[Kde-scm-interest] svn2git 209e6ce can corrupt history

[Pauli Virtanen]

Hi all,

I noticed that svn2git (from http://gitorius.org/svn2git) commit 

        209e6ce Try and branch from the correct svn revision rather then the last one.

can cause corrupt trees to appear in the repository history.

It occurs with this repository: http://svn.scipy.org/svn/numpy 
(svn dump here, if you want to play with it:
http://pav.iki.fi/tmp/numpy.svndump.gz)

With this config:

        create repository numpy
        end repository
        
        match /trunk/
            repository numpy
            branch master
        end match
        
        match /branches/([^/]+)/
            repository numpy
            branch branches/\1
        end match
        
        match /tags/([^/]+)/
          repository numpy
          branch svntags/\1
        end match
        
        match /([^/]+)/
            repository numpy
            branch svnroot/\1
        end match

and exhibits as

        $ git diff svntags/1.0.3..svntags/1.0.3.1|diffstat|tail -n1
         90 files changed, 5836 insertions(+), 2303 deletions(-)
        $ svn diff file://$PWD/numpy-svn/tags/1.0.3 file://$PWD/numpy-svn/tags/1.0.3.1 | diffstat
         numpy/distutils/misc_util.py |   30 ++++++++++++++++++++++++++++--
         numpy/version.py             |    2 +-
         setup.py                     |    7 +++----
         3 files changed, 32 insertions(+), 7 deletions(-)

For some reason, the tree corresponding to tag 1.0.3.1 is different in
the generated Git repository than what it is in SVN. The branch point
seems to be detected wrong, and for some reason this also affects the
content of the corresponding revision. Perhaps there are also other
discrepancies, but this I managed to spot.

With 209e6ce I see this behavior, but it works correctly with 209e6ce^
so the bug seems to be in 209e6ce.

Please let me know if you need other information.

Best regards,
Pauli Virtanen