Issue #13553: False positive in FallThroughCheck on last case

[Lmh-java]

Fix for Issue #13553, Continue of #14016

Config

<module name="FallThrough">
    <property name="checkLastCaseGroup" value="true"/>
</module>

Regression Report

https://checkstyle-diff-reports.s3.us-east-2.amazonaws.com/8a0a354_2024060435/reports/diff/checkstyle/index.html
5 violations are removed in Checkstyle code base. Differences in this report are intended.

For example:

checkstyle/src/test/resources/com/puppycrawl/tools/checkstyle/checks/coding/fallthrough/InputFallThrough4.java

Line 109 in 4fe8118

void method5(int i, int j, boolean cond) {

void method5(int i, int j, boolean cond) {
      while (true) {
          switch (i){
          case 5: // violation 'Fall\ through from the last branch of the switch statement'
              i++;
              /* block */ /* fallthru */ // comment
          }
      }
   }

This is no longer a violation in this case. Same for the other 4 cases.

Diff Regression config: https://gist.githubusercontent.com/Lmh-java/55bdd9e289522aff23374ad7b5571c2c/raw/4e52c69d3888eff5f976f0ae87c058e98c0dd72d/input-fall-thru-config.xml

[Lmh-java]

GitHub, generate report

[Lmh-java]

Refactor hasReliefComment

[Lmh-java]

Remove the sentence in the doc as required

[rnveach]

Update in doc to remove:

The comment containing these words must be all on one line

I don't remember this conversation regarding this but does this mean if I have a suppression word as "suppress me", it will match things like "suppress \n me"? Do you have a test example showing this?

[rnveach]

Your updated documentation says:

...on the same line before the ...

How is that different then ...must be all on one line...?

[Lmh-java]

Update in doc to remove:

The comment containing these words must be all on one line

I don't remember this conversation regarding this but does this mean if I have a suppression word as "suppress me", it will match things like "suppress \n me"? Do you have a test example showing this?

No, I don't think "suppress \n me" will match.
This is matching by the pattern

    private Pattern reliefPattern = Pattern.compile("falls?[ -]?thr(u|ough)");

[Lmh-java]

Your updated documentation says:

...on the same line before the ...

How is that different then ...must be all on one line...?

I didn't see the conversion about this part either, but I think "must be all on one line" is redundant and confusing. I don't really understand what it means. The subject of this sentence is "the comment containing these words". However, we only need one "comment" containing these words to suppress this check for that branch.

If we remove this "must be all on one line.", and updated it to

The comment containing these words must be on the last non-empty line
before the {@code case} triggering the warning or on the same line before
the {@code case}(ugly, but possible).

I think this might be able to explain everything?

[rnveach]

It looks like comment in issue came from #12966 (comment) . This was about "the last comment on that line". This is why the issue has /* block */ /* fallthru */ // comment.

If you agree, please restore original wording and add some wording that it can be any comment on that line.

Can you point to keywords where the suppression text is not the last comment on the line. If we don't have one, then create it and point to it.

Example:
/* trigger */ /* comment */ // last
/* comment */ /* fallthru */ // last

[github-actions]

https://checkstyle-diff-reports.s3.us-east-2.amazonaws.com/8a0a354_2024060435/reports/diff/index.html

[Lmh-java]

@nrmancuso @romani please take a look.

[nrmancuso]

Rebasing on latest master, good to merge

[romani]

Ok to merge

[rnveach]

Edit: I am removing this. Property is additive, not either/or.

[Lmh-java]

GitHub, generate report

[github-actions]

https://checkstyle-diff-reports.s3.us-east-2.amazonaws.com/7e9dba3_2024170824/reports/diff/index.html