You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Peter Kahn <ci...@gmail.com> on 2008/10/22 19:11:30 UTC
Can I convince svn 1.5 that a random directory is really a branch off of trunk?
I recently used the Polarion converter to import my MKS repository to svn.
I had to make some changes to support a cutoff point based on date for the
export out of MKS. I confused the importer and now my svn repository has
branches the are really adds of 1000s of files instead of a proper branch.
Is there a way for me to tell SVN that "project/branch/1.0_integration" came
from trunk @ revision 450?
Right now, when I try to merge between branches merge the svn application
takes 100% cpu and gobbles up all of my ram before terminating with an
out-of-memory error.
I can dump my repository and rebuild it with a skeleton of tags branching
where appropriate, but I'll loose the who-did-what history and I'd really
like to keep that.
So, anyone have any ideas? Or can someone point me at the right place for
me to RTFM?
Thanks.
--
Peter Kahn
citizenkahn@gmail.com
pkahnpie1@AIM
http://citizenkahn.myplaxo.com
Awareness - Intention - Action
Re: Can I convince svn 1.5 that a random directory is really a branch off of trunk?
Posted by Peter Kahn <ci...@gmail.com>.
Well, it looks like it works except I cannot use the --reintegratee option
when merging to trunk because the branches are not ancestral. So, its not
perfect, but it seems viable in a small case.
Anyone know if collabnet made a tool to alter the ancestral relationship of
branches?
On Thu, Oct 23, 2008 at 11:34 AM, Peter Kahn <ci...@gmail.com> wrote:
> Ah, so the answer is appears to be "yes": manipulate the mergeinfo for
> trunk and the "branch" such that at branch origin is its recorded as
> synchronized with trunk.
>
> r1000 branch created via add by converter
> r1001-HEAD both branch and trunk modified by various modifications
>
> modify trunk's mergeinfo with /branch/foo:1-1000
> modify branch's mergeinfo with trunk:1-1000
>
> Now branch and trunk should assume that they are insync as of branch
> creation. Now, merges between them should work properly.
>
> I came up with this plan after reading some of this doc:
>
> http://www.collab.net/community/subversion/articles/merge-info.html
>
>
> *Does this make sense?*
>
>
>
> On Wed, Oct 22, 2008 at 3:11 PM, Peter Kahn <ci...@gmail.com> wrote:
>
>>
>> I recently used the Polarion converter to import my MKS repository to
>> svn. I had to make some changes to support a cutoff point based on date for
>> the export out of MKS. I confused the importer and now my svn repository
>> has branches the are really adds of 1000s of files instead of a proper
>> branch.
>>
>> Is there a way for me to tell SVN that "project/branch/1.0_integration"
>> came from trunk @ revision 450?
>>
>>
>> Right now, when I try to merge between branches merge the svn application
>> takes 100% cpu and gobbles up all of my ram before terminating with an
>> out-of-memory error.
>>
>> I can dump my repository and rebuild it with a skeleton of tags branching
>> where appropriate, but I'll loose the who-did-what history and I'd really
>> like to keep that.
>>
>> So, anyone have any ideas? Or can someone point me at the right place for
>> me to RTFM?
>>
>> Thanks.
>>
>> --
>> Peter Kahn
>> citizenkahn@gmail.com
>> pkahnpie1@AIM
>> http://citizenkahn.myplaxo.com
>> Awareness - Intention - Action
>>
>
>
>
> --
> Peter Kahn
> citizenkahn@gmail.com
> pkahnpie1@AIM
> http://citizenkahn.myplaxo.com
> Awareness - Intention - Action
>
--
Peter Kahn
citizenkahn@gmail.com
pkahnpie1@AIM
http://citizenkahn.myplaxo.com
Awareness - Intention - Action
Re: Can I convince svn 1.5 that a random directory is really a branch off of trunk?
Posted by Peter Kahn <ci...@gmail.com>.
Ah, so the answer is appears to be "yes": manipulate the mergeinfo for trunk
and the "branch" such that at branch origin is its recorded as synchronized
with trunk.
r1000 branch created via add by converter
r1001-HEAD both branch and trunk modified by various modifications
modify trunk's mergeinfo with /branch/foo:1-1000
modify branch's mergeinfo with trunk:1-1000
Now branch and trunk should assume that they are insync as of branch
creation. Now, merges between them should work properly.
I came up with this plan after reading some of this doc:
http://www.collab.net/community/subversion/articles/merge-info.html
*Does this make sense?*
On Wed, Oct 22, 2008 at 3:11 PM, Peter Kahn <ci...@gmail.com> wrote:
>
> I recently used the Polarion converter to import my MKS repository to svn.
> I had to make some changes to support a cutoff point based on date for the
> export out of MKS. I confused the importer and now my svn repository has
> branches the are really adds of 1000s of files instead of a proper branch.
>
> Is there a way for me to tell SVN that "project/branch/1.0_integration"
> came from trunk @ revision 450?
>
>
> Right now, when I try to merge between branches merge the svn application
> takes 100% cpu and gobbles up all of my ram before terminating with an
> out-of-memory error.
>
> I can dump my repository and rebuild it with a skeleton of tags branching
> where appropriate, but I'll loose the who-did-what history and I'd really
> like to keep that.
>
> So, anyone have any ideas? Or can someone point me at the right place for
> me to RTFM?
>
> Thanks.
>
> --
> Peter Kahn
> citizenkahn@gmail.com
> pkahnpie1@AIM
> http://citizenkahn.myplaxo.com
> Awareness - Intention - Action
>
--
Peter Kahn
citizenkahn@gmail.com
pkahnpie1@AIM
http://citizenkahn.myplaxo.com
Awareness - Intention - Action
Re: Can I convince svn 1.5 that a random directory is really a branch off of trunk?
Posted by Peter Kahn <ci...@gmail.com>.
Yeah, I tired this in a micro test with a couple of files and it looked
good. In the macro test with my massive bad imports it failed. The problem
is that the converter created two branches or a branch and a tag as a single
checking of 20,000 adds. That's a big commit to parse.
So, i'm going to have to create a skeleton from siginifcant tags instead.
History aggregated by build not by changeset is what we'll wind up with. I
don't love it, but it should be viable.
On Thu, Oct 23, 2008 at 1:32 PM, Konstantin Kolinko
<kn...@gmail.com>wrote:
> 2008/10/22 Peter Kahn <ci...@gmail.com>:
> >
> > I recently used the Polarion converter to import my MKS repository to
> svn.
> > I had to make some changes to support a cutoff point based on date for
> the
> > export out of MKS. I confused the importer and now my svn repository has
> > branches the are really adds of 1000s of files instead of a proper
> branch.
> >
> > Is there a way for me to tell SVN that "project/branch/1.0_integration"
> came
> > from trunk @ revision 450?
> >
> I am sure that it is not possible.
>
> If a "branch" was created by add or import operations instead of svn copy,
> the only way that I know to workaround that is to use the
> '--ignore-ancestry' option.
>
> That is my knowledge since svn 1.4, and svn:mergeinfo should not
> make much difference here.
>
> > Right now, when I try to merge between branches merge the svn application
> > takes 100% cpu and gobbles up all of my ram before terminating with an
> > out-of-memory error.
> >
> It might be a different issue.
>
> (2Peter: resending. Sorry: forgot to CC the list at the first time.)
>
> Best regards,
> Konstantin Kolinko
>
--
Peter Kahn
citizenkahn@gmail.com
pkahnpie1@AIM
http://citizenkahn.myplaxo.com
Awareness - Intention - Action
Re: Can I convince svn 1.5 that a random directory is really a branch off of trunk?
Posted by Konstantin Kolinko <kn...@gmail.com>.
2008/10/22 Peter Kahn <ci...@gmail.com>:
>
> I recently used the Polarion converter to import my MKS repository to svn.
> I had to make some changes to support a cutoff point based on date for the
> export out of MKS. I confused the importer and now my svn repository has
> branches the are really adds of 1000s of files instead of a proper branch.
>
> Is there a way for me to tell SVN that "project/branch/1.0_integration" came
> from trunk @ revision 450?
>
I am sure that it is not possible.
If a "branch" was created by add or import operations instead of svn copy,
the only way that I know to workaround that is to use the
'--ignore-ancestry' option.
That is my knowledge since svn 1.4, and svn:mergeinfo should not
make much difference here.
> Right now, when I try to merge between branches merge the svn application
> takes 100% cpu and gobbles up all of my ram before terminating with an
> out-of-memory error.
>
It might be a different issue.
(2Peter: resending. Sorry: forgot to CC the list at the first time.)
Best regards,
Konstantin Kolinko
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org