Partial metadata rebuilding For highly active subversion repositories, it can be excruciatingly slow to pull updates one at a time from subversion. One way around this is to setup another mercurial repo that pulls new commits from svn periodicly (say every 5 minutes). When you want to update your repository, you can pull commits from this mercurial repository via native mercurial protocols, which will be much faster than pulling directly from svn. Unfortunately, your metadata will be out of date after doing so. Highly active repositories also tend to be very large, which means that it takes a long time to rebuild your metadata from scratch. To address this, this adds support to do a partial rebuild on the metadata by processing only revisions that have been added to the repository after the last revision we processed. With the rev map 1k revisions (~2 days) behind tip updatemeta is dramatically faster than rebuild meta: $ hg --time svn updatemeta Time: real 0.570 secs (user 0.480+0.000 sys 0.060+0.000) $ hg --time svn rebuildmeta Time: real 129.160 secs (user 128.570+0.000 sys 0.320+0.000)

#!/bin/sh
#
# Generate filecase.svndump
# WARNING: this script must be run on a case-sensitive file system
#

mkdir temp
cd temp

mkdir project-orig
cd project-orig
mkdir trunk
cd ..

svnadmin create testrepo
svnurl=file://`pwd`/testrepo
svn import project-orig $svnurl -m "init project"

svn co $svnurl project
cd project/trunk
# Test files and directories differing in case only
echo a > a
echo A > A
echo b > b
mkdir d
echo a > d/a
mkdir D
echo a > D/a
mkdir e
echo a > e/a
mkdir f
echo a > f/a
echo F > F
svn add a A b d D e f F
svn ci -m 'add files'
# Rename files and directories, changing only their case
svn mv b B
svn mv d/a d/A
svn mv e E
svn ci -m 'change case'
cd ../..

svnadmin dump testrepo > ../filecase.svndump