[
https://issues.apache.org/jira/browse/YETUS-922?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17052305#comment-17052305
]
Nick Dimiduk commented on YETUS-922:
------------------------------------
bq. This is literally the point of {{--tests-filter}}
[~aw] I disagree. {{--tests-filter}} says "I want to run these checks, but
regardless of the result, I want them to vote at worst -0." What I'm asking for
is the voting system to say, "hey, it's not perfect, but it's better than you
found it. Good work, +1." If the patch produces more total infractions than
were already present, I still want a -1, which {{--tests-filter}} denies me.
As I said in the description, one can refactor/restructure code without making
material changes to the implementation therein. The check-tool-output-diff
approach isn't smart enough to track lines through a history, so it's unable to
be certain if a given change is new to the codebase or not.
> Precommit qualitative checks vote be -0 on overall improvement
> --------------------------------------------------------------
>
> Key: YETUS-922
> URL: https://issues.apache.org/jira/browse/YETUS-922
> Project: Yetus
> Issue Type: Improvement
> Components: Precommit
> Reporter: Nick Dimiduk
> Priority: Minor
>
> Looking at the output over on
> [hbase/775|https://github.com/apache/hbase/pull/775#issuecomment-547725241],
> I think the -1 votes on qualitative checks are a bit harsh. The tests I'm
> looking at are {{javac}} and {checkstyle}}, where we have a qualitative
> measure of change in quality. In this case, the patch improved quality by
> reducing the overall number of failure occurrences. I think these should be
> voted as -0 rather than -1. I suspect the reasoning behind the -1 vote is
> that the patch is viewed to have introduced new failures. The thing is, with
> patches that refactor code, this simple diff isn't able to distinguish
> between an actual new failure and a moved failure.
> I could also argue that they should actually be +1 when {{total}} is less
> than {{previous}} because it's positive trajectory for the code base.
> {noformat}
> javac | hbase-server generated 1 new + 3 unchanged - 3 fixed = 4 total (was 6)
> checkstyle | hbase-server: The patch generated 12 new + 270 unchanged - 37
> fixed = 282 total (was 307)
> {noformat}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
