[SPARK-4193][BUILD] Disable doclint in Java 8 to prevent from build error.

[ueshin]

No description provided.

[SparkQA]

Test build #22727 has started for PR 3058 at commit b548017.

• This patch merges cleanly.

[SparkQA]

Test build #22727 has finished for PR 3058 at commit b548017.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/22727/
Test PASSed.

[srowen]

I think we will ultimately have to turn this on, since javadoc 8 generates so much noise, but:

• Are there javadoc errors that are fixable? I have tried to clean up a lot of them
• Can we just disable warnings to cut the noise? or disable just some of the warnings instead of using :none?
• Can we make errors not fail the build?

[ueshin]

@srowen, Thank you for your comment.

The answers are:

• Yes, there are 3 errors and 63 warnings and the errors, which are usage of HTML entities issues, are fixable and the warnings, which are mainly missing of Javadoc annotations like @param or @return, are fixable if developers write Javadocs correctly.
• The checks are grouped and we can choose which groups should be checked, e.g. if we want to omit the missing issue mentioned above, we should use -Xdoclint:-missing insteand of -Xdoclint:none. The details are here. We have to make policies of Javadoc to choose the groups.
• Yes, add configuration <failOnError>false</failOnError> to maven-javadoc-plugin instead of <additionalparam>-Xdoclint:none</additionalparam>, or combination of them if we choose check groups.

[srowen]

So instead, how about fixing the errors? then the build works, no? the warnings are annoying though, but can be disabled with just -missing then? Fixing errors and quieting this kind of warning seem better than completely turning off doclint.

[ueshin]

Ah, I agree with you.
I could fix the errors and the build works now. And the warnings currently annoying go away with -missing.

[SparkQA]

Test build #22877 has started for PR 3058 at commit 923e2f0.

• This patch merges cleanly.

[SparkQA]

Test build #22877 has finished for PR 3058 at commit 923e2f0.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/22877/
Test PASSed.

[srowen]

@ueshin LGTM

[SparkQA]

Test build #23873 has started for PR 3058 at commit 6762ec2.

• This patch merges cleanly.

[SparkQA]

Test build #23873 has finished for PR 3058 at commit 6762ec2.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/23873/
Test PASSed.

[srowen]

@ueshin One last tiny thing. I think this plugin should be declared in <pluginManagement>, and should add <version>2.10.1</version>, but would not have <configuration> of course. Then this is left exactly as-is.

All -- I think this may help the problem in SPARK-4543? It is good to commit this fix anyway but may also simply silence Java 8's warnings on missing javadoc as desired.

[ueshin]

@srowen Thank you for your suggestion and I agree with it. I'll add it to the <pluginManagement>.

[SparkQA]

Test build #23889 has started for PR 3058 at commit e096bb1.

• This patch merges cleanly.

[SparkQA]

Test build #23889 has finished for PR 3058 at commit e096bb1.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/23889/
Test PASSed.

[pwendell]

This LGTM - thanks for the fix.