使用“git svn clone”从 sourceforge 克隆一个大型 svn 存储库

Question

I'm trying to clone a large subversion repository from sourceforge using git svn clone . The cloning process periodically gets stuck: SF just stops sending data. Eventually it times out, but it takes forever . After this, it's a cycle of:

• Run git svn fetch
• Wait for it to timeout
• Repeat as necessary

Looking at the git svn man page, there doesn't appear to be an obvious way to set a timeout for i/o operations.

Is there any way to make this process more efficient?

Answer 1

Let's me give you some advice on speeding things up. I just finished running them myself so I'm sure they work. The main trick is to copy Subversion repository to the local filesystem and then run git svn .

First, SourceForge allows to backup repositories using rsync (see the last command at the very bottom). So start with

rsync -ahPv svn.code.sf.net::p/vice-emu/code vice-emu.svn

In a general case where rsync access is not available you can use svnrdump to dump the repository:

svnrdump dump --non-interactive https://svn.code.sf.net/p/vice-emu/code > vice-emu.svndump

and then load the dump into a local repository:

svnadmin create vice-emu.svn
svnadmin load vice-emu.svn < vice-emu.svndump

Either way you can use git svn on the local repository:

git svn clone file://`pwd`/vice-emu.svn vice-emu

And finally yet another trick: offload the import to Github .

My results are: https://github.com/phdru/vice-emu-ghi (Github Importer) and https://github.com/phdru/vice-emu-gitsvn ( rsync + git svn + postprocessing).

PS. If you gonna fork one or both of my results please ping me afterward and I'll remove my temporary repositories. Else I'll remove then soon.

Answer 2

You can resume the process at the revision that you saw that it was able to pull fine (or say, 100 revisions before that) with -r

git svn fetch -r some-revision:HEAD

That way you don't restart from scratch.