I’ve previously had an SVN repository filled with a mixture of projects that I’m now splitting out into separate repositories, so am in the process of creating various scripts based on “svnadmin dump”/”svnadmin load” to process repositories.
Along the way, I had a need to dump only selected revisions from a repository, using a command such as:
svnadmin dump -r 1234:HEAD /path/to/repo > repo.dump
However, when this is loaded back into a new repository (“svnadmin load mynewrepo < repo.dump”), the revisions are renumbered starting at 1, so that what was revision 1234 becomes revision 1. This is undesirable, as I have existing bugs and changelogs referring to SVN revision numbers, so I created a small script (svn-generate-empty-revisions) to create a number of empty revisions.
In use, it’s most useful to splice its output into the start of an SVN dump, for example:
svnadmin dump -r 1234:HEAD /path/to/repo > repo.dump # Extract the first few lines of the dump, which contain metadata head -n 4 repo.dump > repo-padded.dump # Splice in some empty "padding" revisions to preserve revision numbering # Note that the first revision is 1234, so we want 1233 empty revisions at start ./svn-generate-empty-revisions.sh 1233 >> repo-padded.dump # Add the rest of the original repository dump to the file tail -n +4 repo.dump >> repo-padded.dump