2015-06-01 19:13 GMT+02:00 Kristian Rosenvold <kristian.rosenv...@gmail.com> :
> Re-running the clone from a backup of asf svn is time consuming but might > be the way to go, because we could probably get the correct layout in one > go. (But it's dooog slow and will be even worse multiplied by X) > Yes. That's indeed the way I chose to follow for Mojo because after twiddling a bit with filter-branch, I was quite sure it would be slower but a lot easier to just git svn clone mojo by mojo (hence the csv file, specifying the tags&branches for each mojo [1]). I also And I can certainly relate about the sluggishness. It took about 2 or 3 minutes even for very small mojos, and I had the whole SVN imported locally. In that current case, I've always seen svn repositories available at GitHub as automated facilities somehow only to be able to file PR against the svn trunk, but not as real Git clone that could be used for a migration. > Alternately one could probably get around the strange layout of the current > git-svn clone by doing some hardcore conditional filter-branching to > rewrite the root directory of only those commits that are not already at > the root. This is not your average git-svn rewrite and we're *way* beyond > powerpoint presentations here. > > But bottom line, it's just a dirty job that someone has to do :) > Yep. But even if I'm not even sure every required commits are here, I'm not sure that would be interesting to do it apart from learning a bit more black magic :). > > Kristian > [1] https://github.com/mojohaus/convert-to-git/blob/master/repo-infos.csv
