing list