So something about SVN and its atomic commits might be futzing with how it does merges — I’m not sure — but it didn’t work for us in this case. I suppose we should spend a bit more time retracing our steps — I will try to put together a follow-up post.

Yeah, that’d be good to know what causes this state.

I’m curious to see if there’s another solution to this problem, since externals seem to have some pretty onerous requirements.

If you can reproduce, you might also take a look at using svnmerge.py to do your merges; the tool tracks some extra metadata properties, and may help address issues like these (it tracks merges in time, so you don’t have to reperform them; I think this may be the problem you’re running into).

Let me know if can help out.

– Your friend build engineer

Say in the instance where I have to push out a config adjustment or a serious bug in a single file, I don’t want to have that single change depending on a whole array of other trunk updates sitting around in the queue — which usually aren’t completely baked yet.

We ending up doing an svn merge _for that file_ between the trunk revision and the current prod revision. That got our fix in, we updated the site, etc. Later, when trying to merge the rest of the trunk changes onto the prod tag, we ran into issues where 10-12 files were in conflict when a) they had not branched at all b) the revision args in the merge command were correct.

So something about SVN and its atomic commits might be futzing with how it does merges — I’m not sure — but it didn’t work for us in this case. I suppose we should spend a bit more time retracing our steps — I will try to put together a follow-up post.

*waits for the “use ____ instead” comments*

Here it comes… what exactly do you mean by “SVN eats itself?” You say “So when we have tried to merge the entire tree after merging single files manually we’ve had some interesting results — conflicts in certain places and an overall headache.” Are these conflicts generated by things you’ve already merged, or things you perhaps don’t care about (but are required to merge anyway?)

The interesting thing is that if I read it correctly, externals are merely labels, but at the directory level. It’s not entirely clear to me that that’s actually what you wanted. It sounds like you want sane merging algorithms when adding to your testing/production tagging directories/areas.

Am I reading that right, or am I misreading?

From our experience I can only say: Try to merge often. It certainly reduces the number of conflicts you run into. Obviously, new development most often causes less trouble than bug fixes. With time you will get a better feeling how much and what you want to have in one merge.