Re: [PATCH] Merge source and target must be related but different branches, v1
From: Julian Foad <julianfoad_at_btopenworld.com>
Date: Thu, 15 Dec 2011 12:53:23 +0000 (GMT)
Paul Burba wrote:
> Julian Foad wrote:
This "sync merge from my own history" operation seems bogus. I notice that (without my patch) it merges changes from the future *and* records mergeinfo for them. Surely we didn't ever intend that?
$ svn up -r1214700
$ svn merge ^/subversion/trunk
$ svn diff
We don't ever intend to record self-referential mergeinfo, do we? I'm assuming that's bogus and the merge should really be rejected or at least the mergeinfo should be elided.
>> A few tests currently fail with this patch -- tests that use the
I'm struggling to see what we expected this kind of syntax to do and why. It seems to have been committed mainly in r864853 and I'm seeing a long email thread <http://svn.haxx.se/dev/archive-2007-04/0957.shtml> about it. I'm not happy about simply removing those test cases until I understand a little better what's wanted.
Also merge_reintegrate_tests.py 5 and 9 fail. That's because the specified source or target location is also the branching point, so one of the locations is equal to the common ancestor. My patch currently rejects that, but I'll simply remove that part of the check for now and then consider later whether it can be tightened a bit. In other words I'm happy about dealing with this part.
> A bit bikesheddy, bit I wonder if it might make for a cleaner error
Sure, that would be nice throughout 'svn', but I won't aim to do that within this patch.
This is an archived mail posted to the Subversion Dev mailing list.