[
https://issues.apache.org/jira/browse/LUCENE-5273?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Steve Rowe updated LUCENE-5273:
-------------------------------
Attachment: LUCENE-5273-slowdown-fixed.patch
Patch addressing slowdown: the old war-building behavior of pulling
{{lucene-*.jar}} from {{lucene/build/*/}} is restored, unless called from the
{{create-package}} target, in which case they are pulled from the
built/unpacked {{lucene-<version>.tgz}}. In this way, {{ant example}} doesn't
have to wait around for javadocs.
Here are timings from my Macbook Pro (OS X 10.8.5) with Oracle Java 1.7.0_25
64-bit - times are *minutes:seconds*:
||Top-level clean?||trunk before any LUCENE-5273 patches committed
(r1531353)||current trunk||trunk with this patch applied||
|Yes|1:11|4:26|1:15|
|No|0:37|0:38|0:37|
So with this patch, timings are back to where they were before I committed
anything for this issue.
This patch provides the additional benefit that Lucene module changes will get
picked up even if {{ant clean}} is not performed first.
Committing shortly.
> Binary artifacts in Lucene and Solr convenience binary distributions
> accompanying a release, including on Maven Central, should be identical
> across all distributions
> ---------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: LUCENE-5273
> URL: https://issues.apache.org/jira/browse/LUCENE-5273
> Project: Lucene - Core
> Issue Type: Bug
> Components: general/build
> Reporter: Steve Rowe
> Assignee: Steve Rowe
> Fix For: 4.6, 5.0
>
> Attachments: LUCENE-5273-optimized.patch, LUCENE-5273.patch,
> LUCENE-5273-slowdown-fixed.patch
>
>
> As mentioned in various issues (e.g. LUCENE-3655, LUCENE-3885, SOLR-4766), we
> release multiple versions of the same artifact: binary Maven artifacts are
> not identical to the ones in the Lucene and Solr binary distributions, and
> the Lucene jars in the Solr binary distribution, including within the war,
> are not identical to the ones in the Lucene binary distribution. This is bad.
> It's (probably always?) not horribly bad, since the differences all appear to
> be caused by the build re-creating manifests and re-building jars and the
> Solr war from their constituents at various points in the release build
> process; as a result, manifest timestamp attributes, as well as archive
> metadata (at least constituent timestamps, maybe other things?), differ each
> time a jar is rebuilt.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
