time to switch from CVS to Git, was: Missing mailings from commits


On Thu, 24 Feb 2011, James Amundson wrote:

< On 02/24/2011 12:29 PM, Leo Butler wrote:
< 
< >   You can see that cvs2git does a rather bad job in
< >   comparison to git-cvsimport: the diff file sizes
< >   are
< > 
< >                    Release 5.22        5.23
< >   git cvsimport    3.3K                3.3K
< >   cvs2git          363K                691K
< > 
< >   I've not bothered with more testing on earlier branches.
< >   Note that cvs2git doesn't like $Id$ tags, and it has lost
< >   files (including in the Attic).
< 
< OK, it looks like I made the wrong choice in resolving attic problems. Let's
< try this version:
< 
< 	http://amundson-long.net/maximav2.git.tar.bz2
< 
< Some notes:
< 	1) We should ignore special RCS comments, e.g.,
< 		-# $Id: texi2html,v 1.5 2007/04/14 21:56:43 robert_dodier Exp
< $
< 		+# $Id$
< 	    They don't have any real meaning in the git repository.
< 
< 	2) Since I'm using a cvs snapshot from Ray, we should be comparing
< against that, not the current CVS. I put a copy of the snapshot here:
< 
< 	http://amundson-long.net/maxima-cvs-2011-15.tar.bz2
 

 Jim, I redid the tests using your git repo, and ignoring diffs
 caused by:
 -RCS-type headers in all files;
 -case changes in .li?sp , .el files
 the results look quite good.

 You can find the results at

 http://www.maths.ed.ac.uk/~lbutler/cvs2git-v2/

 I put the script used to do the tests in there, too.
 I note though, that the tags+branches of each are not
 exactly the same.

 I didn't understand your comment (2), because I wanted to 
 check how well the git repo's tags+branches compared to the
 cvs repo's and I (think I) need access to the cvs repo
 for that, not just a snapshot.

 Leo

-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.