[
https://issues.apache.org/jira/browse/LUCENE-10353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17469220#comment-17469220
]
Robert Muir commented on LUCENE-10353:
--------------------------------------
+1, this sounds good to me. Then we can add missing {{Objects.requireNonNull}}
to all the buggy places until the test is happy again.
I agree with adding NPE to the allowed list from the ctor. There might also be
some cases where we throw IllegalArgumentException when it should really be
NullPointerException. We can fix these too, NPEs have nice error messages these
days. There is no excuse anymore, even for people who lose the stacktraces :)
> Add null injection to analyzer integration tests (e.g. TestRandomChains)
> ------------------------------------------------------------------------
>
> Key: LUCENE-10353
> URL: https://issues.apache.org/jira/browse/LUCENE-10353
> Project: Lucene - Core
> Issue Type: Task
> Reporter: Robert Muir
> Assignee: Uwe Schindler
> Priority: Major
>
> These tests inject random parameter values (from argumentProviders). Some
> generated values may be illegal and IllegalArgumentException is "allowed" if
> the constructor returns it. None of the values should cause failures at
> runtime.
> But for object types, we never inject null values (unless the
> argumentProvider were to do it itself). We should do this some low % of the
> time, and "allow" ctors to return NPE too.
> I see bugs in some of the analyzers where they are just a missing null check
> in the constructor. It is important to fail on invalid configuration up-front
> in the ctor, rather than failing e.g. at index time.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
