I think it's great to encourage people to use it if they feel the need or desire. Certainly for longer patches (new features especially) where you have a lot of comments to make, reviewboard is useful. However, everyone has their own workflow. For me, I always download every patch I review and inspect it in my editor, along with running a subset of related tests. If I then have comments to make about individual lines, I'll often upload it to RB to make the comments, but frequently the comments are either of a more general nature (this doesn't solve the bug it purports to address) or there are no comments necessary. Experience has shown me that trying to do good code reviews without actually looking at the patch in the context of my IDE is error
What I'm not clear on is why you want to make a rule that we must do this. Is there some larger problem you see that you think this would help us solve? Are reviews not transparent enough? Is the process too slow? Are we missing errors because we don't have good review tools? You seem to be proposing a solution to a problem that no one has complained about. Anyway, we should certainly update the wiki to explain how to use RB and when it is recommended to do so (such as when proposing a new feature). I think you should be able to do this, if you are willing. -----Original Message----- From: Thomas Koch [mailto:[email protected]] Sent: Thursday, September 22, 2011 5:23 AM To: [email protected] Subject: Make ReviewBoard obligatory? Hi, as Camille suggested[1], I've used ReviewBoard in the last weeks for a couple of issues. I believe it's a very good tool and helps a lot. Actually I ask myself, how one can do an effective code review without such a tool? It's kind of time-consuming to download the patch file, inspect it in an editor and post comments to jira, copy and pasting code lines or typing line numbers. What do you think? Would it be good to strongly encourage the use of ReviewBoard for every change whose patch file is longer then ~30 lines? I also think, that the current process of using ReviewBoard is time-consuming. But if that should be the reason to reject a review tool, then you might have a look to my suggestion of using Gerrit at the ASF[2]. I scanned the wiki[3][4] and didn't find ReviewBoard mentioned. ZOOKEEPER-1172 is an example of an (I believe) new contributor, who didn't know about ReviewBoard and also didn't correctly fill the ReviewRequest. I believe that the review process could become easier for the committers, if people would default to open review requests. [1] http://comments.gmane.org/gmane.comp.java.hadoop.zookeeper.devel/10095 [2] http://permalink.gmane.org/gmane.comp.apache.infrastructure.devel/1361 [3] https://cwiki.apache.org/confluence/display/ZOOKEEPER/HowToContribute [4] https://cwiki.apache.org/confluence/display/ZOOKEEPER/Committing+changes Regards, Thomas Koch, http://www.koch.ro
