Branko Čibej wrote:
> Ben Collins-Sussman wrote:
>> BTW, I believe that SVK's patch format is basically 'svn diff' plus a
>> big binary hunk thrown at the end, which is just perl's serialization
>> of an editor drive. (Is that right?)
>
> I keep wondering if that (editor drive) dives enough context to perform
> fuzzy tree modifications (the same way unidiff has enough context for
> fuzzy patching), and if something like that is even feasible without all
> history being available.
The patch will typically be generated with reference to the contents of
some SVN repository. And typically the repository will also be
available when the patch is applied. So why not record the patch's
reference version within the patch (to the extent it is known and
well-defined) and have the smart patching tool get all of the history
that it needs from the repository? It could of course fall back to the
amnesic algorithm if the repository is not available.
(This technique could be used whether or not the patch is represented as
an editor drive.)
Michael
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Apr 1 19:49:08 2007