At Sat, 22 Jan 2011 08:01:04 -0500, John Wiegley wrote:
On Jan 14, 2011, at 2:43 PM, John Wiegley wrote:
I wanted to chime in here and say hello; and that yes, I'm working on the modularization project with input from both Dave and Eric. I have the beginnings of the migration script up here:
https://github.com/jwiegley/boost-migrate
It's very rough right now, as I'm still exploring the completeness of the Subversion 'dump' format, and how to use Git plumbing to avoid the migration process taking days upon days to complete.
Quick status update: Direct conversion from a Subversion flat-filesystem to an identical Git flat-filesystem now works.
So now you have a linear sequence of commits that reflect the state of the entire SVN tree?
It takes 8 GB of RAM and a lot of time to run, but that can be optimized fairly easily.
And incrementalized, by any chance?
Next step is to read in a corrected branches.txt file (this is currently generated based on hueristics by the 'branches' subcommand), and then use that information to output a branchified object hierarchy instead of a flat one. This step should be very easy to implement.
After that is reading Eric's manifest.txt file and using the information to produce multiple submodules during the repository conversion process. This step is quite a bit trickery, and will require a few days to get right.
Sounds like you're making great progress; keep it up! -- Dave Abrahams BoostPro Computing http://www.boostpro.com