Greg Stein <gstein@lyra.org> writes:
> On Sat, Jun 08, 2002 at 08:59:26AM -0500, Ben Collins-Sussman wrote:
> > anders@gnulinux.dk writes:
> >
> > > We checked our source-code in and began to play with the different svn
> > > options. svn diff -r x:y segmentation faults on our archive
> >
> > It was a client-side bug that I wrote, and just fixed in r2128.
> >
> > Go ahead and update/rebuild your client, it should work now.
>
> Sounds like we should also fix the client/server to *not* send the data on a
> simple diff.
Hmmm, not sure how we would do that. There is no "simple diff". :-)
'svn diff -r5 wcpath', which compares your working wcpath against r5
of it, is done by running an update under the hood. The working copy
is described to the server, and when parsing the response, we GET
tmpfiles and compare to working files.
'svn diff URL1 URL2' runs a switch under the hood (instead of
update), since the urls aren't necessarily related. This method is
also used for the 'svn diff -rX:Y [URL|wcpath]' case.
As far as mod_dav_svn is concerned, it only sends the "extra"
vsn-rsc-url data for a switch. That was a recent change.
Really, we're getting wcprops sent back to us *embedded* in the update
and switch responses anyway... we always have. It's just that
operations like 'svn diff' and 'svn st -u' deliberately set the
set_wcprop() callback to NULL, so the vsn-rsc-urls are ignored during
parsing.
Maybe we could add some custom request header that will tell
mod_dav_svn to leave out all the vsn-rsc-urls, both within the response
and in the resource-walk? Do we really care?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Jun 9 16:00:53 2002