From: Andreas H. <haf...@in...> - 2010-12-29 10:44:54
|
> On 2010-12-29 11:43:19.015177, Andreas Haferburg wrote: > > I suggest you commit the fix in this form, before our next release. > > > > Could you give a concrete example of what happens when the bug occurs? I don't quite understand yet what I did wrong. Why was it so easy to break? > > > > Also, I think that the class SarosSession is not the place where something like OT should be able to break. We should consider moving this code somewhere else, maybe to the ActivitySequencer? SarosSession should be about participants, roles, and the shared projects, not OT. I forgot: Good catch! It would have taken me ages to fix that one, good job. - Andreas ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: http://saros-build.imp.fu-berlin.de/reviews/r/162/#review371 ----------------------------------------------------------- On 2010-12-25 21:17:37.771388, Michael wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://saros-build.imp.fu-berlin.de/reviews/r/162/ > ----------------------------------------------------------- > > (Updated 2010-12-25 21:17:37.771388) > > > Review request for All Saros. > > > Summary > ------- > > This patch fixes the broken operational transformation. The problem was introduced by a new dispatcher in the SarosSession for the SVN operations. As a consequence a bunch of received activities was not executed consecutively anymore in the SWT thread and local actions could be executed in between - milliseconds after the transformation but before the execution of the transformation. That's why the problem worsened in case of high load. > > Merry Christmas!!!!! > > [FIX] #3098992 - Inconsistency when concurrently writing at same position > > > Diffs > ----- > > /src/de/fu_berlin/inf/dpp/project/internal/SarosSession.java 2866 > > Diff: http://saros-build.imp.fu-berlin.de/reviews/r/162/diff > > > Testing > ------- > > > Thanks, > > Michael > > |