[Svnmerge] Merging blocked revisions between branches

[Blair Zajac]

David James wrote:
> On 3/22/06, Blair Zajac <blair at orcaware.com> wrote:
> 
>>David James wrote:
>>
>>>Currently, svnmerge.py is not consistent about whether blocked
>>>revisions are merged between branches. Currently, blocked revisions
>>>are sometimes merged between branches, but we silently ignore the
>>>merge if any conflicts arise. I'd like to setup more consistent
>>>behavior.
>>>
>>>Here's two options, both of which would be consistent:
>>>1) Never merge 'blocked revisions' between branches. If you ask
>>>svnmerge.py to not merge r10 from branchA to branchB, and then merge
>>>changes from branchB to branchC, you've only asked svnmerge to block a
>>>merge of r10 from branchA to branchB. r10 is still available to merge
>>>from branchA to branchC.
>>>2) Always merge 'blocked revisions'  between branches. If you ask
>>>svnmerge.py to not merge r10 from branchA to branchB, and then merge
>>>changes from branchB to branchC, you've therefore asked svnmerge.py to
>>>not merge r10 from branchA to branchC.
>>>
>>>Which choice do you think would be better? I tend to favour option (2).
>>
>>Can you explain where this would occur?
> 
> 
> Here's the exact scenario:
> 1. Block r10 from merging from branch A -> branch B
> 2. Merge changes from branch B -> branch C.
> 3. Merge changes from branch A -> branch C.
> 
> Should (3) attempt to merge r10 from branch A -> branch C? This
> depends on whether the block of 'r10' is merged when you merge from
> branch B -> branch C.

I agree with your initial assessment to not merge r10 over for the reason that 
it has the principal of least surprise that svnmerge.py does not remove any 
changes in one changeset when merging them over, even its own properties.

Given the ancestory of merges, then branchC should not recieve r10.

Regards,
Blair

-- 
Blair Zajac, Ph.D.
CTO, OrcaWare Technologies
<blair at orcaware.com>
Subversion training, consulting and support
http://www.orcaware.com/svn/