[
https://issues.apache.org/jira/browse/BIGTOP-39?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13712724#comment-13712724
]
Sean Mackrory edited comment on BIGTOP-39 at 7/18/13 8:06 PM:
--------------------------------------------------------------
Like Bruno's attachment, this is *not* something I want to propose for a commit
just yet. It has some known flaws but I think it's a good starting point to
packaging Avro. Some points to consider:
* Fedora already has a package named "avro", so I think we should avoid using
that name. On a similar note, should we generally isolate our package namespace
from that of the distribution? I don't want to prefix everything with "bigtop-"
but it seems that as Linux distributions ship their own Hadoop packages (that
may or may not be based on Bigtop's), we're going to eventually have some
collisions...
* I've done very basic testing that this installs "avro-tools" for use on the
command-line.
* I've also included all the other JARs and other language libraries. Which
bindings do we want to include? Another JIRA (BIGTOP-898) recently requested
including avro-mapred in Hadoop's classpath, and it was suggested maybe it's
time we packaged Avro. Without harmonizing the Avro version across components,
it seems like this isn't the solution either... Thoughts?
* The paranemer maven plugin was being troublesome and I"m not very familiar
with it. I just disabled it with a patch for now. One of the things I intend to
fix before seriously submitting this for review...
* The C/C++ libraries are also troublesome in that they require a newer version
of CMake than some of our supported distributions have, and there's /usr/local
is affected by some Debian packaging policies that I'm not yet familiar with.
was (Author: mackrorysd):
Like Bruno's attachment, this is *not* something I want to propose for a
commit just yet. It has some known flaws but I think it's a good starting point
to packaging Avro. Some points to consider:
* Fedora already has a package named "avro", so I think we should avoid using
that name. On a similar note, should we generally isolate our package namespace
from that of the distribution? I don't want to prefix everything with "bigtop-"
but it seems that as Linux distributions ship their own Hadoop packages (that
may or may not be based on Bigtop's), we're going to eventually have some
collisions...
* I've done very basic testing that this installs "avro-tools" for use on the
command-line.
* I've also included all the other JARs and other language libraries. Which
bindings do we want to include? Another JIRA recently requested including
avro-mapred in Hadoop's classpath, and it was suggested maybe it's time we
packaged Avro. Without harmonizing the Avro version across components, it seems
like this isn't the solution either... Thoughts?
* The paranemer maven plugin was being troublesome and I"m not very familiar
with it. I just disabled it with a patch for now. One of the things I intend to
fix before seriously submitting this for review...
* The C/C++ libraries are also troublesome in that they require a newer version
of CMake than some of our supported distributions have, and there's /usr/local
is affected by some Debian packaging policies that I'm not yet familiar with.
> Include Apache Avro in Bigtop
> -----------------------------
>
> Key: BIGTOP-39
> URL: https://issues.apache.org/jira/browse/BIGTOP-39
> Project: Bigtop
> Issue Type: Sub-task
> Components: Debian, RPM
> Reporter: Jeff Hammerbacher
> Assignee: Peter Linnell
> Attachments: 0001-Avro-packaging.patch, avro.spec
>
>
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
