Re: Playing with a sandbox repository

Greg Ward <gward@mems-exchange.org> writes:

> I want to do some playing around with a sandbox repository -- ie. I want
> to do the same operation (or sequence of operations) several times,
> starting from the same not-quite-blank slate each time. Seems like the
> obvious thing to do is keep a backup copy of the repository somewhere
> handy (but not seen by Apache or Subversion), and then untar or "cp -rp"
> it into place each time I restart.

Or you could use svnadmin dump/load. If you are dealing with a large
repository then 'cp -rp' is probably fastest, particulary if you skip
the log files that can be archived.

> Are there any caveats I should be aware of? Eg. should I shutdown
> Apache before replacing the sandbox, and restart it afterwards? Or just
> restart it after replacing the sandbox? Or can I be as lazy and
> careless as if this were a CVS repository?

The client/cmdline regression tests did have a problem at one stage.
Before switching to using dump/load the tests simply copied
repositories. The problem arose when a repository was populated using
'svn import' and was then immediately copied. It was most easily
triggered using ra_svn; I believe the copy started before the svnserve
process had flushed everything to disk, so effectively the repository
changed during the copy. This problem probably won't affect you as it
doesn't sound as if your "backup" repository will change.

As to whether shutting down Apache is necessary: it probably safest to
do so, but I don't do it in my testing :)

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org