[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Ordering of output for diff tests #29 and #31

From: Justin Erenkrantz <justin_at_erenkrantz.com>
Date: 2006-03-28 08:02:10 CEST

On 3/27/06, Peter Samuelson <peter@p12n.org> wrote:
>
> [Justin Erenkrantz]
> > Our regression tests for diff #29 (diff to BASE with local property
> > mods) and #31 (diff a property change plus a local prop edit) expect
> > a specific ordering that the client will return the diff in.
>
> I think deterministic ordering is important for readable diffs, and I
> prefer subdirs-first (that is, depth-first).

It is depth first. The question is whether we mandate pre-order or
post-order. =)

> Would it be too hard to
> make ra_serf do this? Forcing the client to reorder the diff in any
> way after the fact kills streaminess.

ra_serf keys its prop changes off the closing </*-directory> tag in
the server's XML response. ra_dav can cheat because it has the
property and text deltas in-line so it emits a diff as soon as it sees
those tags - not the closing tag of the directory or file. Also, the
delta editor requires that propchanges be the first or the last thing
done in a directory - since ra_serf doesn't necessarily have the
properties when we open the children (as we may have to issue a
PROPFIND), ra_serf needs to issue prop changes at close time. --
justin

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Mar 28 08:02:32 2006

This is an archived mail posted to the Subversion Dev mailing list.