On Wed, Oct 13, 2010 at 07:34:56PM -0000, cmpilato_at_apache.org wrote:
> Author: cmpilato
> Date: Wed Oct 13 19:34:55 2010
> New Revision: 1022255
> URL: http://svn.apache.org/viewvc?rev=1022255&view=rev
> Propose (without yet adding my own strong vote) the backport of
> Modified: subversion/branches/1.6.x/STATUS
> URL: http://svn.apache.org/viewvc/subversion/branches/1.6.x/STATUS?rev=1022255&r1=1022254&r2=1022255&view=diff
> --- subversion/branches/1.6.x/STATUS (original)
> +++ subversion/branches/1.6.x/STATUS Wed Oct 13 19:34:55 2010
> @@ -174,6 +174,25 @@ Candidate changes:
> +1: cmpilato
> + * ^/subversion/1.6.x-no-wcng-check
> + Remove the function subversion/libsvn_wc/questions.c:is_inside_wc_ng()
> + and its uses.
> + Justification:
> + First, per issue #3729, the performance cost of this check is
> + atrocious (as it seems to happen many, many times in the course
> + of a checkout) and, according to some CollabNet customers,
> + unbearable. Secondly, the workaround stsp added to allow the
> + 1.6.x test suite to run inside a trunk-managed working copy tells
> + us that it's not necessarily the case that creating a 1.6 working
> + copy below some 1.7+ wcroot is problematic.
I added that workaround so I could manage a working copy of our 1.6.x
branch with a trunk svn client. It was impossible to run the tests
> It can be concluded,
> + then, that we might just be trying to stop a problem that doesn't
> + exist, and making every 1.6.6+ user pay a high cost for our
> + preventative efforts. That's bad.
I think the only reason for this check was to prevent users from
accidentally creating 1.6 working copies within a 1.7 working copy.
There is value in such a check, but it's not worth the performance
overhead you describe.
Received on 2010-10-13 22:22:18 CEST