[
https://issues.apache.org/jira/browse/LUCENE-9819?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17293212#comment-17293212
]
Michael Sokolov commented on LUCENE-9819:
-----------------------------------------
The problem arises when the graph builder's `maxConn` is at its minimum value
of 2, which we sometimes randomly choose in the test:
MIN = 2;
{{if (random().nextBoolean()) {}}
{{ maxConn = random().nextInt(256) + MIN;}}
{{ }}}
In that case we can get a graph where some nodes are unreachable from any other
node. Setting MIN to 3 avoids that case. With that setting, I was able to run
1000 iterations without a failure. I'll push a fix.
> Look into TestKnnGraph random failures
> --------------------------------------
>
> Key: LUCENE-9819
> URL: https://issues.apache.org/jira/browse/LUCENE-9819
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Robert Muir
> Priority: Major
>
> If you apply enough pressure to this test it will provoke failures. Let's
> beast it from some specific git hash and figure out what is happening.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
