[ 
https://issues.apache.org/jira/browse/HADOOP-10325?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13891580#comment-13891580
 ] 

Andrew Wang commented on HADOOP-10325:
--------------------------------------

I looked at the previous and new proposed behavior for this patch, and I don't 
think either are really counting the number of Javadoc warnings. The old 
behavior looks at the full maven output, chops off everything after the first 
"Javadoc Warnings", and then counts anything with "[WARNING]" and "warning" in 
it. Unfortunately, that includes javac warnings as well. The new behavior 
doesn't bother with the chop, and thus results in counting more warnings, but 
still not the actual number of Javadoc warnings.

I tried this, and it seems to work more reliably:

{code}
cat patchJavadocWarnings.txt | egrep "^[0-9]+ warnings$" | awk '{sum+=$1} END 
{print sum}'
{code}

We could also be more sure by doing something like "egrep -A 1 ..." and making 
sure the same number of " # warnings" lines matches the number of lines with 
"[WARNING] Javadoc Warnings" (i.e., each "# warning" line has an immediately 
following "[WARNING] Javadoc Warnings").

> improve jenkins javadoc warnings from test-patch.sh
> ---------------------------------------------------
>
>                 Key: HADOOP-10325
>                 URL: https://issues.apache.org/jira/browse/HADOOP-10325
>             Project: Hadoop Common
>          Issue Type: Bug
>    Affects Versions: 3.0.0
>            Reporter: Colin Patrick McCabe
>            Assignee: Colin Patrick McCabe
>         Attachments: HADOOP-10325.001.patch, HADOOP-10325.002.patch
>
>
> Currently test-patch.sh uses {{OK_JAVADOC_WARNINGS}} to know how many 
> warnings trunk is expected to have.  However, this is a fragile and difficult 
> to use system, since different build slaves may generate different numbers of 
> warnings (based on compiler revision, etc.).  Also, programmers must remember 
> to update {{OK_JAVADOC_WARNINGS}}, which they don't always.  Finally, there 
> is no easy way to find what the *new* javadoc warnings are in the huge pile 
> of warnings.
> We should change this to work the same way the javac warnings code does: to 
> simply build with and without the patch and do a diff.  The diff should be 
> saved for easy perusal.  We also should not complain about warnings being 
> removed.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Reply via email to