Speedup and Caveats with Git cvsimport?

Question

hi all,

we're trying our hand at migrating from cvs to git.

Our goals are 1) perfect current codebase 2) usable history. If it's missing an entry in an old branch, we don't care so much.

Two comments:

1. Even with a local copy of the cvs repository, it takes a very long time. The dry run took over 24 hours (7.5 gb cvs codebase; P4 2.0 ghz machine, 2gb RAM, 10K drives. Git scrunched it down to 1.8 gb). Because CVS itself takes up 99% of the cpu, I imagine it's cpu bound. Any way to speed it up?

2. Stdout has a bunch of warnings. Which ones matter? Which ones don't?

WARNING: Invalid PatchSet 5763, Tag pre-merge-pe-2-3-merge-26: conf/peTEST.conf:1.4=after, src/java/com/participate/util/XSLUtilTEST.java:1.1=before. Treated as 'before' revision 1.167.2.11 of file derived/workflow/xml/bpd.xml is tagged but not present revision 1.106 of file derived/workflow/xml/bpd.xml is tagged but not present Branch PE-2-3 already exists! Warning: commit message does not conform to UTF-8.

thanks in advance

will

Answer 1

"git cvsimport" has many known bugs and should not be considered a reliable means of converting from CVS to git. I recommend cvs2svn (in particular its cvs2git mode), which is very flexible and will preserve all of your CVS history. I would use the trunk version of cvs2svn, as it has many improvements compared to the latest released version (2.3.0).