On Nov 3, 2009, at 2:28 AM, Bert Huijben wrote:
>> -----Original Message-----
>> From: Hyrum K. Wright [mailto:
hyrum@...]
>> Sent: maandag 2 november 2009 21:28
>> To:
svn@...
>> Subject: svn commit: r40357 - trunk/subversion/libsvn_wc
>>
>> Author: hwright
>> Date: Mon Nov 2 12:28:21 2009
>> New Revision: 40357
>>
>> Log:
>> * subversion/libsvn_wc/props.c
>> (svn_wc__merge_props): Flush a couple of log accumulators.
>
> This function did not flush its own log accumulaters. It is always
> flushed
> by its callers. (Which use it in transactions).
>
> This change can install a conflict for a update/merge that is never
> written
> to the working copy (e.g. when the file interactive conflict handler
> cancels
> the operation).
>
> These operations should be moved up in the call chain to the final
> location
> where the db, the working file, the properties, the conflict and the
> dav
> properties are updated at once (all or nothing).
So, barring the time/desire/knowledge to immediately move this
operation up the call stack, should this just be reverted, then?
-Hyrum
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2414154
