Boost logo

Boost :

Subject: Re: [boost] SVN workflow (Was: Maintenance suspended)
From: John Maddock (boost.regex_at_[hidden])
Date: 2010-07-09 07:01:45


>> I tried, I really did, to resolve the trunk and release branches of my
>> libraries for the upcoming
>> release. But the truth is that I'm finding Boost's current
>> interdependent structure and the
>> sluggishness of SVN on a project of our size just makes everything too
>> hard to do.
>
> Maybe, you could describe your SVN workflow? While SVN certainly could be
> a bit faster, I find
> that since server upgrade to 1.6 and first merge with 1.6, subsequent
> merges take less time that I
> need to deal with a cup of tea. And it's once per release cycle.

Hmmm, I'm still finding merges very slow for large trees - Boost.Math takes
a good hour or more to merge for example - I would expect Boost.Python to be
the same.

That said, I only merge when I've accumulated enough changes to make a
"release" for that library worthwhile, and it's not like I have to hold it's
hand or anything while the merge is going on. So it's basically 5 minutes
and then get on with something else, then when the merge is complete I
visually inspect the diff and run the tests on the release branch - again
the latter can take quite a while, but again it's just a case of fire off
the command and then leave it to it. So the amount of my time spent on
merging is very minimal, provided of course you're not in a hurry and don't
desperately need the machine cycles for something else. This also assuming
you're doing a "bring the release branch into synch with Trunk" kind of
merge - I haven't tried merging specific patches or revisions - I assume
that would take rather more figuring out.

John.


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk