Jeff Barrett wrote:
> If I have a good sized working copy (11k files, 200mb) checked out on
> windows that has not had any svn operations performed on it for a while,
> it will take 5+ minutes to perform an "svn status" or "svnversion"
> However, once that 5+ minute invocation of "svn status" has completed,
> if I run it again it takes < 5 seconds.
>
> I'm assuming there is local caching on disk, am I right? How does it
> work? Are there things I can do to make it faster for SVN to
> create/update that cache? We've already done things like disabled
> Windows file-system indexing and turned off virus checkers.
>
> Thanks.
For details you should read the svn book, but yes, svn caches a pristine
copy of each file to compare your changes against when you do, for
instance, svn diff.
An svn status command must first check the timestamps on all the files
in the wc against the saved timestamps in the .svn directories. If you
have 20,000 files, fetching all of those timestamps can take quite some
time with a cold cache. I know of nothing that can tune windows/ntfs to
perform better with this kind of workload. Most filesystems perform
very poorly with large numbers of small files.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Wed Mar 14 21:14:42 2007