[
https://issues.apache.org/jira/browse/HADOOP-11790?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15397831#comment-15397831
]
Alan Burlison commented on HADOOP-11790:
----------------------------------------
I do have a recipe to get leveldb to build on Solaris, it needs a patch
applying to do so, and some hand-copying of .so files after it is built.
The real problem however is leveldbjni. Because it includes .so and DLL files
inside the JAR file, you need those libraries for *all* the platforms it can
run on. You also have to force it to statically link against both snappy and
leveldb. And there's no tarball containing the JNI code, you have to pull it
down with maven. And then of course it's hard-coded to use a specific autoconf
version, so you have to regen all that, and then you have to patch the POM so
it knows about the new platforms you are adding.
Basically, if you don't have access to *all* the platforms it's exceedingly
difficult to build. And even if you *do* have access to all the platforms, you
are in a world of hurt anyway, building and then copying libraries and JAR
files between platforms.
In my opinion, if there's any way leveldb can be removed as a dependency, it
should be removed. It's pretty much toxic if you need to either build from
ground up yourself, or are adding support for a new platform.
> leveldb usage should be disabled by default or smarter about platforms
> ----------------------------------------------------------------------
>
> Key: HADOOP-11790
> URL: https://issues.apache.org/jira/browse/HADOOP-11790
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 2.6.0
> Environment: * any non-x86
> * any OS that isn't Linux, OSX, Windows
> Reporter: Ayappan
> Priority: Critical
>
> The leveldbjni artifact in maven repository has been built for only x86
> architecture. Due to which some of the testcases are failing in PowerPC. The
> leveldbjni community has no plans to support other platforms [
> https://github.com/fusesource/leveldbjni/issues/54 ]. Right now , the
> approach is we need to locally built leveldbjni prior to running hadoop
> testcases. Pushing a PowerPC-specific leveldbjni artifact in central maven
> repository and making pom.xml to pickup it up while running in PowerPC is
> another option but i don't know whether this is a suitable one . Any other
> alternative/solution is there ?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
