Re: Patch ping, 5 years later: removing properties

Hi Alexey.

In your original email <http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=741864>, you described the problem this way. You have a tree imported into SVN from CVS, and another copy of it in which you have added Subversion properties such as svn:executable to lots of files and made a few other changes. Now you want to see a diff between the two trees, ignoring the properties to make it easier to focus on the other changes.

I can see why you want to ignore the properties in this case, but I don't think Subversion needs to do this for you, as it's not a general need. I think "svn diff | filterdiff --clean" is the right solution to your problem.

You wrote:
> > But this [filterdiff] is still unreliable: if property
> > text includes a diff-like chunk, it will go through.

That was true in Subversion 1.6, but in 1.7 a multi-line property is diffed line-by-line just like a multi-line file, so no output line will ever start with "@@", so "patch" and "filterdiff" will not be confused. Even with 1.6, this wasn't a problem in practice was it?

C. Michael Pilato wrote:
> Would you be interested in a new 'svn diff --patch' flavor,
> which generates only diff output that's suitable
> for consumption with GNU patch?  Implied by --patch
> would be no property diff output and the display of copied
> items as additions (like --show-copies-as-adds does).

That is potentially a good direction to go in the user interface, adding options focused on a specific high-level goal. With the current low-level options (--no-diff-deleted, --show-copies-as-adds, and --notice-ancestry), it can be rather hard for a user (myself included) to figure out what options need to be specified, if indeed it can be done at all, to give a particular result such as a fully two-way diff or a diff that's as compatible as possible with standard 'patch'.

- Julian

Alexey Neyman wrote:
> > What is the sentiment about implementing --no-property-diff
> > option to 'svn diff'? I can implement it if there's
> > agreement such option is needed.
Received on 2011-11-04 11:19:48 CET