Allen, I am not questioning whether checkstyle is a requirement. My concern is that some of the rules in there are arcane/unnecessary - we should use a better list of rules, in my opinion. This is why I filed HADOOP-11869. As pointed out by Jason, this is not a dupe of HADOOP-11866 <https://issues.apache.org/jira/browse/HADOOP-11866> unless scope is being increased.
thanks, -Sidharta On Fri, Apr 24, 2015 at 12:55 AM, Allen Wittenauer <a...@altiscale.com> wrote: > > > > On Apr 23, 2015, at 7:57 PM, Sidharta Seethana <sidharta.apa...@gmail.com> > wrote: > > > About (3.) , a lot of the check style rules seem to be > arcane/unnecessary. > > Please see : https://issues.apache.org/jira/browse/HADOOP-11869 > > > a) I've closed it as a dupe of HADOOP-11866 to keep everything in one > place. > > b) I've had HADOOP-11778 open for a while to update checkstyle to a more > modern version…. which will also likely fix HADOOP-11546. > > c) According to our commit guidelines, checkstyle is a requirement for > commitment. If we want to remove that requirement, we need to modify the > guidelines and comment out the registration line in the checkstyle.sh > plugin. We've been ignoring it for whatever reasons, likely because the > code in the old test-patch.sh was pretty broken to the point of being > disabled. > > > Personally, while some view this as a "minor formatting issue", it > reflects poorly on the project to have every file formatted differently. > check style is meant to enforce those rules. This *is* a quality check. > > Given that branch-2 went from "stable" (~2.4) to "beta" (~2.6) to > "alpha" (officially 2.7), well… I guess I shouldn't be surprised that > quality is going down though.
