[
https://issues.apache.org/jira/browse/LUCENE-6502?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
David Smiley updated LUCENE-6502:
---------------------------------
Attachment: LUCENE-6502.patch
Here's a patch. It improves the clarity of the class -- e.g. "minLaps" should
have been "maxLaps", etc. I reconsidered on demanding just 1 occurrence each
time, as it's really dependent on the number of laps. The problem is that in
rare cases with path & polygon based on Geo3D's random test shape generator as
currently written, it may be necessary to try many, many, attempts before
finding a "Within" predicate occurrence. So I bumped up the max laps to 200k
and for the case of Within, we'll just look for 1. We can always revisit the
200k figure in the future if it needs to be larger.
What do you think [~daddywri]?
> Spatial RectIntersectionTestHelper should only require one of each relation
> type to complete
> --------------------------------------------------------------------------------------------
>
> Key: LUCENE-6502
> URL: https://issues.apache.org/jira/browse/LUCENE-6502
> Project: Lucene - Core
> Issue Type: Improvement
> Components: modules/spatial
> Reporter: David Smiley
> Assignee: David Smiley
> Attachments: LUCENE-6502.patch
>
>
> The RectIntersectionTestHelper requires a minimum number of occurrences of
> each relation type before it passes, and there are a minimum number of
> attempts. But this can be a bit much, and too often it can cause a spurious
> test failure that isn't really a bug. Instead, it should simply try to find
> at least one of every case in a minimum number of tries.
> This would solve this bug today:
> http://jenkins.thetaphi.de/job/Lucene-Solr-trunk-Linux/12825/
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
