Owen O'Malley wrote:
On Dec 12, 2007, at 1:36 PM, Andrzej Bialecki wrote:
Ted Dunning wrote:
Hadoop *normally* uses the Sun JDK. Using gcj successfully would be
a bit
of a surprise.
GCJ 4.2 does NOT work. With minor tweaks it's possible to compile all
Hadoop classes, including contrib, but it doesn't run properly. The
offending class is org.apache.hadoop.io.Text (CharacterEncoder works
differently from the Sun implementation, perhaps it's broken). This
class (Text) is widely used throughout Hadoop, so it won't work with
GCJ for now ...
If anyone knows of specific problems or workarounds, it would be great
to share them. I thought that gcj was still missing a lot of the the
java 1.5 libraries...
AFAIK few GUI applications run successfully - AWT / Swing support is
still shaky, but most other APIs are in good shape.
Actually, the piece I'd love to see working under gcj is the hdfs
client. I bet gcj would perform better that using jni in libhdfs.
I tried to find some info about this bug in GCJ 4.3, perhaps it's fixed
(I don't know what input caused this error, but there was some work done
on CharacterEncoder since 4.2 release).
--
Best regards,
Andrzej Bialecki <><
___. ___ ___ ___ _ _ __________________________________
[__ || __|__/|__||\/| Information Retrieval, Semantic Web
___|||__|| \| || | Embedded Unix, System Integration
http://www.sigram.com Contact: info at sigram dot com