[Svnmerge] [PATCH] Eliminate spurious svnmerge-integratedproperty conflicts
Raman Gupta
rocketraman at fastmail.fm
Thu Feb 15 11:22:07 PST 2007
Blair Zajac wrote:
> Patrick Oor wrote:
>> I hope the patch will be accepted soon. Is not there any expectation
>> of when this can happen? It is certainly a useful patch for us.
>
> IIRC, this patch, while simple, has some issues in that information is
> lost, and so won't be applied to svnmerge.py.
For MOST use cases (for example, Auke's case and any other use case
that reduces to unidirectional or bidirectional merging between
branches), this patch does *not* lose any relevant information.
For the remaining use cases (what we can call "graph" merging), though
the integrated property values currently get merged back and forth
between branches, making it *seem* like svnmerge supports graph-based
merging, svnmerge has never claimed to support graph merging and does
not in actuality do it correctly anyway. In order to support this, we
would need to manually merge the integrated property as I described in
earlier posts. [1]
One other point: the entries in the integrated property that are
removed by the patch will only be related to branches that have never
been initialized with the target branch. My opinion is that we should
interpret this as meaning the user is not interested in, nor do they
want to see, that information anyway. The removal of this extraneous
information is a feature, not a bug! This behavior should be valid
even when using graph-based merging.
Daniel, when you originally brought up your concerns, you specifically
mentioned Auke's use case. We have shown that Auke's use case is
better off with the patch than without it. Do you have any other
concerns with it?
Cheers,
Raman
[1] IMHO, graph merging is rare compared to uni and bi-directional
merging (or transitive merging that reduces to uni or bi-directional
merging as in Auke's case). Considering svn 1.5 with its superior
merge support is imminent, I believe it is not necessary to implement
graph merging in svnmerge.py -- however we can have this discussion
independent of whether this patch should be applied or not.
More information about the Svnmerge
mailing list