Hi Kumar,
It's hard to use preview features in javac/javadoc because of the
bootstrap N-1 problem,
but yes, DocCommentInfo could be a record; also Utils.Pair is only used
in option decoding
for `-linkoffline` and could/should be moved there.
I'd also like the review the use of 3 separate maps to get doc comment
info from elements:
there's one down in com.sun.source world, one in CommentUtils and
another in Utils.
-- Jon
On 6/9/20 1:43 PM, Kumar Srinivasan wrote:
Hi Jon,
Looks good.
Out of curiosity, is it worthwhile to replace DocCommentInfo with
Record ?
There may be other places where these Pair-like containers were
introduced in the doclet and javadoc tool.
Kumar
On Jun 8, 2020, at 9:48 PM, Jonathan Gibbons
<jonathan.gibb...@oracle.com <mailto:jonathan.gibb...@oracle.com>> wrote:
Please review a partial fix for this issue, which is about different
doclint results
when invoking doclint from javac and javadoc.
The reason for the partial fix is explained here:
https://bugs.openjdk.java.net/browse/JDK-8242607?focusedCommentId=14344614&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14344614
Within the patch, the primary change is the restructuring of the
if-statement at about line 2706
in Utils.java. Specifically, the call
configuration.workArounds.runDocLint(path);
is moved outside of the test
if (docCommentTree != null
which is the root cause of the error.
At the same time, the patch includes a minor cleanup to rename the
`DocCommentDuo`
class to a slightly more abstract name `DocCommentInfo`. The old
`DocCommentDuo`
inappropriately reflects some of the historical origins of the name.
The change percolates
into some local variable names and member names as well. I also
tried to cleanup some
of the comments in this area, without changing functionality. The net
result is a
perceived need for more cleanup, including functionality cleanup, in
another changeset.
-- Jon
JBS: https://bugs.openjdk.java.net/browse/JDK-8242607
Webrev: http://cr.openjdk.java.net/~jjg/8242607/webrev.00/
