[
https://issues.apache.org/jira/browse/LUCENE-8281?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16456017#comment-16456017
]
Karl Wright commented on LUCENE-8281:
-------------------------------------
[~ivera], ok good.
The test case I made for the ticket originally was resolved by removing those
same two lines of code you removed earlier. My debugging patch includes that
removal, and it looks like you applied it, so the original test case now passes
-- because it can use test point 1, and doesn't need to fall back to test point
2.
Basically, the problem is that test point 2's in-set determination is
unreliable, and I don't understand why. I think if we can analyze the
identified failure (testLUCENE8276_case1) we should be able to solve the
problem. This is where things stand:
(1) IF we assume that the individual Edge XYZBounds are computed correctly,
then we are seeing the right Edge objects are showing up in the iterator for
each travel leg.
(2) We still get a single intersection going from test point 1 -> test point 2,
and no intersections going from test point 2 -> test point 1. So we are
(possibly?) missing an intersection we should find on the test point 2 -> test
point 1 traverse. I really don't know whether Test Point 2 should be in-set or
not -- maybe you can figure that out?
(3) The edge endpoints are far enough from all travel planes so that it does
not appear to be a case of a simple precision error.
(4) All potential intersections that get rejected, get rejected because of
bounds. I'm working to figure out if any of these rejections is close enough
that conceivably it shouldn't be rejected.
> Random polygon test failures
> ----------------------------
>
> Key: LUCENE-8281
> URL: https://issues.apache.org/jira/browse/LUCENE-8281
> Project: Lucene - Core
> Issue Type: Bug
> Components: modules/spatial3d
> Reporter: Karl Wright
> Assignee: Karl Wright
> Priority: Major
> Attachments: LUCENE-8281-debugging.patch, LUCENE-8281.jpg
>
>
> Reproduce here:
> {code}
> ant test -Dtestcase=RandomGeoPolygonTest
> -Dtests.method=testCompareSmallPolygons -Dtests.seed=42573983280EE568
> -Dtests.multiplier=3 -Dtests.slow=true -Dtests.locale=jmc-TZ
> -Dtests.timezone=US/Alaska -Dtests.asserts=true -Dtests.file.encoding=US-ASCII
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
