Migrations...

I've spent the last few weeks migrating from two servers to one. (Ok, it's really a consolidation.) Everything now sits on the Compaq W6000 workstation and it seems to be stable.

I've also begun some serious repository conversion from CVS to Subversion. Pretty slick, but I've had to write a few scripts to deal with oddities. I've opted not to create a separate subversion repository for each of my projects (way too much overhead there) and instead am just using a single repository. I figure if something grows too big (or needs to be shared) that I will simply dump the data out and then import it into the new repository.

The structure I've chosen has, at a high level, a partitioning scheme based on the type of thing being stored. For instance, I have java-devel, unix-devel, and server-config at this time. Under each of these is the typical trunk/tags/branches directories. Under that exists whatever structure makes sense. java-devel is organized as Eclipse proejcts, server-config is organized by servers (which is now down to one), and unix-devel has files just dumped into trunk. I don't do too much Unix development anymore. :-/

Anyway, the partitioning scheme seems to make cvs2svn barf with an error. I've opted instead to use cvs2svn to create a dumpfile which gets tweaked and then loaded in via svnadmin. Just incase it's useful to others, here's a current copy of the script:

#!/bin/bash

# Import CVS projects into new SVN repository in my weird structure...
#    /Grouping/trunk/Project/...
#              tags/Project/...
#              branches/Project/...

# Usage:
#    

GROUPING="$1"
PROJECT="$2"

if [ -z "$GROUPING" -o -z "$PROJECT" ]
then
   echo "Please include both the project grouping and the CVS project name."
   exit 1
fi

# Dump out the project...
/usr/local/cvs2svn/cvs2svn --dumpfile=/tmp/${PROJECT}.dump         --trunk=trunk/${PROJECT} --tags=tags/${PROJECT}         --branches=branches/${PROJECT} /home/cvsroot/${PROJECT}

# We need to drop the trunk/branches/tags directory creation as they are
# already present, hence the gawk script.
cat /tmp/${PROJECT}.dump | gawk '
BEGIN {
   ignore = 0;
}
{
   if ($0 == "Node-path: trunk" || $0 == "Node-path: branches" || $0 == "Node-path: tags") {
       ignore = 5;
   }
   if (ignore > 0) {
       ignore--;
   } else {
       print $0;
   }
}' | svnadmin load --parent-dir ${GROUPING} /home/svnroot/general