hadoop-common-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Bieniosek <mich...@powerset.com>
Subject Re: jdk6 on darwin
Date Fri, 12 Oct 2007 19:00:56 GMT
Ah, okay.  I thought Owen was suggesting that we move to requiring java 1.6 for compilation
or runtime, but it appears it is still a little early for that in terms of availability of
java 1.6.


On 10/12/07 11:34 AM, "Doug Cutting" <cutting@apache.org> wrote:

Colin Evans wrote:
> I'm a bit confused by this discussion though.  How would compiling the
> jars with Java 1.5 and running on 1.6 degrade performance (assuming that
> the jars don't use any new 1.6 APIs)?

It won't.  The claim is just that running with Java 1.5 degrades
performance significantly.  The source code is compatible with Java 1.5,
and code compiled with Java 1.5 runs efficiently in Java 1.6.

So recompilation will make 0.14.2 work in Java 1.5 and Java 1.6.  This
has been the way we normally produce Hadoop releases: compile with Java
1.5 and recommend Java 1.6 as the runtime.  I made a mistake when
building 0.14.2, and used Java 1.6 to compile it, and no one noticed
until after the release was published.

The question is whether this warrants a 0.14.3 release, compiled with
Java 1.5.  I'm currently leaning that way.


  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message