subversion-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Paul Burba <ptbu...@gmail.com>
Subject Re: 1.6.7 up for signing/testing
Date Mon, 04 Jan 2010 17:01:04 GMT
On Tue, Dec 29, 2009 at 1:20 PM, Peter Samuelson <peter@p12n.org> wrote:
>
> [Branko Cibej]
>> I thought we were supposed to be more tolerant of mis-fomatted
>> mergeinfo, not suddenly change the mergeinfo format? What am I missing?
>
> I thought it was both - parse incorrect mergeinfo, but when writing it
> out, fix it.  I think that's why the test expectations had to be
> changed.

Peter is correct.  With the backport of r889840, when we produce
svn_mergeinfo_t from a string representation, we tolerate relative
merge source paths, but convert those paths to absolute paths in the
resulting mergeinfo.  We also do the same when converting
svn_mergeinfo_t to a string representation.

However, r889840 has nothing to do with the failing Python test,
rather it is the fix for issue #3547 "'svnadmin load --parent-dir
PATH' can destroy mergeinfo" that caused the test to fail.  However,
the test was wrong, it was *expecting* incorrect behavior, see
http://svn.haxx.se/dev/archive-2010-01/0018.shtml for a full
explanation.

Paul

> --
> Peter Samuelson | org-tld!p12n!peter | http://p12n.org/
>

Mime
View raw message