[Fedora-packaging] Packing Hadoop

[Patrice Dumas]

On Fri, Jan 16, 2009 at 09:07:50PM +0100, Fabian Deutsch wrote:
> Hello,
> 
> There are a couple of issues I ran into:
> - Can I explicitly require a java package not just to be installed, but
> also to be the the selected alternative?
> (openjdk ist required at build- and runtime)

I did that for tetex-tex4ht, copying what was done in freecol.

> - What about missing dependencies?
>   There are some jars required which are not packed up in Fedora yet.
> I'm not allowed to use the binary-jars distributed in the hadoop
> tarball?

Indeed you are not. They should be packaged in fedora.

> - Scripts
>   There are scripts to run hadoop and dfs (bin/start-all, stop-all, ...)
>   Where shall I put them? Create a custom user, as postgres does?

If hadoop runs as a daemon, it should not run as root but as a
special user. And you can do an initscript for fedora, to integrate well
with init daemon management. But since I have no ieda what hadoop is, 
I may be telling nonsensical stuff.

In any case, and especially for your first package and an hard package,
you can submit a review request with a non-perfect package, as long as 
you tell it clearly in a comment.

As a side note, in general this list should not be used for  
packaging questions, unless you have in mind a needed change in 
guidelines to accomodate your package. The development questions should 
better go to fedora-devel-list, since here it is more to discuss 
packaging guidelines than for help on existing guidelines.

--
Pat