[
https://issues.apache.org/jira/browse/LUCENE-5954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14574480#comment-14574480
]
Michael McCandless commented on LUCENE-5954:
--------------------------------------------
bq. It seems weird to have two ways of comparing versions.
Good point, I'll remove the compareTo and just use the existing onOrAfter.
bq. Is there somewhere we could have a more direct test than deletion policy
tests?
I'll try to make a more direct test (TestSegmentInfos)...
bq. We dont technically need to change file formats to implement this, it could
be computed from the segments on read in the 4.0-5.2 case?
That's what I do in the patch ... if SIS was written by >= 5.3, I pull the min
version that we wrote into the header (and throw IndexTooOldExc if it's too
old), else I re-compute it on load.
> Store lucene version in segment_N
> ---------------------------------
>
> Key: LUCENE-5954
> URL: https://issues.apache.org/jira/browse/LUCENE-5954
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Ryan Ernst
> Assignee: Michael McCandless
> Attachments: LUCENE-5954.patch
>
>
> It would be nice to have the version of lucene that wrote segments_N, so that
> we can use this to determine which major version an index was written with
> (for upgrading across major versions). I think this could be squeezed in
> just after the segments_N header.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
