So, the commit in question as a diff comes out to about 4,200 lines:
[email protected] s018 /tmp/k2/kafka> svn diff -cr1156232  | wc -l
   4160

To me, regardless of how the diff was created, this disqualifies it
from being a trivial change, particularly since it was done via
automation, which tends to be the source of many unintended
consequences.  But of course, it's possible to make a huge change via
a single line, so defining trivial just by size alone is not
sufficient.  For me, 'trivial' changes are one-line spelling fixes,
quick commits to get the build compiling again due to a syntax error,
or adding a single license header to a file.

Re: reviewboard.  Having done hundreds of patch reviews in Apache and
now being required to use RB as a Hive contributor, I very much
dislike it.  It adds an extra level of paperwork to each patch,
clutters the JIRA with impossible-to-trace comments from rb, and can
inadvertently lead to the wrong patch being committed (see HIVE-2276
for an example).  We've reached a happy medium in Hadoop where, those
who would like a review via RB can open one a request, but it's not
required and the reviewers are free to just post their comments within
the JIRA itself.
-Jakob

On Wed, Aug 10, 2011 at 12:45 PM, Henry Saputra <[email protected]> wrote:
> Hi Guys,
>
> I just got "pinged" by Jakob about the RTC =)
>
> I am not sure how strict we want to impose this rule?
>
> What considered as trivial changes? I guess if we want to strictly
> impose this we should sign up to https://reviews.apache.org/ for Kafka
> to help review the patch and diff.
>
> We use the reviews site for Apache Shindig  and Gora.
>
> Thoughts?
>
> - Henry
>
> On Mon, Aug 1, 2011 at 4:26 PM, Jun Rao <[email protected]> wrote:
>> Hi, Everyone,
>>
>> I propose that we use the "review then commit" process for future changes in
>> Kafka. Other than trivial changes, all patches will be reviewed in JIRA. Any
>> objections to this?
>>
>> Thanks,
>>
>> Jun
>>
>

Reply via email to