On 11/4/2005 3:28 PM, Gale, David wrote:
> Duncan Murdoch wrote:
>> On 11/4/2005 2:07 PM, Berlin Brown wrote:
>>> Am I wrong in saying that, it seems easier to delete a branch and
>>> then use svn copy to recreate that branch than it is to constantly
>>> do a merge and commit to keep up to date with for example a trunk?
>>
>> Maybe I'm misunderstanding you, but this sounds as though it won't
>> work. If you have committed changes on both the branch and the
>> trunk and you haven't merged them at the time you do the delete, they
>> won't be merged when you recreate the branch, whether you copy from
>> the trunk or the deleted branch.
>>>
>>> Am I wrong here?
>>
>> I think so...
>>
>> Duncan Murdoch
>>>
>>> Is there a lot wasted by doing constant copying.
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
>>> For additional commands, e-mail: users-help@subversion.tigris.org
>>>
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
>> For additional commands, e-mail: users-help@subversion.tigris.org
>
> I think Berlin's question is more of dealing with the lack of
> merge-tracking in Subversion. So, if you have a personal branch for
> your dev work, and periodically merge your changes back into the trunk,
> according to the book you need to track the revision at which the last
> merge happened (by putting a comment into the log, for instance), and
> use that as the base of your next merge command. This is, of course,
> harder than finding the revision at which a branch was created, since
> svn log has a nice "stop on copy" parameter. To avoid this pain,
> Berlin's thinking of deleting the branch after merging, and then
> re-creating it, so every merge is effectively "from creation of the
> branch until now".
>
> This is, of course, the same thing as just creating a new branch with a
> new name, but recycling the old name. Given the nature of svn branches,
> there shouldn't be any adverse effect to doing it this way, other than
> having difficulty retrieving old versions of the dev branch (since you'd
> need to use peg revisions). If that's not something you plan on doing
> often, I can't think of any reasons not to do this...any one else?
>
> -David
>
Thanks (and thanks to Joshua) for the explanations.
The second part of Berlin's question, "Is there a lot wasted by doing
constant copying", would seem to depend on how Subversion stores the
deltas internally. I've read about that before (how FSFS is different
than using BDB), but I don't remember the details well enough, and can't
find the reference.
Can anyone point it out to me? I.e. a document describing what is
stored in the deltas.
Duncan Murdoch
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Fri Nov 4 21:58:50 2005